基于图像特征选择的田间籽棉成熟度与品级判别技术研究
|
摘要
我国棉花生产发展目标是,棉花等级相符率不断提高,异性纤维问题、掺杂使假事件、棉包一致性差现象不断减少等。严把“采摘、收购、加工”关,减少异性纤维含量是解决这些问题的主要措施。本研究把现代机器视觉和模式识别等技术引入传统的采摘作业进行田间籽棉成熟度与品级的判别,为采摘机器人具备精确的作业能力奠定基础,以适应棉花品种的多样性、避免农业化学品引起的环境污染、降低劳动强度、减少农业投入成本等。取得的主要结论归纳如下：
1.将籽棉与背景视为二个类别,基于竞争学习网络分割了籽棉图像。在HSI、Lab、 Ohta、RGB颜色空间下,用带类别标签的训练样本计算k-均值聚类的误分类率,选取了误分类率最小的RGB颜色空间,其B值较优。在RGB颜色空间下,用带类别标签的训练样本的B值一次性地训练了竞争学习网络模型。结合形态学滤波,用所训练的竞争学习网络模型分割907张新图像的准确率达92.94%。该方法结合了有监督的学习算法,使图像分割在先验知识的指导下进行,同时,避免了K-均值聚类的反复迭代和过拟合现象,降低了计算开销,提高了图像分割的效率和精度。
2.基于频域提取了描述籽棉边界轮廓的频谱特征集。用8-邻域跟踪法获取了籽棉区域边界轮廓的坐标序列,基于傅里叶变换提取64个频谱特征,其相关系数都不超过0.9；单因子方差分析表明,大多数频谱特征在成熟和未成熟籽棉上的均值具有显著的差异。
基于空域提取了个描述籽棉形状的几何结构特征集。用一组同心园外接或切割籽棉区域,基于比值法提取了15个几何结构特征,它们是,全局特征、径向切割区域特征1-3、圆周向切割区域特征1～5、径向切割线特征1-2、圆周向切割线特征1-3以及计盒维数,其相关系数超过了0.9；单因子方差分析表明,大多数几何结构特征在成熟和未成熟籽棉上的均值具有显著的差异。
将籽棉是否成熟视为二类问题,基于过滤器模式从上述特征集中启发式搜索了特征子集,评估函数为类可分性准则。最优标量特征组合特征子集时,在训练集上计算每个特征的类可分性测量值并降序排列,将前l个(l=1、2、…、64或15)特征组合形成l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为14,该特征子集在预测集上的识别率为77.78%；几何结构特征集的最佳特征子集容量为8,该特征子集在预测集上的识别率为85.56%。浮动搜索特征子集时,在训练集上浮动搜索具有最大类可分性测量值的l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为22,该特征子集在预测集上的识别率为82.22%；几何结构特征集的最佳特征子集容量为8,该特征子集在预测集上的识别率为85.56%。
将籽棉是否成熟视为二类问题,基于封装器模式从上述特征集中穷举搜索了特征子集,评估函数为分类器的误分类率。在训练集上组合所有可能的l维(l=1、2、64或15)特征子集建立贝叶斯判别模型,并选择在验证集上具有最小误分类率的l维特征子集,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的最小误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为15,该特征子集在预测集上的识别率为83.33%；几何结构特征集的最佳特征子集容量为6,该特征子集在预测集上的识别率为88.89%。
3.基于相关分析选取了测量籽棉色泽的颜色空间。根据中国籽棉品级分级文字标准,在Hunter、Lab和HSI颜色空间下度量籽棉的色泽,相关分析表明,籽棉的色泽特征呈虚假相关；经灰度修正后,HSI颜色空间消除了虚假相关,可度量籽棉的色泽。
基于空域提取了描述籽棉品级的纹理、形状特征集。在HSI颜色空间下,基于亮度Ⅰ和饱和度S的直方图的均值、标准差、平滑度、三阶矩、一致性和熵,提取了12个纹理特征以描述籽棉的色泽、雨锈/轻霜/僵瓣/污染/烂桃等杂质；基于比值法提取了16个形状特征以描述籽棉的大小、几何结构分布,它们是,棉瓣个数、全局特征、径向切割区域特征1～3、圆周向切割区域特征1～5、径向切割线特征1～2、圆周向切割线特征1～3以及计盒维数；剔除相关系数超过0.9的纹理、形状特征后,基于特征集在1～7级样本上均值与品级值之间的回归分析,在0.05的显著性水平下选取了线性相关的有效纹理、形状特征集,它们是,亮度直方图的三阶矩、饱和度直方图的均值、标准差和熵、棉瓣个数、全局特征、圆周向切割区域特征(1,4)、径向切割线特征1、圆周向切割线特征3、计盒维数。
将成熟籽棉品级视为七类问题,基于10折交叉验证和过滤器模式从上述有效特征集中启发式搜索了特征子集。最优标量特征组合特征子集时,在每一个训练集上计算每个特征的类可分性测量值并降序排列,将前l个(l=1、2、…、11)特征组合形成l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率不断减小,10个对应的验证集上的平均误分类率不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为7,10个训练集在此处建立了10个模型,第5个训练集在此处建立的模型较好,它在预测集上的识别率为74.71%。浮动搜索特征子集时,在每一个训练集上搜索具有最大类可分性测量值的l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率不断减小,10个对应的验证集上的平均误分类率不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为6,10个训练集在此处建立了10个模型,第9个训练集建立的模型较好,它在预测集上的识别率为70.11%。
将成熟籽棉品级视为七类问题,基于10折交叉验证和封装器模式从上述有效特征集中穷举搜索了特征子集。在每一个训练集上组合所有可能的l维特征子集建立贝叶斯判别模型,并选择在对应验证集上具有最小误分类率的l维特征子集(l=1、2、11),随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率逐渐减小,10个对应的验证集上的最小误分类率的平均值不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为9,10个训练集在此处建立了10个模型,第6个训练集建立的模型较好,它在对应验证集上的最小误分类率最小,在预测集上的识别率为79.31%。
4.基于封装器模式穷举搜索的特征子集最优,它能够有效地剔除冗余特征、辨识关键特征,但计算开销大、速度慢。基于过滤器模式启发式搜索的特征子集次优,可能选出冗余特征,但计算量小,速度快,并可能搜索到过于乐观的特征子集,其验证集误差小于训练集误差,或者预测集误差小于验证集误差。基于交叉验证选择的模型稳定、可靠。
为了判别田间籽棉的成熟度,最终选择在空域中基于封装器模式穷举搜索的几何结构特征子集,该特征子集是径向切割区域特征(1,2)、圆周向切割区域特征(1,3,4)、径向切割线特征2,揭示了“成熟籽棉内部绽开外围饱满,未成熟籽棉内部紧实外围瘦小。”这一关键特性,它在预测集上的识别率为88.89%。
为了判别田间籽棉的品级,最终选择基于交叉验证和封装器模式穷举搜索的纹理、形状特征子集,该特征子集是白度三阶矩、黄度均值、黄度标准差、黄度熵、棉瓣个数、全局特征、圆周向切割区域特征1、径向切割线特征1和计盒维数,揭示了高品级棉花白度高、黄度低、杂质少、棉瓣紧凑/肥大/蓬松,低品级棉花白度低、黄度高、杂质多、棉瓣松散/瘦小/僵硬、淡灰棉/污染棉/烂桃棉多等特性,它在预测集上的识别率为79.31%。籽棉通常被误分入相邻品级,高品级籽棉和低品级籽棉比较容易被识别。
The development goals of cotton production were including, such as the corresponding rate of cotton quality grade improved greatly, the foreign fiber problem, the adulteration event, and the inconsistent phenomenon of cotton baling decreased continuously, etc. The main measures of solving these problems were supervising strictly picking, purchase, and process cotton and decreasing foreign fiber content. Modern technology, for example, machine vision and pattern recognition, was applied in traditional picking task to discriminate ripeness and quality grade of raw cotton, and thus laying a foundation for picking robot to exert exactly. Using robot to pick cotton could adapt various cotton varieties, avoid pollution caused by agriculture chemicals, and reduce labor waste and agriculture cost. The main and creative achievements as follows:
1. Raw cotton and its background were regarded as two classes and segmented based on competitive learning network. The training data with class label was classified into two classes based on K-means clustering in HSI, Lab, Ohta, RGB color space, and the error rate of the training data was lowest in RGB color space-particularly B value. The competitive learning network with B of the training data was trained onetime in RGB color space.907cotton images were segmented with an accuracy of92.94%based on the competitive learning network and morphological filtering. With the help of prior knowledge and supervisory learning, the image segmentation arithmetic obtained high precision and efficiency without iterative and over-fitting of K-means clustering as well as high computing cost.
2. The spectrum feature set used to describe boundary of raw cotton was extracted in frequency domain. A sequence of coordinates of boundary contour of cotton were obtained by using8-neighbour tracking, from which64spectrum features without0.9correlation coefficient were extracted based on Fourier transform. Single factor anova analysis showed that the differences of the average of most spectrum features of ripe and unripe cottons were significant at the0.05level.
The geometric structure feature set used to express cotton shape was extracted in spatial domain. Cotton region was circumscribed or cut by a set of concentric circles, from which15geometric structure features with more0.9correlation coefficient were extracted based on ratio method, including global feature, three radial cut region features, five circumferential cut region features, two radial cut line features, three circumferential cut line features, and box-counting dimension, Single factor anova analysis showed that the differences of the average of most geometric structure features of ripe and unripe cottons were significant at the0.05level.
Ripeness and under-ripeness of raw cotton were regarded as the issue of two-class, and the feature subset was selected from the above feature set based on heuristic searching and filter pattern with an assessing function of class separability criterion. To select the feature subset based on optimal scalar feature, class separability measure of every scalar feature was calculated and sorted by descending on training set, and the frontal l scale features were assembled l feature subset to establish a Bayes-criterion based discrimination model (l=1,2,...,64or15) on training set. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that spectrum feature subset was selected at the best feature subset capacity of14, based on which the classification rate on prediction set was77.78%; geometric structure feature subset was selected at the best feature subset capacity of8, based on which the classification rate on prediction set was85.56%. To select the feature set based on floating searching,l feature subset (l=1,2,...,64or15) with maximal class separability measure was searched backward and forward to establish a Bayes-criterion based discrimination model on training set. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the spectrum feature subset was selected at the best subset capacity of22, based on which the classification rate on prediction set was82.22%; the geometric structure feature subset was selected at the best subset capacity of8, based on which the classification rate on prediction set was85.56%.
Ripeness and under-ripeness of raw cotton were considered as the issue of two-class, and the feature subset was selected from the above feature set based on exhaustive searching and wrapper pattern with an assessing function of error rate. All of l feature subset were used to establish Bayes-criterion based discrimination models on training set and l feature subset (l=1,2,...,64or15) with the minimum error rate on validation set was selected. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the minimum error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the spectrum feature subset was selected at the best subset capacity of15, based on which the classification rate on prediction set was83.33%; the geometric structure feature subset was selected at the best subset capacity of6, based on which the classification rate on prediction set was88.89%.
3. The color space measuring cotton color was selected based on correlation analysis. Cotton color was measured in Hunter, Lab, HSI color space according to Chinese government grading standard of raw cotton; and correlation analysis showed that some correlation coefficients of cotton color are false in three color space. With adjusting the component of color space in gray level, all of correlation coefficient are true in HSI color space, which can be used to measure cotton color.
The texture and shape feature set used to discriminate quality grade of raw cotton was extracted in spatial domain.12texture features, including mean, standard deviation, smoothness, the3rd-order moments, consistency, and entropy, were extracted based on the histogram of S and I in HSI color space to describe cotton color and impurity content, including rain rust, hoarfrost, stiff petal, stain, and rot boll.16shape features, including the number of cotton petals, global feature, three radial cut region features, five circumferential cut region features, two radial cut line features, three circumferential cut line features, and box-counting dimension, were extracted based on ratio method to describe cotton size and geometric structure distributing. A valid feature subset without more0.9correlation coefficient was selected based on the correlation of the average texture/shape feature and quality grade from1to7at the0.05level, including the3rd-order moment of I, mean of S, standard deviation of S, entropy of S, the number of cotton petals, global feature, the1st and4th circumferential cut region feature, the1st radial cut line feature, the3rd circumferential cut line feature, and box-counting dimension.
Quality grade of raw cotton was regarded as the issue of seven-class, and the feature subset was selected from the above valid feature set based on10-fold cross-validation, heuristic searching and filter pattern with an assessing function of class separability criterion. To select the feature subset based on optimal scalar feature, class separability measure of every scalar feature was calculated and sorted by descending on every training set, and the frontal l scale features were assembled l feature subset to establish a Bayes-criterion based discrimination model (1=1,2,...,64or15) on every training set. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of7on10training sets respectively, and the fifth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was74.71%. To select the feature set based on floating searching,l feature subset (l=1,2,...,11) with maximal class separability measure was searched backward and forward to establish a Bayes-criterion based discrimination model on every training set. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of6on10training sets respectively, and the ninth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was70.11%.
Quality grade of raw cotton was considered as the issue of seven-class, and the feature subset was selected from the above valid feature set based on10-fold cross-validation, exhaustive searching and wrapper pattern with an assessing function of error rate. All of l feature subset were used to establish Bayes-criterion based discrimination models on every training set and l feature subset (l=1,2,...,11) with the minimum error rate on the corresponding validation set was selected. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average minimum error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of9on10training sets respectively, and the sixth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was79.31%.
4. The feature subset based on exhaustive searching and wrapper pattern is optimal, which can availably eliminate redundant features and identify key features with large computing cost and low speed. The feature subset based on heuristic searching and filter pattern is suboptimal, which may select redundant features with little computing cost and high speed. And an optimistic feature subset may be selected, the error rate of which on validation set was less than that on training set, possibly, prediction set less than validation set.10-fold cross-validation based discrimination model is more steady and reliable.
In order to discriminate ripeness of raw cotton, the geometric structure feature subset was selected with a generalization precision of88.89%based on wrapper pattern and exhaustive searching finally, including the1st and2nd radial cut region feature, circumferential cut region feature (1st,3rd and4th), the2nd radial cut line feature, which revealed that the inner of ripe cotton is dissilient while outer is plump, and the inner of unripe cotton is close while outer is lank.
In order to discriminate quality grade of raw cotton, the texture and shape feature subset was selected with a generalization precision of79.31%based on10-fold cross-validation, wrapper pattern and exhaustive searching finally, including the3rd-order moment of I, mean of S, standard deviation of S, and entropy of S, the number of cotton petals, global feature, the1st circumferential cut region feature, the1st radial cut line feature, and box-counting dimension. The feature subset revealed that high grade cotton petal is whiter, lower yellow, less impurity, more compact, thicker, and fluffier, and that the low grade cotton petal is more light-grey, higher yellow, more impurity, sparser, thinner, stiffer, much stain, and much rot boll. Raw cotton was usually misclassified to neighboring grade, and the higher and lower grade raw cotton were identified exactly.
1.将籽棉与背景视为二个类别,基于竞争学习网络分割了籽棉图像。在HSI、Lab、 Ohta、RGB颜色空间下,用带类别标签的训练样本计算k-均值聚类的误分类率,选取了误分类率最小的RGB颜色空间,其B值较优。在RGB颜色空间下,用带类别标签的训练样本的B值一次性地训练了竞争学习网络模型。结合形态学滤波,用所训练的竞争学习网络模型分割907张新图像的准确率达92.94%。该方法结合了有监督的学习算法,使图像分割在先验知识的指导下进行,同时,避免了K-均值聚类的反复迭代和过拟合现象,降低了计算开销,提高了图像分割的效率和精度。
2.基于频域提取了描述籽棉边界轮廓的频谱特征集。用8-邻域跟踪法获取了籽棉区域边界轮廓的坐标序列,基于傅里叶变换提取64个频谱特征,其相关系数都不超过0.9；单因子方差分析表明,大多数频谱特征在成熟和未成熟籽棉上的均值具有显著的差异。
基于空域提取了个描述籽棉形状的几何结构特征集。用一组同心园外接或切割籽棉区域,基于比值法提取了15个几何结构特征,它们是,全局特征、径向切割区域特征1-3、圆周向切割区域特征1～5、径向切割线特征1-2、圆周向切割线特征1-3以及计盒维数,其相关系数超过了0.9；单因子方差分析表明,大多数几何结构特征在成熟和未成熟籽棉上的均值具有显著的差异。
将籽棉是否成熟视为二类问题,基于过滤器模式从上述特征集中启发式搜索了特征子集,评估函数为类可分性准则。最优标量特征组合特征子集时,在训练集上计算每个特征的类可分性测量值并降序排列,将前l个(l=1、2、…、64或15)特征组合形成l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为14,该特征子集在预测集上的识别率为77.78%；几何结构特征集的最佳特征子集容量为8,该特征子集在预测集上的识别率为85.56%。浮动搜索特征子集时,在训练集上浮动搜索具有最大类可分性测量值的l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为22,该特征子集在预测集上的识别率为82.22%；几何结构特征集的最佳特征子集容量为8,该特征子集在预测集上的识别率为85.56%。
将籽棉是否成熟视为二类问题,基于封装器模式从上述特征集中穷举搜索了特征子集,评估函数为分类器的误分类率。在训练集上组合所有可能的l维(l=1、2、64或15)特征子集建立贝叶斯判别模型,并选择在验证集上具有最小误分类率的l维特征子集,随着特征子集容量l的增加,l维特征子集在训练集上的误分类率不断减小,验证集上的最小误分类率不再减小处为最佳特征子集容量,实验结果表明,频谱特征集的最佳特征子集容量为15,该特征子集在预测集上的识别率为83.33%；几何结构特征集的最佳特征子集容量为6,该特征子集在预测集上的识别率为88.89%。
3.基于相关分析选取了测量籽棉色泽的颜色空间。根据中国籽棉品级分级文字标准,在Hunter、Lab和HSI颜色空间下度量籽棉的色泽,相关分析表明,籽棉的色泽特征呈虚假相关；经灰度修正后,HSI颜色空间消除了虚假相关,可度量籽棉的色泽。
基于空域提取了描述籽棉品级的纹理、形状特征集。在HSI颜色空间下,基于亮度Ⅰ和饱和度S的直方图的均值、标准差、平滑度、三阶矩、一致性和熵,提取了12个纹理特征以描述籽棉的色泽、雨锈/轻霜/僵瓣/污染/烂桃等杂质；基于比值法提取了16个形状特征以描述籽棉的大小、几何结构分布,它们是,棉瓣个数、全局特征、径向切割区域特征1～3、圆周向切割区域特征1～5、径向切割线特征1～2、圆周向切割线特征1～3以及计盒维数；剔除相关系数超过0.9的纹理、形状特征后,基于特征集在1～7级样本上均值与品级值之间的回归分析,在0.05的显著性水平下选取了线性相关的有效纹理、形状特征集,它们是,亮度直方图的三阶矩、饱和度直方图的均值、标准差和熵、棉瓣个数、全局特征、圆周向切割区域特征(1,4)、径向切割线特征1、圆周向切割线特征3、计盒维数。
将成熟籽棉品级视为七类问题,基于10折交叉验证和过滤器模式从上述有效特征集中启发式搜索了特征子集。最优标量特征组合特征子集时,在每一个训练集上计算每个特征的类可分性测量值并降序排列,将前l个(l=1、2、…、11)特征组合形成l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率不断减小,10个对应的验证集上的平均误分类率不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为7,10个训练集在此处建立了10个模型,第5个训练集在此处建立的模型较好,它在预测集上的识别率为74.71%。浮动搜索特征子集时,在每一个训练集上搜索具有最大类可分性测量值的l维特征子集建立贝叶斯判别模型,随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率不断减小,10个对应的验证集上的平均误分类率不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为6,10个训练集在此处建立了10个模型,第9个训练集建立的模型较好,它在预测集上的识别率为70.11%。
将成熟籽棉品级视为七类问题,基于10折交叉验证和封装器模式从上述有效特征集中穷举搜索了特征子集。在每一个训练集上组合所有可能的l维特征子集建立贝叶斯判别模型,并选择在对应验证集上具有最小误分类率的l维特征子集(l=1、2、11),随着特征子集容量l的增加,l维特征子集在10个训练集上的平均误分类率逐渐减小,10个对应的验证集上的最小误分类率的平均值不再减小处为最佳特征子集容量,实验结果表明,有效特征集的最佳特征子集容量为9,10个训练集在此处建立了10个模型,第6个训练集建立的模型较好,它在对应验证集上的最小误分类率最小,在预测集上的识别率为79.31%。
4.基于封装器模式穷举搜索的特征子集最优,它能够有效地剔除冗余特征、辨识关键特征,但计算开销大、速度慢。基于过滤器模式启发式搜索的特征子集次优,可能选出冗余特征,但计算量小,速度快,并可能搜索到过于乐观的特征子集,其验证集误差小于训练集误差,或者预测集误差小于验证集误差。基于交叉验证选择的模型稳定、可靠。
为了判别田间籽棉的成熟度,最终选择在空域中基于封装器模式穷举搜索的几何结构特征子集,该特征子集是径向切割区域特征(1,2)、圆周向切割区域特征(1,3,4)、径向切割线特征2,揭示了“成熟籽棉内部绽开外围饱满,未成熟籽棉内部紧实外围瘦小。”这一关键特性,它在预测集上的识别率为88.89%。
为了判别田间籽棉的品级,最终选择基于交叉验证和封装器模式穷举搜索的纹理、形状特征子集,该特征子集是白度三阶矩、黄度均值、黄度标准差、黄度熵、棉瓣个数、全局特征、圆周向切割区域特征1、径向切割线特征1和计盒维数,揭示了高品级棉花白度高、黄度低、杂质少、棉瓣紧凑/肥大/蓬松,低品级棉花白度低、黄度高、杂质多、棉瓣松散/瘦小/僵硬、淡灰棉/污染棉/烂桃棉多等特性,它在预测集上的识别率为79.31%。籽棉通常被误分入相邻品级,高品级籽棉和低品级籽棉比较容易被识别。
The development goals of cotton production were including, such as the corresponding rate of cotton quality grade improved greatly, the foreign fiber problem, the adulteration event, and the inconsistent phenomenon of cotton baling decreased continuously, etc. The main measures of solving these problems were supervising strictly picking, purchase, and process cotton and decreasing foreign fiber content. Modern technology, for example, machine vision and pattern recognition, was applied in traditional picking task to discriminate ripeness and quality grade of raw cotton, and thus laying a foundation for picking robot to exert exactly. Using robot to pick cotton could adapt various cotton varieties, avoid pollution caused by agriculture chemicals, and reduce labor waste and agriculture cost. The main and creative achievements as follows:
1. Raw cotton and its background were regarded as two classes and segmented based on competitive learning network. The training data with class label was classified into two classes based on K-means clustering in HSI, Lab, Ohta, RGB color space, and the error rate of the training data was lowest in RGB color space-particularly B value. The competitive learning network with B of the training data was trained onetime in RGB color space.907cotton images were segmented with an accuracy of92.94%based on the competitive learning network and morphological filtering. With the help of prior knowledge and supervisory learning, the image segmentation arithmetic obtained high precision and efficiency without iterative and over-fitting of K-means clustering as well as high computing cost.
2. The spectrum feature set used to describe boundary of raw cotton was extracted in frequency domain. A sequence of coordinates of boundary contour of cotton were obtained by using8-neighbour tracking, from which64spectrum features without0.9correlation coefficient were extracted based on Fourier transform. Single factor anova analysis showed that the differences of the average of most spectrum features of ripe and unripe cottons were significant at the0.05level.
The geometric structure feature set used to express cotton shape was extracted in spatial domain. Cotton region was circumscribed or cut by a set of concentric circles, from which15geometric structure features with more0.9correlation coefficient were extracted based on ratio method, including global feature, three radial cut region features, five circumferential cut region features, two radial cut line features, three circumferential cut line features, and box-counting dimension, Single factor anova analysis showed that the differences of the average of most geometric structure features of ripe and unripe cottons were significant at the0.05level.
Ripeness and under-ripeness of raw cotton were regarded as the issue of two-class, and the feature subset was selected from the above feature set based on heuristic searching and filter pattern with an assessing function of class separability criterion. To select the feature subset based on optimal scalar feature, class separability measure of every scalar feature was calculated and sorted by descending on training set, and the frontal l scale features were assembled l feature subset to establish a Bayes-criterion based discrimination model (l=1,2,...,64or15) on training set. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that spectrum feature subset was selected at the best feature subset capacity of14, based on which the classification rate on prediction set was77.78%; geometric structure feature subset was selected at the best feature subset capacity of8, based on which the classification rate on prediction set was85.56%. To select the feature set based on floating searching,l feature subset (l=1,2,...,64or15) with maximal class separability measure was searched backward and forward to establish a Bayes-criterion based discrimination model on training set. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the spectrum feature subset was selected at the best subset capacity of22, based on which the classification rate on prediction set was82.22%; the geometric structure feature subset was selected at the best subset capacity of8, based on which the classification rate on prediction set was85.56%.
Ripeness and under-ripeness of raw cotton were considered as the issue of two-class, and the feature subset was selected from the above feature set based on exhaustive searching and wrapper pattern with an assessing function of error rate. All of l feature subset were used to establish Bayes-criterion based discrimination models on training set and l feature subset (l=1,2,...,64or15) with the minimum error rate on validation set was selected. With increasing feature subset capacity of l, the error rate on training set kept decreasing while the minimum error rate on validation set changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the spectrum feature subset was selected at the best subset capacity of15, based on which the classification rate on prediction set was83.33%; the geometric structure feature subset was selected at the best subset capacity of6, based on which the classification rate on prediction set was88.89%.
3. The color space measuring cotton color was selected based on correlation analysis. Cotton color was measured in Hunter, Lab, HSI color space according to Chinese government grading standard of raw cotton; and correlation analysis showed that some correlation coefficients of cotton color are false in three color space. With adjusting the component of color space in gray level, all of correlation coefficient are true in HSI color space, which can be used to measure cotton color.
The texture and shape feature set used to discriminate quality grade of raw cotton was extracted in spatial domain.12texture features, including mean, standard deviation, smoothness, the3rd-order moments, consistency, and entropy, were extracted based on the histogram of S and I in HSI color space to describe cotton color and impurity content, including rain rust, hoarfrost, stiff petal, stain, and rot boll.16shape features, including the number of cotton petals, global feature, three radial cut region features, five circumferential cut region features, two radial cut line features, three circumferential cut line features, and box-counting dimension, were extracted based on ratio method to describe cotton size and geometric structure distributing. A valid feature subset without more0.9correlation coefficient was selected based on the correlation of the average texture/shape feature and quality grade from1to7at the0.05level, including the3rd-order moment of I, mean of S, standard deviation of S, entropy of S, the number of cotton petals, global feature, the1st and4th circumferential cut region feature, the1st radial cut line feature, the3rd circumferential cut line feature, and box-counting dimension.
Quality grade of raw cotton was regarded as the issue of seven-class, and the feature subset was selected from the above valid feature set based on10-fold cross-validation, heuristic searching and filter pattern with an assessing function of class separability criterion. To select the feature subset based on optimal scalar feature, class separability measure of every scalar feature was calculated and sorted by descending on every training set, and the frontal l scale features were assembled l feature subset to establish a Bayes-criterion based discrimination model (1=1,2,...,64or15) on every training set. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of7on10training sets respectively, and the fifth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was74.71%. To select the feature set based on floating searching,l feature subset (l=1,2,...,11) with maximal class separability measure was searched backward and forward to establish a Bayes-criterion based discrimination model on every training set. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of6on10training sets respectively, and the ninth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was70.11%.
Quality grade of raw cotton was considered as the issue of seven-class, and the feature subset was selected from the above valid feature set based on10-fold cross-validation, exhaustive searching and wrapper pattern with an assessing function of error rate. All of l feature subset were used to establish Bayes-criterion based discrimination models on every training set and l feature subset (l=1,2,...,11) with the minimum error rate on the corresponding validation set was selected. With increasing feature subset capacity of l, the average error rate on10training sets kept decreasing while the average minimum error rate on the10corresponding validation sets changed from decreasing to increasing, and the inflexion of which was used for the best subset capacity in practice. Experiment results showed that the10valid feature subsets were selected at the best subset capacity of9on10training sets respectively, and the sixth feature subset was selected with the minimum error rate on the corresponding validation set, based on which the classification rate on prediction set was79.31%.
4. The feature subset based on exhaustive searching and wrapper pattern is optimal, which can availably eliminate redundant features and identify key features with large computing cost and low speed. The feature subset based on heuristic searching and filter pattern is suboptimal, which may select redundant features with little computing cost and high speed. And an optimistic feature subset may be selected, the error rate of which on validation set was less than that on training set, possibly, prediction set less than validation set.10-fold cross-validation based discrimination model is more steady and reliable.
In order to discriminate ripeness of raw cotton, the geometric structure feature subset was selected with a generalization precision of88.89%based on wrapper pattern and exhaustive searching finally, including the1st and2nd radial cut region feature, circumferential cut region feature (1st,3rd and4th), the2nd radial cut line feature, which revealed that the inner of ripe cotton is dissilient while outer is plump, and the inner of unripe cotton is close while outer is lank.
In order to discriminate quality grade of raw cotton, the texture and shape feature subset was selected with a generalization precision of79.31%based on10-fold cross-validation, wrapper pattern and exhaustive searching finally, including the3rd-order moment of I, mean of S, standard deviation of S, and entropy of S, the number of cotton petals, global feature, the1st circumferential cut region feature, the1st radial cut line feature, and box-counting dimension. The feature subset revealed that high grade cotton petal is whiter, lower yellow, less impurity, more compact, thicker, and fluffier, and that the low grade cotton petal is more light-grey, higher yellow, more impurity, sparser, thinner, stiffer, much stain, and much rot boll. Raw cotton was usually misclassified to neighboring grade, and the higher and lower grade raw cotton were identified exactly.
引文
[1]USDA棉花展望[EB/OL].[2009-09-24].http://news.ctei.gov.cn/zxzx/200185.htm.
[2]USDA棉花展望[EB/OL].[2009-07-20].http://news.ctei.gov.cn/zxzx/187509.htm.
[3]USDA棉花市场分析报告[EB/OL].[2009-09-17].http://news.ctei.gov.cn/zxzx/197994.htm.
[4]USDA棉花市场分析报告[EB/OL].[2009-07-16].http://news.ctei.gov.cn/zxzx/187073.htm.
[5]柯炳生.中国棉花生产的波动变化与发展趋势[C].2003年中国国际棉花会议文集,2003：16-18,61-66.
[6]沈家涛.新疆兵团机采棉技术推广存在的问题及原因分析[J].中国棉花,2005,32(3)：42-43.
[7]冯绍斌,谷成超.棉花采摘新概念：人工快采+机器清花[J].新疆农垦经济,2002,(4)：51-52.
[8]李伟明,林永增.采摘时间对中长绒棉纤维品质的影响[J].河北农业科学,2004,8(1)：117-118.
[9]许建,丁志毅.棉花化学脱叶催熟效果及其对棉子质量影响初报[J].新疆农业科学,2003,40(2)：116-118.
[10]沈玲凤.兵团农场采棉机械化发展的制约因素及对策[J].新疆农机化,2004,(4)：46-47.
[11]方建军.移动式采摘机器人研究现状与进展[J].农业工程学报,2004,20(2)：273-278.
[12]Iida M, Maekawa T, Umeda M. Automatic follow-up vehicle system for agriculture[J]. Journal of the Japanese Society of Agricultural,1999,61(1):99-106.
[13]张军.国内外棉花检验技术现状分析[J].中国棉花加工,2002,(1)：30-31.
[14]Harrell R C, Adsit P D, Munilla R D, Slaughter D C. Robotic picking of citrus[J]. Robotica,1990, 8(4):269-278.
[15]Jimenez A R, Ceres R, Pons J L. A survey of computer vision methods for locating fruit on trees[J]. Transactions of the ASAE,2000,43 (6):1911-1920.
[16]方如明,蔡健荣.计算机图像处理技术及其在农业工程中的应用[M].北京：清华大学出版社,1999：25-26.
[17]Umeda M, Kubota S, Iida M. Development of "STORK", a watermelon-harvesting robot[J]. Artificial Life and Robotics,1999,3(3):143-147.
[18]郭峰,曹其新,谢国俊等.基于OHTA颜色空间的瓜果轮廓提取方法[J].农业机械学报,2005,36(11)：113-116.
[19]赵杰文,刘木华,杨国彬.基于HSI颜色特征的田间成熟番茄识别技术[J].农业机械学报, 2004,35(5)：122-124.
[20]梁喜凤,苗香雯,崔绍荣等.番茄收获机械手运动学优化与仿真试验[J].农业机械学报,2005,36(7)：96-100.
[21]汤修映,张铁中.果蔬收获机器人研究综述[J].机器人,2005,27(1)：90-96.
[22]宋健,张铁中,徐丽明.汤修映果蔬采摘机器人研究进展与展望[J].农业机械学报,2006,37(5)：158-162.
[23]蔡自兴.机器人学[M].北京：清华大学出版社,2000：76.
[24]冯建农,柳明,吴捷.自主移动机器人智能导航研究进展[J].机器人,1997,19(6)：468-478.
[25]Zhang N, Wang M, Wang N. Precision agriculture-a wordwide overview[J]. Computers and Electronics in Agriculture,2002, (36):113-132.
[26]徐丽明,张铁中.果蔬果实收获机器人的研究现状及关键问题和对策[J].农业工程学报,2004,20(5)：38-42.
[27]Kondo N, Ting K C. Robotics for bioproduction system[J]. American Society of Agricultural Engineers,1998:325.
[28]周增产,Bontsema J荷兰黄瓜收获机器人的研究开发[JJ.农业工程学报,2001,17(6)：77-80.
[29]蔡健荣,周小军,李玉良,范军.基于机器视觉自然场景下成熟柑橘识别[J].农业工程学报,2008,24(1)：175-178.
[30]郑南宁.计算机视觉与模式识别[M].北京：国防工业出版社,1998：49-192.
[31]Gonzalez R C, Woods R E数字图像处理[M].2版.阮秋琦译.北京：电子工业出版社,2003：224-254,460-510.
[32]Pentland A. Fractal based decomposition of natural scenes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(6):661-674.
[33]Wuncsh P, Laine A F. Wavelet descriptors for multiresolution recognition of handwritten characters[J]. Pattern Recognition,1995,28(8):1237-1249.
[34]Peleg S, Naor J, Hartley R, Anvir D. Multiple resolution texture analysis and classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(4):518-523.
[35]Wuncsh P, Laine A F. Wavelet descriptors for multiresolution recognition of handwritten characters[J]. Pattern Recognition,1995,28(8):1237-1249.
[36]Arbter K, Snyder W E, Burkhardt H, Hirzinger G. Application of affine-invariant Fourier descriptors to recognition of 3-D objects[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1990,12(7):640-647.
[37]Crimmins T R. A complete set of Fourier descriptors for two dimensional shapes[J]. IEEE Transactions on Systems, Man Cybernetics,1982,12(6):848-855.
[38]罗志增,蒋静坪.机器人感觉与多信息融合[M].北京：机械工业出版社,2002：120-133.
[39]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004：83-109,451-454.
[40]Hastie T, Tibshirani R, Friedmar J统计学习基础——数据挖掘、推理与预测[M].范明,柴玉梅译.北京：电子工业出版社,2004：135-150,305-356.
[41]Sarle W S. Generalization [ai-faq/neural-nets/part3] [EB/OL]. [2001-05-21]. ftp://ftp.sas.com/pub/neural/FAQ3.html
[42]李宁,刘东波,臧英明.中国棉花分级标准与国外棉花分级标准差异的研究[J].大连轻工业学院学报,2001,20(4)：309-312.
[43]梁湘.世界原棉分级技术(下)[J].北京纺织,1996,(2)：54-56.
[44]梁湘.世界原棉分级技术(上)[J].北京纺织,1996,(1)：54-58.
[45]钱林森,汤魁中,王成礼.世界棉花产销与检验[M].济南：山东科学技术出版社,1996：173-210.
[46]杨宝富.中美棉花分级标准异同分析[J].中国棉花加工,2000,(6)：19-20.
[47]Kang T J, Kim S C. Objective evaluation of the trash and color of raw cotton by image processing and neural network[J]. Textile Research Journal,2002,72(9):776-782.
[48]Xu B, Fang C, Watson M D. Investigating new factors in cotton color grading[J]. Textile Research Journal,1998,68(11):779-787.
[49]Xu B, Su J, Dale D S, Watson M D. Cotton color grading with a neural network[J]. Textile Research Journal,2000,70(5):430-436.
[50]Xu B, Dale D S, Huang Y, Watson M D. Cotton color classification by fuzzy logic[J]. Textile Research Journal,2002,72(6):504-509.
[51]Sarraf H S, Hequet E F, Abidi N, Dai Y, Chan H Y, Jasso M F, Morris B. Image processing algorithm for automatic assessment of fabric shrinkage[C]. Proceedings of SPIE-The International Society for Optical Engineering, Vol.4664:89-96.
[52]Han Y J, Cho Y J, Lambert W E, Bragg C K. Identification and measurement of convolutions in cotton fiber using image analysis[J]. Artificial Intelligence Review,1998,12(1-3):201-211.
[53]Wang Q, Fan X R, Gao W D, Chen J. Characterization of bioscoured cotton fabrics using FT-IR ATR spectroscopy and microscopy techniques [J]. Carbohydrate Research,2006,341(12):2170-2175.
[54]Leigh R M, Cauthen D A. Determining cotton fiber maturity by image analysis:Part I:direct measurement of cotton fiber characteristics [J]. Textile Research Journal,1994,64(9):534-544.
[55]Rajasekaran K, Thibodeaux D P. Fidelity of cotton fiber maturity measurement by image analysis[J]. Journal of New Seeds,2002,4(3):25-36.
[56]Wu D, Ma C W. The support vector machine (SVM) based near-Infrared spectrum recognition of leaves infected by the leafminers[C]. Proceedings of the First International Conference on Innovative Computing, Information and Control,2006,3:448-451.
[57]Thomasson J A, Shearer S A, Byler R K. Image-processing solution to cotton color measurement problems:part I-instrument design and construction[J]. Transactions of the ASAE,2005,48 (2): 421-438.
[58]刘立海.棉花等级分类的预处理和特征提取[J].数据采集与处理,2001,16(2)：145-149.
[59]丁天怀,郏东耀.利用多颜色空间特征融合方法检测近似目标[J].清华大学学报(自然科学版),2006,46(2)：176-179.
[60]许鹤群,黄健.棉纤维成熟度的图象分析[J].中国纺织大学学报,1992,8(1)：78-83.
[61]王勇,沈明霞,姬长英.采摘期成熟棉花不同部位颜色识别分析[J].农业工程学报,2007,23(4)：183-185.
[62]王勇,姬长英.田间早期成熟棉花识别研究[J].江西农业学报,2006,18(6)：141-143.
[63]王勇,沈明霞,姬长英.基于颜色信息和形状特征的棉桃识别方法[J].农业机械学报,2007,38(11)：77-79,87.
[64]王勇,沈明霞,姬长英.基于色差信息的田间成熟棉花识别[J].浙江农业学报,2007,19(5)：385-388.
[65]禹素萍,曾培峰,陈建平.基于图像掩码的棉纤维截面分割[J].计算机工程,2007,33(13)：188-190.
[66]禹素萍,曾培峰,陈建平.棉纤维截面图像的自适应阈值Canny边缘检测[J].计算机工程,2007,33(11)：168-203.
[67]张教海,李亚兵,别墅,沈杰,夏松波,王孝刚.数字图像处理在棉花形态特征提取上的应用[J].湖北农业科学,2007,46(3)：369-371.
[68]何晓峰,刘文耀,徐守东.游离棉纤维长度的计算机测量[J].棉花学报,2003,15(6)：339-343.
[69]何晓峰,刘文耀,吴国新.棉纤维品质长度的图像测量[J].棉花学报,2005,17(6)：348-352.
[70]宋新平.应用微机图象处理技术检测纤维材料[J].纺织标准与质量,2000,(4)：25-26.
[71]胡宝林.从异性纤维状况谈清理技术措施[J].中国纤检,2006,(12)：51-52.
[72]张晓璐,韩力群.计算机视觉技术在纤维检验中的应用[J].北京工商大学学报(自然科学版),2005,23(2)：43-45.
[73]郏东耀,丁天怀.检测皮棉内部杂质的透射成像影响因素分析[J].农业机械学报,2005,36(2)：65-69.
[74]付江,于伟东.国外棉花含杂量检测及相关测量[J].中国纤检,2006,(8)：43-48.
[75]Barker R L, Lyons D W. Instrumented procedures for analysis of non-lint particles in cotton[J]. Journal of Engnieering for Industry,1976,98 (3):845-848.
[76]Lyons D W, Barker R L. Comparative analysis of trash contamination in cotton by optical scanning[J]. Journal of Engineering for Industry,1977,99 (1):46-50.
[77]Taylor R A. Using high-speed image analysis to estimate trash in cotton[J]. Journal of Engineering for Industry,1985,107(2):206-219.
[78]Taylor R A, Godbey L C. Using digital images to measure and discriminate small particles in cotton[C]. Proceedings of SPIE-The International Society for Optical Engineering,1991:16-27.
[79]Taylor R A. Measuring leaf material in ginned cotton from surface images[C]. Proceedings of SPIE-The International Society for Optical Engineering,1995:218-225.
[80]Taylor R A用视频扫杂质分析仪测定棉花杂质[J].冯平译.中国纤检,1992,(1)：24-28.
[81]Xu B, Fang C Y, Huang R, Watson M D. Chromatic image analysis for cotton trash and color measurement[J]. Textile Research Journal,1997,67 (12):881-890.
[82]Veit D, Hormes I, Bergmann J, Wulfhorst B. Image processing as a tool to improve machine performance and process control[J]. International Journal of Clothing Science and Technology, 1996,8(1-2):66-72.
[83]Xu B, Fang C, Watson M D. Clustering analysis for cotton trash classification[J]. Textile Research Journal,1999,69(9):656-662.
[84]Gururajan A, Sarraf H S, Hequet E F. Statistical modeling, detection, and segmentation of stains in digitized fabric images[C]. Proceedings of SPIE-The International Society for Optical Engineering-Machine vision applications in industrial inspection,2007, Vol.6503.
[85]Dogan M S, Sarraf H S, Hequet E F. Cotton trash assessment in radiographic X-ray images with scale-space filtering and stereo analysis[C]. Proceedings of Machine Vision Applications in Industrial Inspection XIII,2005, Vol.5679:276-287.
[86]Kang T J, Kim S C. Objection evaluation of the trash and color of raw cotton by image processing and neural network [J]. Textile Research Journal,2002,36(11):124-128.
[87]Tilocca A, Borzone P, Carosio S, Durante A. Detecting fabric defects with a neural network using two kinds of optical patterns[J]. Textile Research Journal,2002,72 (6):545-550.
[88]Pelletier M G, Barker G L, Baker R V. Image processing for stripper harvested cotton trash content measurement a progress report[C]. Proceedings Beltwide Cotton Conferences,1999,2:1412-1415.
[89]Lieberman M A, Bragg C K, Brennan S N. Determining gravimetric bark content in cotton with machine vision[J]. Textile Research Journal,1998,68(2):94-104.
[90]Siddaiah M N, Lieberman M, Hughs S, Foulk J. Trash measurements in ginned cotton[C]. Proceedings Beltwide Cotton Conferences,2005:1926-1937.
[91]Siddaiah M N, Lieberman M, Hughs S, Foulk J. Comparison of small trash measurements between imaging techniques and AFIS[C]. Proceedings Beltwide Cotton Conferences,2005:2362-2372.
[92]Siddaiah M N, LAS C N, Lieberman M, Hughs S. High speed trash measurements[C]. Proceedings Beltwide Cotton Conferences,2003:1-10.
[93]Frydrych R, Krifa M, Tamime O, Giner M, Gourlot J P. Detection and counting of two cotton contaminants:seed coat fragments and honeydew depostts[C]. Proceedings Beltwide Cotton Conferences,1999,1:695-698.
[94]Jia D Y, Ding T H. Detection of foreign materials in cotton using a multi-wavelength imaging method[J]. Meas Sci Technol,2005,16:1355-1362.
[95]Fitzgerald G J, Maas S J, Detar W R. Early detection of spider mites in cotton using multispectral remote sensing[C]. Proceedings Beltwide Cotton Conferences,1999,2:1022-1023.
[96]Pavani S K, Mehmet S D, Hamed S S, Eric F H. Segmentation and classification of four common cotton contaminants in X-Ray micro-tomographic images[C]. Conference on Machine Vision Applications in Industrial Inspection,2004:1-13.
[97]Anthony W S. Comparison of video color/trash meters for measurements on cotton[J]. Apllied Engineering in Agriculture,1989,5 (1):18-23.
[98]Anthony W S. Online assessment of foreign matter in cotton during ginning[J]. Applied Engineering in Agriculture,1990,6 (3):267-272.
[99]Farah B D. Image processing in the measurement of trash contents and grades in cotton[C]. Proceedings IWISP, Manchester, United Kingdom,1996:629-632.
[100]Jensen K L, Carstensen M. Fuzz and pills evaluated on knitted textiles by image analysis[J]. Textile Research Journal,2002,72(1):34-50.
[101]Pelletier M G. High speed machine vision sensing of cotton lint trash[C]. Cotton Production & Processing Research Unit,2007:1-13.
[102]Tantaswadi P, Vilainatre J, Tamaree N, Viraivan P. Machine vision for automated visual inspection of cotton quality in textile industries using color isodiscrimination contour[J]. Computers & Industrial Engineering.1999,37(1-2):347-350.
[103]Brainard, D H. Colorimeter[M]. Bass M, Editor, Optical Society of America:Handbook of Optics, 1994, Vol.1, pp.26.1-26.54.
[104]丁天怀,李勇,苗君哲,郑颖航.基于BP神经网络的皮棉杂质在线检测方法[J].农业工程学报,2003,19(2)：137-139.
[105]苗君哲.用图象处理法在线检测棉花杂质的试验研究[D].北京：清华大学,2000,20-65.
[106]冯显英,任长志,黄燕云.基于机器视觉的异性纤维检测系统[J].山东大学学报(工学版), 2006,36(4)：5-8.
[107]郭彩霞,袁建畅,金守峰.基于BP网络的棉花中异性纤维识别算法[J].西安工程科技学院学报,2006,20(5)：542-544.
[108]符宝鼎,袁建畅,郭彩霞.基于RGB颜色模型棉花杂质的识别算法[J].北京纺织,2005,26(5)：48-51.
[109]符宝鼎.异性纤维识别算法的研究[D].西安：西安工程科技学院,2006：1-75.
[110]罗德坡,朱邦太,李勋.紫外线荧光效应及其在棉花异性纤维分拣系统中的应用[J].河南科技大学学报(自然科学版),2007,28(2)：63-66.
[111]郏东耀,丁天怀.纤维红外吸收特性及其在皮棉杂质检测中的应用[J].农业工程学报,2004,20(3)：104-108.
[112]郏东耀,丁天怀.棉花中异性纤维的多光谱检测[J].清华大学学报(自然科学版),2005,45(2)：193-196.
[113]郏东耀,丁天怀.利用纤维红外吸收特性的皮棉杂质检测新方法[J].红外与毫米波学报,2005,24(2)：147-150.
[114]金守峰.异性纤维检测装置的研究[D].西安：西安工程科技学院,2005,1-60
[115]金守峰,袁建畅,张慧.基于Matlab的棉花异性纤维检测及定位算法[J].北京纺织,2004,25(6)：48-50.
[116]金守峰,袁建畅,张慧Matlab在棉纤维杂质检测中的应用[J].现代纺织技术,2004,12(3)：5-7.
[117]王宏坡.基于DSP的原棉杂质实时检测系统[D].天津：天津工业大学,2005：1-62.
[118]吕宇涛.异性纤维检测系统研究及硬件实现[D].北京：北京工业大学,2004：1-71.
[119]李碧丹,丁天怀,郏东耀.皮棉异性纤维剔除系统设计[J].农业机械学报,2006,37(1)：107-110.
[120]常蕾.基于DSP的棉花异性纤维检测系统[D].合肥：合肥工业大学,2006：1-43.
[121]赵鹏.异性纤维捡出机系统及其软件算法的实现[D].曲阜：曲阜师范大学,2006：1-45.
[122]张海容.棉花异性纤维挑拣技术研究[D].合肥：合肥工业大学,2006：1-35.
[123]姚远.异性纤维检测装置基本算法及可行性研究[D].西安：西安工程科技学院,2002：1-56.
[124]张慧,袁建畅,金守峰.棉花杂质检测定位算法的研究[J].计量与测试技术,2004,31(12)：44-45.
[125]艾世一.计算机图像识别技术在异纤在线检测中的应用[D].天津：天津工业大学,2004：1-67.
[126]艾世一,王跃存,刘迪.基于MATLAB的图像识别技术的原棉异纤在线检测[J].天津工业大学学报,2004,23(2)：70-73.
[1]Underwood S A, Aggarwal J K. Interactive computer analysis of aerial color infrared photographs[J]. Computer Graphics and Image Processing,1977,6 (1):1-24.
[2]Kurugollu F, Sankur B, Harmanci A E. Color image segmentation using histogram multith resholding and fusion[J]. Image and Vision Computing,2001,19(13):915-928.
[3]Celenk M. A color clustering technique for image segmentation[J]. Computer Vision, Graphics and Image Processing,1990,52(2):145-170.
[4]Ahuja N, Rosenfeld A, Haralick R M. Neighbor gray levels as features in pixel classification[J]. Pattern Recognition,1980,13(4):251-260.
[5]刘重庆,程华.分割彩色图像的一种有效聚类方法[J].模式识别与人工智能,1995,8(S1)：133-138.
[6]阴国富.基于阈值法的图像分割技术[J].现代电子技术,2007,(23)：107-108.
[7]Nevatia R. A color edge detector and its use in scene segmentation[J]. IEEE Transactions on Systems, Man and Cybernetics,1977,7(11):820-826.
[8]Ohta Y I, Kanade T, Sakai T. Color information for region segmentation[J]. Computer Graphics and Image Processing,1980,13(3):222-241.
[9]Ohlander R, Price K, Reddy D R. Picture segmentation using a recursive region splitting method[J]. Computer Graphics and Image Processing,1978,8(3):313-333.
[10]柳萍,阳爱民.一种基于区域的彩色图像分割方法[J].计算机工程与应用,2007,43(6)：37-39,64.
[11]潘红艳,敬忠良,肖刚,郭超.融合颜色和区域信息的彩色图像分割新方法[J].计算机工程与应用,2007,43(25)：72-74.
[12]Vincent L, Soille P. Watersheds in digital spaces:an efficient algorithm based on immersion simulations[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1991,13(6): 583-598.
[13]Lezoray O, Cardot H. Cooperation of color pixel classification schemes and color watershed:a study for microscopic images[J]. IEEE Transactions on Image Processing,2002,11(7):783-789.
[14]Huang C L, Cheng T Y, Chen C C. Color images segmentation using scale space filter and Markov random field[J]. Pattern Recognition,1992,25(10):1217-1229.
[15]Liu J, Yang Y H. Multiresolution color image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1994,16(7):689-700.
[16]黄玉明,徐光,叶培健.基于物理模型的彩色图像分割[J].自动化学报,1992,18(4)：421-429.
[17]Tizhoosh H R. Fuzzy image processing[M]. Berlin:Springer Verlag,1997.
[18]Carevic D, Caelli T. Region-based coding of color image using Karhunen-Loeve Transform[J]. Graphics Models and Image Processing,1997,59(1):27-38.
[19]王兴伟,沈兰荪,卫保国等.基于改进的K均值聚类和数学形态学的彩色眼科图像病灶分割[J].中国生物医学工程学报,2002,21(5)：443-448.
[20]Bezdek J C. Pattern recognition with fuzzy objective function algorithms[M]. New York:Plenum Press,1981.
[21]刘木华,赵杰文,张海东.基于模糊C均值聚类的牛肉图像中脂肪和肌肉区域分割技术[J].农业工程学报,2004,20(2)：161-163.
[22]朱伟华,曹其新.基于模糊彩色聚类方法的西红柿缺陷分割研究[J].农业工程学报,2003,19(3)：133-136.
[23]周天娟,张铁中,杨丽,赵金英.基于数学形态学的相接触草莓果实的分割方法及比较研究[J].农业工程学报,2007,23(9)：164-168.
[24]Ong S H, Yeo N C, Lee K H, Venkatesh Y V, Cao D M. Segmentation of color images using a two stage self-organizing network[J]. Image and Vision Computing,2002,20(4):279-289.
[25]Uchiyama T, Arbib M A. Color image segmentation using competitive learning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1994,16(12):1197-1206.
[26]常发亮,刘静,乔谊正.基于自组织神经网络的彩色图像自适应聚类分割[J].控制与决策,2006,21(4)：449-452.
[27]徐海荣,田国会,李晓磊.一种基于自组织神经网络的彩色图像分割方法[J].山东大学学报(工学版),2005,35(6)：67-70.
[28]黎宁,毛四新,李有福.用于彩色图像分割的有效特征分析[J].南京航空航天大学学报,2001,18(2)：206-212.
[29]石晓龙,许志榜,许进.基于自组织特征映射的图像分割算法研究[J].华中科技大学学报(自然科学版),2005,33(9)：13-25.
[30]况菲,王耀南,余洪山等.一种基于学习向量量化神经网络的图像分割方法[J].计算机工程与应用,2005,(14)：34-36.
[31]Huang C L. Parallel image segmentation using modified hopfield model[J]. Pattern Recognition Letters,1993,13(5):345-353.
[32]Campadelli P, Medici D, Schettini R. Color image segmentation using hopfield networks[J]. Image and Vision Computing,1997,15(3):161-166.
[33]Lescure P, Yedid V M, Dupoisot H, Stamon G. Color segmentation on biological microscope images[C]. Proceedings of SPIE-The International Society for Optical Engineering,1999, Vol. 3647:182-193.
[34]王勇,沈明霞,姬长英.采摘期成熟棉花不同部位颜色识别分析[J].农业工程学报,2007,23(4)：183-185.
[35]Ohta Y I, Kanade T, Sakai T. Color information for region segmentation[J]. Computer Graphics and Image Processing,1980,13(3):222-241.
[36]林开颜,吴军辉,徐立鸿.彩色图像分割方法综述[J].中国图像图形学报,2005,10(1)：1-10.
[37]Gonzalez R C, Woods R E, Eddins S L数字图像处理[M]. Matlab版.阮秋琦等译.北京：电子工业出版社,2005：144-178.
[38]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004,83-109,451-454.
[39]Macadam D L. Visual sensitivites to color differences in daylight[J]. Journal of the Optical Society of America,1942,32(5):247-274.
[40]Hastie T, Tibshirani R, Friedmar J统计学习基础——数据挖掘、推理与预测[M].北京：电子工业出版社,2004,135-150,305-356.
[41]张敏,于剑.基于划分的模糊聚类算法[J].软件学报,2004,15(6)：858-868.
[42]Cheng H D, Jiang X H, Sun Y, Wang J. Color image segmentation:advances and prospects[J]. Pattern Recognition,2001,34(12):2259-2281.
[43]Li N, Li Y F. Feature encoding for unsupervised segmentation of color images[J]. IEEE Transactions on Systems, Man and Cybernetics,2003,33(3):438-447.
[44]Sergios T, Konstantinos K模式识别[M].2版.北京：电子工业出版社,2004：257-283.
[45]徐飞,施晓红Matlab应用图像处理[M].西安：西安电子科技大学出版社,2002：101-214.
[46]陈杨,陈荣娟,郭颖辉Matlab 6. X图形编程与图像处理[M].西安：西安电子科技大学出版社,2002：162-187.
[47]蔡健荣,周小军,李玉良,范军.基于机器视觉自然场景下成熟柑橘识别[J].农业工程学报,2008,24(1)：175-178.
[1]陈友,程学旗,李洋,戴磊.基于特征选择的轻量级入侵检测系统[J].软件学报,2007,18(7)：1639-1651.
[2]毛勇,周晓波,夏铮,尹征,孙优贤.特征选择算法研究综述[J].模式识别与人工智能,2007,20(2)：211-218.
[3]崔自峰,徐宝文,张卫丰,徐峻岭.一种近似Markov Blanket最优特征选择算法[J].计算机学报,2007,30(12)：2074-2081.
[4]付峰,应义斌.生物图像阈值分割方法的研究[J].浙江大学学报(农业与生命科学版),2003,29(1)：108-112.
[5]Webb A R.统计模式识别[M].2版.北京：电子工业出版社,2004：7-30&342-346.
[6]Kuhl F P, Giardina C R. Elliptic Fourier features of a closed contour[J]. Computer Graphics and Image Processing,1982,18(3):236-258.
[7]Lin C S, Hwang C L. New forms of shape invariants from elliptic Fourier descriptors[J]. Pattern Recognition,1987,20(5):535-545.
[8]李秀智.基于机器视觉的苹果形状分级系统研究[D].南京：南京农业大学,2003：17-21.
[9]Taxt T, Olafsdottir J B, Daechlen M. Recognition of hand written symbols[J]. Pattern Recognition, 1990,23(11):1155-1166.
[10]Fu K S, Persoon E. Shape discrimination using Fourier descriptors[J]. IEEE Trans. Pattern Anal Mach. Intellig.1986,8(3):388-397.
[11]Pitas I. Digital image processing algorithms and applications[M]. Prentice Hall,1992:400-413.
[12]Sergios T, Konstantinos K模式识别[M].2版.北京：电子工业出版社,2004：257-283.
[13]黄心晔,王茂祥,富煜清,陆佶人.基于结构分析的手写体数字识别算法[J].电子工程师.1999,11(5)：23-26.
[14]张济忠.分形[M].北京：清华大学出版社,1995：111-141.
[1]李宁,刘东波,臧英明.中国棉花分级标准与国外棉花分级标准差异的研究[J].大连轻工业学院学报,2001,20(4)：309-312.
[2]《棉花仪器化检验》(草案)起草小组.棉花品级与色特征级的辩证关系[J].中国纤检,2005,(5)：5-7.
[3]中国纤维检验局棉花质量监督处.《棉花细绒棉》新标准系列问答：棉花品级、长度18问[J].高端·视点.2007,(8)：12-15.
[4]熊宗伟.我国棉花纤维质量及颜色等级划分研究[D].北京：中国农业大学,2005：10-25.
[5]熊宗伟.棉花色特征[J].中国棉花,1995,22(4)：39-40.
[6]郑芝奖.我国棉花品级与棉花色征指标的关系[J].纺织学报,2001,22(2)：23-24.
[7]马兆仁,邓明辉,詹建斌,周献珠,陈贺,江洪.新疆棉花色泽特征、马克隆值及品质分析[J].中国纤检,2006,(5)：26-29.
[8]Kang T J, Kim S C. Objective evaluation of the trash and color of raw cotton by image processing and neural network[J]. Textile Research Journal,2002,72(9):776-782.
[9]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004：83-109,451-454.
[10]Gonzalez R C, Woods R E, Eddins S L数字图像处理[M]. Matlab版.阮秋琦等译.北京：电子工业出版社,2005：144-178.
[11]Webb A R统计模式识别[M].2版.北京：电子工业出版社,2004：7-30&342-346.
[12]Cantoni E, Field C, Flemming J M, Ronchetti E. Longitudinal variable selection by cross-validation in the case of many covariates[J]. Statistics in Medicine,2007,26(4):919-930.
[13]Trevor H, Robert T, Jerome F.统计学习基础——数据挖掘、推理与预测[M].北京：电子工业出版社,2004：135-150,305-356.
[2]USDA棉花展望[EB/OL].[2009-07-20].http://news.ctei.gov.cn/zxzx/187509.htm.
[3]USDA棉花市场分析报告[EB/OL].[2009-09-17].http://news.ctei.gov.cn/zxzx/197994.htm.
[4]USDA棉花市场分析报告[EB/OL].[2009-07-16].http://news.ctei.gov.cn/zxzx/187073.htm.
[5]柯炳生.中国棉花生产的波动变化与发展趋势[C].2003年中国国际棉花会议文集,2003：16-18,61-66.
[6]沈家涛.新疆兵团机采棉技术推广存在的问题及原因分析[J].中国棉花,2005,32(3)：42-43.
[7]冯绍斌,谷成超.棉花采摘新概念：人工快采+机器清花[J].新疆农垦经济,2002,(4)：51-52.
[8]李伟明,林永增.采摘时间对中长绒棉纤维品质的影响[J].河北农业科学,2004,8(1)：117-118.
[9]许建,丁志毅.棉花化学脱叶催熟效果及其对棉子质量影响初报[J].新疆农业科学,2003,40(2)：116-118.
[10]沈玲凤.兵团农场采棉机械化发展的制约因素及对策[J].新疆农机化,2004,(4)：46-47.
[11]方建军.移动式采摘机器人研究现状与进展[J].农业工程学报,2004,20(2)：273-278.
[12]Iida M, Maekawa T, Umeda M. Automatic follow-up vehicle system for agriculture[J]. Journal of the Japanese Society of Agricultural,1999,61(1):99-106.
[13]张军.国内外棉花检验技术现状分析[J].中国棉花加工,2002,(1)：30-31.
[14]Harrell R C, Adsit P D, Munilla R D, Slaughter D C. Robotic picking of citrus[J]. Robotica,1990, 8(4):269-278.
[15]Jimenez A R, Ceres R, Pons J L. A survey of computer vision methods for locating fruit on trees[J]. Transactions of the ASAE,2000,43 (6):1911-1920.
[16]方如明,蔡健荣.计算机图像处理技术及其在农业工程中的应用[M].北京：清华大学出版社,1999：25-26.
[17]Umeda M, Kubota S, Iida M. Development of "STORK", a watermelon-harvesting robot[J]. Artificial Life and Robotics,1999,3(3):143-147.
[18]郭峰,曹其新,谢国俊等.基于OHTA颜色空间的瓜果轮廓提取方法[J].农业机械学报,2005,36(11)：113-116.
[19]赵杰文,刘木华,杨国彬.基于HSI颜色特征的田间成熟番茄识别技术[J].农业机械学报, 2004,35(5)：122-124.
[20]梁喜凤,苗香雯,崔绍荣等.番茄收获机械手运动学优化与仿真试验[J].农业机械学报,2005,36(7)：96-100.
[21]汤修映,张铁中.果蔬收获机器人研究综述[J].机器人,2005,27(1)：90-96.
[22]宋健,张铁中,徐丽明.汤修映果蔬采摘机器人研究进展与展望[J].农业机械学报,2006,37(5)：158-162.
[23]蔡自兴.机器人学[M].北京：清华大学出版社,2000：76.
[24]冯建农,柳明,吴捷.自主移动机器人智能导航研究进展[J].机器人,1997,19(6)：468-478.
[25]Zhang N, Wang M, Wang N. Precision agriculture-a wordwide overview[J]. Computers and Electronics in Agriculture,2002, (36):113-132.
[26]徐丽明,张铁中.果蔬果实收获机器人的研究现状及关键问题和对策[J].农业工程学报,2004,20(5)：38-42.
[27]Kondo N, Ting K C. Robotics for bioproduction system[J]. American Society of Agricultural Engineers,1998:325.
[28]周增产,Bontsema J荷兰黄瓜收获机器人的研究开发[JJ.农业工程学报,2001,17(6)：77-80.
[29]蔡健荣,周小军,李玉良,范军.基于机器视觉自然场景下成熟柑橘识别[J].农业工程学报,2008,24(1)：175-178.
[30]郑南宁.计算机视觉与模式识别[M].北京：国防工业出版社,1998：49-192.
[31]Gonzalez R C, Woods R E数字图像处理[M].2版.阮秋琦译.北京：电子工业出版社,2003：224-254,460-510.
[32]Pentland A. Fractal based decomposition of natural scenes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(6):661-674.
[33]Wuncsh P, Laine A F. Wavelet descriptors for multiresolution recognition of handwritten characters[J]. Pattern Recognition,1995,28(8):1237-1249.
[34]Peleg S, Naor J, Hartley R, Anvir D. Multiple resolution texture analysis and classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(4):518-523.
[35]Wuncsh P, Laine A F. Wavelet descriptors for multiresolution recognition of handwritten characters[J]. Pattern Recognition,1995,28(8):1237-1249.
[36]Arbter K, Snyder W E, Burkhardt H, Hirzinger G. Application of affine-invariant Fourier descriptors to recognition of 3-D objects[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1990,12(7):640-647.
[37]Crimmins T R. A complete set of Fourier descriptors for two dimensional shapes[J]. IEEE Transactions on Systems, Man Cybernetics,1982,12(6):848-855.
[38]罗志增,蒋静坪.机器人感觉与多信息融合[M].北京：机械工业出版社,2002：120-133.
[39]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004：83-109,451-454.
[40]Hastie T, Tibshirani R, Friedmar J统计学习基础——数据挖掘、推理与预测[M].范明,柴玉梅译.北京：电子工业出版社,2004：135-150,305-356.
[41]Sarle W S. Generalization [ai-faq/neural-nets/part3] [EB/OL]. [2001-05-21]. ftp://ftp.sas.com/pub/neural/FAQ3.html
[42]李宁,刘东波,臧英明.中国棉花分级标准与国外棉花分级标准差异的研究[J].大连轻工业学院学报,2001,20(4)：309-312.
[43]梁湘.世界原棉分级技术(下)[J].北京纺织,1996,(2)：54-56.
[44]梁湘.世界原棉分级技术(上)[J].北京纺织,1996,(1)：54-58.
[45]钱林森,汤魁中,王成礼.世界棉花产销与检验[M].济南：山东科学技术出版社,1996：173-210.
[46]杨宝富.中美棉花分级标准异同分析[J].中国棉花加工,2000,(6)：19-20.
[47]Kang T J, Kim S C. Objective evaluation of the trash and color of raw cotton by image processing and neural network[J]. Textile Research Journal,2002,72(9):776-782.
[48]Xu B, Fang C, Watson M D. Investigating new factors in cotton color grading[J]. Textile Research Journal,1998,68(11):779-787.
[49]Xu B, Su J, Dale D S, Watson M D. Cotton color grading with a neural network[J]. Textile Research Journal,2000,70(5):430-436.
[50]Xu B, Dale D S, Huang Y, Watson M D. Cotton color classification by fuzzy logic[J]. Textile Research Journal,2002,72(6):504-509.
[51]Sarraf H S, Hequet E F, Abidi N, Dai Y, Chan H Y, Jasso M F, Morris B. Image processing algorithm for automatic assessment of fabric shrinkage[C]. Proceedings of SPIE-The International Society for Optical Engineering, Vol.4664:89-96.
[52]Han Y J, Cho Y J, Lambert W E, Bragg C K. Identification and measurement of convolutions in cotton fiber using image analysis[J]. Artificial Intelligence Review,1998,12(1-3):201-211.
[53]Wang Q, Fan X R, Gao W D, Chen J. Characterization of bioscoured cotton fabrics using FT-IR ATR spectroscopy and microscopy techniques [J]. Carbohydrate Research,2006,341(12):2170-2175.
[54]Leigh R M, Cauthen D A. Determining cotton fiber maturity by image analysis:Part I:direct measurement of cotton fiber characteristics [J]. Textile Research Journal,1994,64(9):534-544.
[55]Rajasekaran K, Thibodeaux D P. Fidelity of cotton fiber maturity measurement by image analysis[J]. Journal of New Seeds,2002,4(3):25-36.
[56]Wu D, Ma C W. The support vector machine (SVM) based near-Infrared spectrum recognition of leaves infected by the leafminers[C]. Proceedings of the First International Conference on Innovative Computing, Information and Control,2006,3:448-451.
[57]Thomasson J A, Shearer S A, Byler R K. Image-processing solution to cotton color measurement problems:part I-instrument design and construction[J]. Transactions of the ASAE,2005,48 (2): 421-438.
[58]刘立海.棉花等级分类的预处理和特征提取[J].数据采集与处理,2001,16(2)：145-149.
[59]丁天怀,郏东耀.利用多颜色空间特征融合方法检测近似目标[J].清华大学学报(自然科学版),2006,46(2)：176-179.
[60]许鹤群,黄健.棉纤维成熟度的图象分析[J].中国纺织大学学报,1992,8(1)：78-83.
[61]王勇,沈明霞,姬长英.采摘期成熟棉花不同部位颜色识别分析[J].农业工程学报,2007,23(4)：183-185.
[62]王勇,姬长英.田间早期成熟棉花识别研究[J].江西农业学报,2006,18(6)：141-143.
[63]王勇,沈明霞,姬长英.基于颜色信息和形状特征的棉桃识别方法[J].农业机械学报,2007,38(11)：77-79,87.
[64]王勇,沈明霞,姬长英.基于色差信息的田间成熟棉花识别[J].浙江农业学报,2007,19(5)：385-388.
[65]禹素萍,曾培峰,陈建平.基于图像掩码的棉纤维截面分割[J].计算机工程,2007,33(13)：188-190.
[66]禹素萍,曾培峰,陈建平.棉纤维截面图像的自适应阈值Canny边缘检测[J].计算机工程,2007,33(11)：168-203.
[67]张教海,李亚兵,别墅,沈杰,夏松波,王孝刚.数字图像处理在棉花形态特征提取上的应用[J].湖北农业科学,2007,46(3)：369-371.
[68]何晓峰,刘文耀,徐守东.游离棉纤维长度的计算机测量[J].棉花学报,2003,15(6)：339-343.
[69]何晓峰,刘文耀,吴国新.棉纤维品质长度的图像测量[J].棉花学报,2005,17(6)：348-352.
[70]宋新平.应用微机图象处理技术检测纤维材料[J].纺织标准与质量,2000,(4)：25-26.
[71]胡宝林.从异性纤维状况谈清理技术措施[J].中国纤检,2006,(12)：51-52.
[72]张晓璐,韩力群.计算机视觉技术在纤维检验中的应用[J].北京工商大学学报(自然科学版),2005,23(2)：43-45.
[73]郏东耀,丁天怀.检测皮棉内部杂质的透射成像影响因素分析[J].农业机械学报,2005,36(2)：65-69.
[74]付江,于伟东.国外棉花含杂量检测及相关测量[J].中国纤检,2006,(8)：43-48.
[75]Barker R L, Lyons D W. Instrumented procedures for analysis of non-lint particles in cotton[J]. Journal of Engnieering for Industry,1976,98 (3):845-848.
[76]Lyons D W, Barker R L. Comparative analysis of trash contamination in cotton by optical scanning[J]. Journal of Engineering for Industry,1977,99 (1):46-50.
[77]Taylor R A. Using high-speed image analysis to estimate trash in cotton[J]. Journal of Engineering for Industry,1985,107(2):206-219.
[78]Taylor R A, Godbey L C. Using digital images to measure and discriminate small particles in cotton[C]. Proceedings of SPIE-The International Society for Optical Engineering,1991:16-27.
[79]Taylor R A. Measuring leaf material in ginned cotton from surface images[C]. Proceedings of SPIE-The International Society for Optical Engineering,1995:218-225.
[80]Taylor R A用视频扫杂质分析仪测定棉花杂质[J].冯平译.中国纤检,1992,(1)：24-28.
[81]Xu B, Fang C Y, Huang R, Watson M D. Chromatic image analysis for cotton trash and color measurement[J]. Textile Research Journal,1997,67 (12):881-890.
[82]Veit D, Hormes I, Bergmann J, Wulfhorst B. Image processing as a tool to improve machine performance and process control[J]. International Journal of Clothing Science and Technology, 1996,8(1-2):66-72.
[83]Xu B, Fang C, Watson M D. Clustering analysis for cotton trash classification[J]. Textile Research Journal,1999,69(9):656-662.
[84]Gururajan A, Sarraf H S, Hequet E F. Statistical modeling, detection, and segmentation of stains in digitized fabric images[C]. Proceedings of SPIE-The International Society for Optical Engineering-Machine vision applications in industrial inspection,2007, Vol.6503.
[85]Dogan M S, Sarraf H S, Hequet E F. Cotton trash assessment in radiographic X-ray images with scale-space filtering and stereo analysis[C]. Proceedings of Machine Vision Applications in Industrial Inspection XIII,2005, Vol.5679:276-287.
[86]Kang T J, Kim S C. Objection evaluation of the trash and color of raw cotton by image processing and neural network [J]. Textile Research Journal,2002,36(11):124-128.
[87]Tilocca A, Borzone P, Carosio S, Durante A. Detecting fabric defects with a neural network using two kinds of optical patterns[J]. Textile Research Journal,2002,72 (6):545-550.
[88]Pelletier M G, Barker G L, Baker R V. Image processing for stripper harvested cotton trash content measurement a progress report[C]. Proceedings Beltwide Cotton Conferences,1999,2:1412-1415.
[89]Lieberman M A, Bragg C K, Brennan S N. Determining gravimetric bark content in cotton with machine vision[J]. Textile Research Journal,1998,68(2):94-104.
[90]Siddaiah M N, Lieberman M, Hughs S, Foulk J. Trash measurements in ginned cotton[C]. Proceedings Beltwide Cotton Conferences,2005:1926-1937.
[91]Siddaiah M N, Lieberman M, Hughs S, Foulk J. Comparison of small trash measurements between imaging techniques and AFIS[C]. Proceedings Beltwide Cotton Conferences,2005:2362-2372.
[92]Siddaiah M N, LAS C N, Lieberman M, Hughs S. High speed trash measurements[C]. Proceedings Beltwide Cotton Conferences,2003:1-10.
[93]Frydrych R, Krifa M, Tamime O, Giner M, Gourlot J P. Detection and counting of two cotton contaminants:seed coat fragments and honeydew depostts[C]. Proceedings Beltwide Cotton Conferences,1999,1:695-698.
[94]Jia D Y, Ding T H. Detection of foreign materials in cotton using a multi-wavelength imaging method[J]. Meas Sci Technol,2005,16:1355-1362.
[95]Fitzgerald G J, Maas S J, Detar W R. Early detection of spider mites in cotton using multispectral remote sensing[C]. Proceedings Beltwide Cotton Conferences,1999,2:1022-1023.
[96]Pavani S K, Mehmet S D, Hamed S S, Eric F H. Segmentation and classification of four common cotton contaminants in X-Ray micro-tomographic images[C]. Conference on Machine Vision Applications in Industrial Inspection,2004:1-13.
[97]Anthony W S. Comparison of video color/trash meters for measurements on cotton[J]. Apllied Engineering in Agriculture,1989,5 (1):18-23.
[98]Anthony W S. Online assessment of foreign matter in cotton during ginning[J]. Applied Engineering in Agriculture,1990,6 (3):267-272.
[99]Farah B D. Image processing in the measurement of trash contents and grades in cotton[C]. Proceedings IWISP, Manchester, United Kingdom,1996:629-632.
[100]Jensen K L, Carstensen M. Fuzz and pills evaluated on knitted textiles by image analysis[J]. Textile Research Journal,2002,72(1):34-50.
[101]Pelletier M G. High speed machine vision sensing of cotton lint trash[C]. Cotton Production & Processing Research Unit,2007:1-13.
[102]Tantaswadi P, Vilainatre J, Tamaree N, Viraivan P. Machine vision for automated visual inspection of cotton quality in textile industries using color isodiscrimination contour[J]. Computers & Industrial Engineering.1999,37(1-2):347-350.
[103]Brainard, D H. Colorimeter[M]. Bass M, Editor, Optical Society of America:Handbook of Optics, 1994, Vol.1, pp.26.1-26.54.
[104]丁天怀,李勇,苗君哲,郑颖航.基于BP神经网络的皮棉杂质在线检测方法[J].农业工程学报,2003,19(2)：137-139.
[105]苗君哲.用图象处理法在线检测棉花杂质的试验研究[D].北京：清华大学,2000,20-65.
[106]冯显英,任长志,黄燕云.基于机器视觉的异性纤维检测系统[J].山东大学学报(工学版), 2006,36(4)：5-8.
[107]郭彩霞,袁建畅,金守峰.基于BP网络的棉花中异性纤维识别算法[J].西安工程科技学院学报,2006,20(5)：542-544.
[108]符宝鼎,袁建畅,郭彩霞.基于RGB颜色模型棉花杂质的识别算法[J].北京纺织,2005,26(5)：48-51.
[109]符宝鼎.异性纤维识别算法的研究[D].西安：西安工程科技学院,2006：1-75.
[110]罗德坡,朱邦太,李勋.紫外线荧光效应及其在棉花异性纤维分拣系统中的应用[J].河南科技大学学报(自然科学版),2007,28(2)：63-66.
[111]郏东耀,丁天怀.纤维红外吸收特性及其在皮棉杂质检测中的应用[J].农业工程学报,2004,20(3)：104-108.
[112]郏东耀,丁天怀.棉花中异性纤维的多光谱检测[J].清华大学学报(自然科学版),2005,45(2)：193-196.
[113]郏东耀,丁天怀.利用纤维红外吸收特性的皮棉杂质检测新方法[J].红外与毫米波学报,2005,24(2)：147-150.
[114]金守峰.异性纤维检测装置的研究[D].西安：西安工程科技学院,2005,1-60
[115]金守峰,袁建畅,张慧.基于Matlab的棉花异性纤维检测及定位算法[J].北京纺织,2004,25(6)：48-50.
[116]金守峰,袁建畅,张慧Matlab在棉纤维杂质检测中的应用[J].现代纺织技术,2004,12(3)：5-7.
[117]王宏坡.基于DSP的原棉杂质实时检测系统[D].天津：天津工业大学,2005：1-62.
[118]吕宇涛.异性纤维检测系统研究及硬件实现[D].北京：北京工业大学,2004：1-71.
[119]李碧丹,丁天怀,郏东耀.皮棉异性纤维剔除系统设计[J].农业机械学报,2006,37(1)：107-110.
[120]常蕾.基于DSP的棉花异性纤维检测系统[D].合肥：合肥工业大学,2006：1-43.
[121]赵鹏.异性纤维捡出机系统及其软件算法的实现[D].曲阜：曲阜师范大学,2006：1-45.
[122]张海容.棉花异性纤维挑拣技术研究[D].合肥：合肥工业大学,2006：1-35.
[123]姚远.异性纤维检测装置基本算法及可行性研究[D].西安：西安工程科技学院,2002：1-56.
[124]张慧,袁建畅,金守峰.棉花杂质检测定位算法的研究[J].计量与测试技术,2004,31(12)：44-45.
[125]艾世一.计算机图像识别技术在异纤在线检测中的应用[D].天津：天津工业大学,2004：1-67.
[126]艾世一,王跃存,刘迪.基于MATLAB的图像识别技术的原棉异纤在线检测[J].天津工业大学学报,2004,23(2)：70-73.
[1]Underwood S A, Aggarwal J K. Interactive computer analysis of aerial color infrared photographs[J]. Computer Graphics and Image Processing,1977,6 (1):1-24.
[2]Kurugollu F, Sankur B, Harmanci A E. Color image segmentation using histogram multith resholding and fusion[J]. Image and Vision Computing,2001,19(13):915-928.
[3]Celenk M. A color clustering technique for image segmentation[J]. Computer Vision, Graphics and Image Processing,1990,52(2):145-170.
[4]Ahuja N, Rosenfeld A, Haralick R M. Neighbor gray levels as features in pixel classification[J]. Pattern Recognition,1980,13(4):251-260.
[5]刘重庆,程华.分割彩色图像的一种有效聚类方法[J].模式识别与人工智能,1995,8(S1)：133-138.
[6]阴国富.基于阈值法的图像分割技术[J].现代电子技术,2007,(23)：107-108.
[7]Nevatia R. A color edge detector and its use in scene segmentation[J]. IEEE Transactions on Systems, Man and Cybernetics,1977,7(11):820-826.
[8]Ohta Y I, Kanade T, Sakai T. Color information for region segmentation[J]. Computer Graphics and Image Processing,1980,13(3):222-241.
[9]Ohlander R, Price K, Reddy D R. Picture segmentation using a recursive region splitting method[J]. Computer Graphics and Image Processing,1978,8(3):313-333.
[10]柳萍,阳爱民.一种基于区域的彩色图像分割方法[J].计算机工程与应用,2007,43(6)：37-39,64.
[11]潘红艳,敬忠良,肖刚,郭超.融合颜色和区域信息的彩色图像分割新方法[J].计算机工程与应用,2007,43(25)：72-74.
[12]Vincent L, Soille P. Watersheds in digital spaces:an efficient algorithm based on immersion simulations[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1991,13(6): 583-598.
[13]Lezoray O, Cardot H. Cooperation of color pixel classification schemes and color watershed:a study for microscopic images[J]. IEEE Transactions on Image Processing,2002,11(7):783-789.
[14]Huang C L, Cheng T Y, Chen C C. Color images segmentation using scale space filter and Markov random field[J]. Pattern Recognition,1992,25(10):1217-1229.
[15]Liu J, Yang Y H. Multiresolution color image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1994,16(7):689-700.
[16]黄玉明,徐光,叶培健.基于物理模型的彩色图像分割[J].自动化学报,1992,18(4)：421-429.
[17]Tizhoosh H R. Fuzzy image processing[M]. Berlin:Springer Verlag,1997.
[18]Carevic D, Caelli T. Region-based coding of color image using Karhunen-Loeve Transform[J]. Graphics Models and Image Processing,1997,59(1):27-38.
[19]王兴伟,沈兰荪,卫保国等.基于改进的K均值聚类和数学形态学的彩色眼科图像病灶分割[J].中国生物医学工程学报,2002,21(5)：443-448.
[20]Bezdek J C. Pattern recognition with fuzzy objective function algorithms[M]. New York:Plenum Press,1981.
[21]刘木华,赵杰文,张海东.基于模糊C均值聚类的牛肉图像中脂肪和肌肉区域分割技术[J].农业工程学报,2004,20(2)：161-163.
[22]朱伟华,曹其新.基于模糊彩色聚类方法的西红柿缺陷分割研究[J].农业工程学报,2003,19(3)：133-136.
[23]周天娟,张铁中,杨丽,赵金英.基于数学形态学的相接触草莓果实的分割方法及比较研究[J].农业工程学报,2007,23(9)：164-168.
[24]Ong S H, Yeo N C, Lee K H, Venkatesh Y V, Cao D M. Segmentation of color images using a two stage self-organizing network[J]. Image and Vision Computing,2002,20(4):279-289.
[25]Uchiyama T, Arbib M A. Color image segmentation using competitive learning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1994,16(12):1197-1206.
[26]常发亮,刘静,乔谊正.基于自组织神经网络的彩色图像自适应聚类分割[J].控制与决策,2006,21(4)：449-452.
[27]徐海荣,田国会,李晓磊.一种基于自组织神经网络的彩色图像分割方法[J].山东大学学报(工学版),2005,35(6)：67-70.
[28]黎宁,毛四新,李有福.用于彩色图像分割的有效特征分析[J].南京航空航天大学学报,2001,18(2)：206-212.
[29]石晓龙,许志榜,许进.基于自组织特征映射的图像分割算法研究[J].华中科技大学学报(自然科学版),2005,33(9)：13-25.
[30]况菲,王耀南,余洪山等.一种基于学习向量量化神经网络的图像分割方法[J].计算机工程与应用,2005,(14)：34-36.
[31]Huang C L. Parallel image segmentation using modified hopfield model[J]. Pattern Recognition Letters,1993,13(5):345-353.
[32]Campadelli P, Medici D, Schettini R. Color image segmentation using hopfield networks[J]. Image and Vision Computing,1997,15(3):161-166.
[33]Lescure P, Yedid V M, Dupoisot H, Stamon G. Color segmentation on biological microscope images[C]. Proceedings of SPIE-The International Society for Optical Engineering,1999, Vol. 3647:182-193.
[34]王勇,沈明霞,姬长英.采摘期成熟棉花不同部位颜色识别分析[J].农业工程学报,2007,23(4)：183-185.
[35]Ohta Y I, Kanade T, Sakai T. Color information for region segmentation[J]. Computer Graphics and Image Processing,1980,13(3):222-241.
[36]林开颜,吴军辉,徐立鸿.彩色图像分割方法综述[J].中国图像图形学报,2005,10(1)：1-10.
[37]Gonzalez R C, Woods R E, Eddins S L数字图像处理[M]. Matlab版.阮秋琦等译.北京：电子工业出版社,2005：144-178.
[38]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004,83-109,451-454.
[39]Macadam D L. Visual sensitivites to color differences in daylight[J]. Journal of the Optical Society of America,1942,32(5):247-274.
[40]Hastie T, Tibshirani R, Friedmar J统计学习基础——数据挖掘、推理与预测[M].北京：电子工业出版社,2004,135-150,305-356.
[41]张敏,于剑.基于划分的模糊聚类算法[J].软件学报,2004,15(6)：858-868.
[42]Cheng H D, Jiang X H, Sun Y, Wang J. Color image segmentation:advances and prospects[J]. Pattern Recognition,2001,34(12):2259-2281.
[43]Li N, Li Y F. Feature encoding for unsupervised segmentation of color images[J]. IEEE Transactions on Systems, Man and Cybernetics,2003,33(3):438-447.
[44]Sergios T, Konstantinos K模式识别[M].2版.北京：电子工业出版社,2004：257-283.
[45]徐飞,施晓红Matlab应用图像处理[M].西安：西安电子科技大学出版社,2002：101-214.
[46]陈杨,陈荣娟,郭颖辉Matlab 6. X图形编程与图像处理[M].西安：西安电子科技大学出版社,2002：162-187.
[47]蔡健荣,周小军,李玉良,范军.基于机器视觉自然场景下成熟柑橘识别[J].农业工程学报,2008,24(1)：175-178.
[1]陈友,程学旗,李洋,戴磊.基于特征选择的轻量级入侵检测系统[J].软件学报,2007,18(7)：1639-1651.
[2]毛勇,周晓波,夏铮,尹征,孙优贤.特征选择算法研究综述[J].模式识别与人工智能,2007,20(2)：211-218.
[3]崔自峰,徐宝文,张卫丰,徐峻岭.一种近似Markov Blanket最优特征选择算法[J].计算机学报,2007,30(12)：2074-2081.
[4]付峰,应义斌.生物图像阈值分割方法的研究[J].浙江大学学报(农业与生命科学版),2003,29(1)：108-112.
[5]Webb A R.统计模式识别[M].2版.北京：电子工业出版社,2004：7-30&342-346.
[6]Kuhl F P, Giardina C R. Elliptic Fourier features of a closed contour[J]. Computer Graphics and Image Processing,1982,18(3):236-258.
[7]Lin C S, Hwang C L. New forms of shape invariants from elliptic Fourier descriptors[J]. Pattern Recognition,1987,20(5):535-545.
[8]李秀智.基于机器视觉的苹果形状分级系统研究[D].南京：南京农业大学,2003：17-21.
[9]Taxt T, Olafsdottir J B, Daechlen M. Recognition of hand written symbols[J]. Pattern Recognition, 1990,23(11):1155-1166.
[10]Fu K S, Persoon E. Shape discrimination using Fourier descriptors[J]. IEEE Trans. Pattern Anal Mach. Intellig.1986,8(3):388-397.
[11]Pitas I. Digital image processing algorithms and applications[M]. Prentice Hall,1992:400-413.
[12]Sergios T, Konstantinos K模式识别[M].2版.北京：电子工业出版社,2004：257-283.
[13]黄心晔,王茂祥,富煜清,陆佶人.基于结构分析的手写体数字识别算法[J].电子工程师.1999,11(5)：23-26.
[14]张济忠.分形[M].北京：清华大学出版社,1995：111-141.
[1]李宁,刘东波,臧英明.中国棉花分级标准与国外棉花分级标准差异的研究[J].大连轻工业学院学报,2001,20(4)：309-312.
[2]《棉花仪器化检验》(草案)起草小组.棉花品级与色特征级的辩证关系[J].中国纤检,2005,(5)：5-7.
[3]中国纤维检验局棉花质量监督处.《棉花细绒棉》新标准系列问答：棉花品级、长度18问[J].高端·视点.2007,(8)：12-15.
[4]熊宗伟.我国棉花纤维质量及颜色等级划分研究[D].北京：中国农业大学,2005：10-25.
[5]熊宗伟.棉花色特征[J].中国棉花,1995,22(4)：39-40.
[6]郑芝奖.我国棉花品级与棉花色征指标的关系[J].纺织学报,2001,22(2)：23-24.
[7]马兆仁,邓明辉,詹建斌,周献珠,陈贺,江洪.新疆棉花色泽特征、马克隆值及品质分析[J].中国纤检,2006,(5)：26-29.
[8]Kang T J, Kim S C. Objective evaluation of the trash and color of raw cotton by image processing and neural network[J]. Textile Research Journal,2002,72(9):776-782.
[9]Forsyth D A, Ponce J计算机视觉：一种现代方法[M].北京：电子工业出版社,2004：83-109,451-454.
[10]Gonzalez R C, Woods R E, Eddins S L数字图像处理[M]. Matlab版.阮秋琦等译.北京：电子工业出版社,2005：144-178.
[11]Webb A R统计模式识别[M].2版.北京：电子工业出版社,2004：7-30&342-346.
[12]Cantoni E, Field C, Flemming J M, Ronchetti E. Longitudinal variable selection by cross-validation in the case of many covariates[J]. Statistics in Medicine,2007,26(4):919-930.
[13]Trevor H, Robert T, Jerome F.统计学习基础——数据挖掘、推理与预测[M].北京：电子工业出版社,2004：135-150,305-356.