玉米病害图像识别系统的设计与实现
|
摘要
玉米是重要的粮食作物之一,在我国玉米的播种面积很大,分布也很广,由于玉米病害的影响,造成玉米减产严重,品质下降。传统的农作物病害诊断检测方法由于技术、人力的不足,不仅费时费力,而且获取信息的滞后性还严重影响病害诊断的准确性。针对玉米病害识别自动化程度不高,识别诊断不及时的问题,本文应用图像识别技术对玉米的病害识别诊断进行了研究,针对常见的玉米病害种类为研究对象,采用了支持向量机(SVM)这种可行的方法进行识别诊断,识别精度较高。为玉米病害自动识别与诊断的相关研究做出了贡献。本文的研究工作主要包括以下几个方面:
(1)以玉米生产中常见的几种危害比较严重的病害(玉米大斑病、玉米灰斑病和玉米小斑病)为研究对象,在自然光条件下,通过数码相机采集了玉米病害的图片,采用相关图像处理软件进行了统一处理,得到了病害图像识别实验样本;首先对获取的原始病害图像进行自适应中值滤波平滑处理,这种算法可以很好的过滤噪声,同时也能保持目标的边缘清晰;然后采用超绿分割特征和Ostu自动阀值分割法来分割图像,超绿特征是植物所特有的一种图像特征,利用超绿特征先将图像进行灰度变换,再利用Ostu自动阀值分割法通过调整进行图像分割,能够很好的分离病斑图像部分;最后利用图像的开闭运算去除冗余斑点;从而最终提取了病斑图像。本文对于上述图像预处理过程进行深入分析和研究,通过对噪声过滤、图像分割、数学形态学等图像处理技术的分析,并结合玉米病害的特点选择了合适的算法,结果表明这些算法适合于米病害图像预处理操作,为后续特征提取和病害识别工作奠定了基础。
(2)在特征提取部分,结合玉米病害图像的特点,本文将图像从RGB颜色空间转化到HSI颜色空间,由于图像颜色分布信息主要集中在颜色矩中低阶矩里面,所以本文选用颜色特征的一阶、二阶和三阶矩来反映图像的颜色分布特征。提取了图像的颜色特征。利用灰度共生矩阵方法提取纹理特征。算法提取的特征维数较低,计算量较小,算法具有较强的纹理分析能力。
(3)在病害识别部分,为了解决玉米病害图像样本数量少的特点,将支持向量机(SVM)方法应用于玉米病害图像识别诊断方面,由于支持向量机(SVM)利用了结构风险最小化原理,同时也兼顾训练误差和泛化能力,它对于解决小样本、非线性高维、局部极小值等模式识别问题有着特有的优势。本文对支持向量机(SVM)方法理论进行详细介绍,在SVM分类器设计方面,采用了基于模糊方法的“一对多”SVM分类器,很好的解决了可能出现的不可识别区问题。试验结果表明:将支持向量机(SVM)方法对灰斑病,大斑病和小斑病三种玉米叶部病害进行分类识别。对病害的识别率达到了83%以上,识别玉米病害类别精度较高。
(4)在上述研究基础上,本文利用C#语言和GDI+技术,在Visual Studio.NET 2008平台下开发研制了玉米病害图像识别系统,在文中,针对系统功能和各模块设计进行了详细的介绍。
(5)最后对整篇论文的主要工作做了总结,并对未来的玉米病害识别研究方向做了展望,指出了研究应在特征选择方面和自动化方面继续加强完善,同时以后应扩展研究的应用范围,譬如可以扩展到多部位综合识别诊断玉米病害,也可以扩展到虫、草等灾害识别防治方面。
本文在玉米病害的识别诊断方面引入了图像识别技术,发展并壮大了图像识别技术的应用范围,为图像识别技术在农业领域的应用提供了借鉴。
Maize is one of the important food crops, a large acreage of maize in China, the distribution is very broad. Maize disease affect the output and quality of crop badly.The traditional crop disease diagnostic test for technical and manpower shortage, not only time consuming, and access to information lag also seriously affect the accuracy of disease diagnosis. This paper researched maize disease diagnosis by the image recognition technology , it show feasible and diagnostic accuracy by this study of maize leaf diseases. It provides a theoretical basis for the automatic identification and diagnosis of crop pests related research . This research work includes the following:
(1) Maize production in several common diseases (maize leaf blight, gray leaf spot disease and maydis) as the research object in this paper. Natural light conditions, digital camera capture images of maize diseases, These images were processed for a unified operation by associated image processing software, in order to get diseased maize samples. Firstly, adaptive median filtering is performed to remove noises in images. This algorithm can filter noise, while maintaining the edge of clear objectives. Secondly, it was used by ultra-green feature and Ostu threshold segmentation method to segment images, Ultra-green features a unique image of plant characteristics, characteristics of first use of super-green image gray-scale transformation, re-use Ostu automatic threshold segmentation method for image segmentation by adjusting, to a good separation of the image part of the lesion. Finally, the image of the opening and closing operation to remove redundant spots; to eventually extract the lesion image. This paper analysis and research image pre-processing process. Through the noise filtering, image segmentation, image processing techniques such as mathematical morphology analysis, combined with the characteristics of maize diseases selected the image processing algorithms, The results show that the algorithms for image pre-processing operation are foundation of follow-up feature extraction and maize disease identification.
(2) In the feature extraction part, combined with the characteristics of the image of maize diseases, this paper transform the image from RGB color space to HIS color space, As the image color distribution of the information concentrated in the low moments in which color moment, so this paper we use the color characteristics of the first, second and third moment to reflect the image of the color distribution. Color features were extracted from the image. This paper using gray level co-occurrence matrix to extract texture features. The algorithm has strong texture analysis and less computation.
(3) In order to solve a small number of image samples of maize diseases , This paper proposed the support vector machine theory is applied to maize disease pattern recognition, the support vector machine (SVM) using the structural risk minimization principle, while taking into account training errors and generalization ability, it is to solve the small sample, nonlinear high-dimensional, local minimum problems has a unique advantage. This paper described in detail in the design of SVM classifier, and using a method based on fuzzy "one to many" SVM classifier, It solved the problem of a possible non-recognition zone. The results showed that: by vector machine (SVM) method ,recognition rate of the maize disease 83%, and identify categories of high precision maize diseases.
(4) The basis of these studies, By using C # language and GDI + technology, Visual Studio.NET 2008 platform, developed a maize disease image recognition system. This paper described functions and modules for system design in detail.
(5) Finally, the entire paper summarizes the major work done, and the future identification of maize disease research are put forward.
In this paper, image recognition technology used in the identification of maize disease diagnosis, image recognition technology to expand the application scope of image recognition technologies for applications in agriculture provide a reference.
(1)以玉米生产中常见的几种危害比较严重的病害(玉米大斑病、玉米灰斑病和玉米小斑病)为研究对象,在自然光条件下,通过数码相机采集了玉米病害的图片,采用相关图像处理软件进行了统一处理,得到了病害图像识别实验样本;首先对获取的原始病害图像进行自适应中值滤波平滑处理,这种算法可以很好的过滤噪声,同时也能保持目标的边缘清晰;然后采用超绿分割特征和Ostu自动阀值分割法来分割图像,超绿特征是植物所特有的一种图像特征,利用超绿特征先将图像进行灰度变换,再利用Ostu自动阀值分割法通过调整进行图像分割,能够很好的分离病斑图像部分;最后利用图像的开闭运算去除冗余斑点;从而最终提取了病斑图像。本文对于上述图像预处理过程进行深入分析和研究,通过对噪声过滤、图像分割、数学形态学等图像处理技术的分析,并结合玉米病害的特点选择了合适的算法,结果表明这些算法适合于米病害图像预处理操作,为后续特征提取和病害识别工作奠定了基础。
(2)在特征提取部分,结合玉米病害图像的特点,本文将图像从RGB颜色空间转化到HSI颜色空间,由于图像颜色分布信息主要集中在颜色矩中低阶矩里面,所以本文选用颜色特征的一阶、二阶和三阶矩来反映图像的颜色分布特征。提取了图像的颜色特征。利用灰度共生矩阵方法提取纹理特征。算法提取的特征维数较低,计算量较小,算法具有较强的纹理分析能力。
(3)在病害识别部分,为了解决玉米病害图像样本数量少的特点,将支持向量机(SVM)方法应用于玉米病害图像识别诊断方面,由于支持向量机(SVM)利用了结构风险最小化原理,同时也兼顾训练误差和泛化能力,它对于解决小样本、非线性高维、局部极小值等模式识别问题有着特有的优势。本文对支持向量机(SVM)方法理论进行详细介绍,在SVM分类器设计方面,采用了基于模糊方法的“一对多”SVM分类器,很好的解决了可能出现的不可识别区问题。试验结果表明:将支持向量机(SVM)方法对灰斑病,大斑病和小斑病三种玉米叶部病害进行分类识别。对病害的识别率达到了83%以上,识别玉米病害类别精度较高。
(4)在上述研究基础上,本文利用C#语言和GDI+技术,在Visual Studio.NET 2008平台下开发研制了玉米病害图像识别系统,在文中,针对系统功能和各模块设计进行了详细的介绍。
(5)最后对整篇论文的主要工作做了总结,并对未来的玉米病害识别研究方向做了展望,指出了研究应在特征选择方面和自动化方面继续加强完善,同时以后应扩展研究的应用范围,譬如可以扩展到多部位综合识别诊断玉米病害,也可以扩展到虫、草等灾害识别防治方面。
本文在玉米病害的识别诊断方面引入了图像识别技术,发展并壮大了图像识别技术的应用范围,为图像识别技术在农业领域的应用提供了借鉴。
Maize is one of the important food crops, a large acreage of maize in China, the distribution is very broad. Maize disease affect the output and quality of crop badly.The traditional crop disease diagnostic test for technical and manpower shortage, not only time consuming, and access to information lag also seriously affect the accuracy of disease diagnosis. This paper researched maize disease diagnosis by the image recognition technology , it show feasible and diagnostic accuracy by this study of maize leaf diseases. It provides a theoretical basis for the automatic identification and diagnosis of crop pests related research . This research work includes the following:
(1) Maize production in several common diseases (maize leaf blight, gray leaf spot disease and maydis) as the research object in this paper. Natural light conditions, digital camera capture images of maize diseases, These images were processed for a unified operation by associated image processing software, in order to get diseased maize samples. Firstly, adaptive median filtering is performed to remove noises in images. This algorithm can filter noise, while maintaining the edge of clear objectives. Secondly, it was used by ultra-green feature and Ostu threshold segmentation method to segment images, Ultra-green features a unique image of plant characteristics, characteristics of first use of super-green image gray-scale transformation, re-use Ostu automatic threshold segmentation method for image segmentation by adjusting, to a good separation of the image part of the lesion. Finally, the image of the opening and closing operation to remove redundant spots; to eventually extract the lesion image. This paper analysis and research image pre-processing process. Through the noise filtering, image segmentation, image processing techniques such as mathematical morphology analysis, combined with the characteristics of maize diseases selected the image processing algorithms, The results show that the algorithms for image pre-processing operation are foundation of follow-up feature extraction and maize disease identification.
(2) In the feature extraction part, combined with the characteristics of the image of maize diseases, this paper transform the image from RGB color space to HIS color space, As the image color distribution of the information concentrated in the low moments in which color moment, so this paper we use the color characteristics of the first, second and third moment to reflect the image of the color distribution. Color features were extracted from the image. This paper using gray level co-occurrence matrix to extract texture features. The algorithm has strong texture analysis and less computation.
(3) In order to solve a small number of image samples of maize diseases , This paper proposed the support vector machine theory is applied to maize disease pattern recognition, the support vector machine (SVM) using the structural risk minimization principle, while taking into account training errors and generalization ability, it is to solve the small sample, nonlinear high-dimensional, local minimum problems has a unique advantage. This paper described in detail in the design of SVM classifier, and using a method based on fuzzy "one to many" SVM classifier, It solved the problem of a possible non-recognition zone. The results showed that: by vector machine (SVM) method ,recognition rate of the maize disease 83%, and identify categories of high precision maize diseases.
(4) The basis of these studies, By using C # language and GDI + technology, Visual Studio.NET 2008 platform, developed a maize disease image recognition system. This paper described functions and modules for system design in detail.
(5) Finally, the entire paper summarizes the major work done, and the future identification of maize disease research are put forward.
In this paper, image recognition technology used in the identification of maize disease diagnosis, image recognition technology to expand the application scope of image recognition technologies for applications in agriculture provide a reference.
引文
[1]尹成杰.粮安天下-全球粮食危机与中国粮食安全[M].北京,中国经济出版社,2009.
[2]洪涛.农产品现代物流需进入快车道[J].中国物流与采购,2006, (7):26-30.
[3]安冈善文.图像处理技术在环境中的应用[C].电气学会杂志特集, 1985,455-458.
[4]Russ J C. The Image Processing Handbook[M].CRC Press,1995.
[5]H.D.Cheng, X.H.Jiang, Y.Sun, et al. Color image segmentation on: advances and prospect[J]. Pattern Recognition,2001,34(7):2259-2281.
[6]Zayas I, Pomeranz Y, Lai F S. Discrimination between arthur and arkan wheat by mage analysis[J]. Cereal Chemistry,1985.62(2):478-480.
[7]Yutaka SASAKI,Tsuguo OKAMOTO,Kenji IMOU,Toru TOR.Automatic Diagnosis of Plant Disease[J].Journal of JSAM.1999.61(2):119-126.
[8]Yutaka SASAKI, Masato SUZUKI. Construction of the Automatic Diagnosis System of Plant Disease Using Genetic Programming Which Paid Its Attention to Variety[C]. ASAE Meeting Presentation(paper No.031049),2003.
[9]El-Helly M,El-Beltagy S.,and Rafea A. Image analysis based interface for diagnostic expert systems[A].Proceedings of the Winter international synposium on information and Communication Technologies[C].Trinity College Dublin,2004,1-6.
[10]Sammany, M, Mohammed El-Beltagy. Optimizing Neural Networks Architecture and Parameters Using Genetic Algorithms for diagnosing Plant Diseases[A].Proceeding of and International Computer Engineering Conference[C],IEEE(Egypt section),2006.
[11]毛罕平,徐贵力,李萍萍.番茄缺素叶片的图像特征提取和优化选择研究[J].农业工程学报,2003,19(2):133-136.
[12]毛罕平,徐贵力,李萍萍.基于计算机视觉的番茄营养元素亏缺的识别[J].农业机械学报,2003,34(2):73-75.
[13]毛罕平,吴雪梅,李萍萍.基于计算机视觉的番茄缺素神经网络识别[J].农业工程学报,2005,21(8):106-109.
[14]吴雪梅,毛罕平.计算机视觉描述缺素番茄叶片颜色变化的研究[J].农机化研究,2004,(5):7-90.
[15]徐贵力,毛罕平,李萍萍.缺素叶片彩色图像颜色特征提取的研究[J].农业工程学报,2002,18(4):150-154.
[16]田有文,张长水,李成华.支持向量机在植物病斑形状识别中的应用研究[J].农业工程学报,2004,20(3):134-136.
[15]田有文,张长水,李成华.基于支持向量机和色度矩的植物病害识别研究[J].农业机械学报,2004,35(3):95-98.
[16]田有文,王滨,唐晓明.基于纹理特征和支持向量机的玉米病害的识别[J].沈阳农业大学学报,2005,36(6):730-732.
[17]田有文,李成华.基于图像处理的日光温室黄瓜病害识别的研究[J].农机化研究,2006,(2):151-153,160.
[18]田有文,李天来,李成华等.基于支持向量机的葡萄病害图像识别方法[J].农业工程学报,2007,23(6):175-180.
[19]王双喜,董晓志,王旭.温室植物病害数字化处理中图像增强方法的研究[J].内蒙古农业大学学报,2007,28(3):15-18.
[20]张静,王双喜,董晓志等.基于温室植物叶片纹理的病害图像处理及特征值提取方法的研究[J].沈阳农业大学学报,2006,37(3):282-285.
[21]张静,王双喜.温室植物病害图像处理技术中图像分割方法的研究[J].内蒙古农业大学学报,2007,28(3):19-21.
[22]王克如.基于图像识别的作物病虫草害诊断研究[D]:[博士].北京:中国农业科学院作物科学研究所,2005.
[23]赵玉霞,王克如,白中英等.贝叶斯方法在玉米叶部病害图像识别中的应用[J].计算机工程与应用,2007,43(5):193-195.
[24]赵玉霞,王克如,白中英等.基于图像识别的玉米叶部病害诊断研究[J].中国农业科学,2007,40(4):698-703.
[25]齐龙.基于图像处理的作物病害诊断及叶片形态参数测量技术的研究[D].吉林:吉林大学生物与农业工程学院,2006.
[26]陈学佺.数字图像分析[M].合肥:中国科学技术大学,2004.
[27]阮秋琦.数字图像处理学[M].北京:电子工业出版社,2001.
[28]王晓峰.植物叶片图像自动识别系统的研究与实现[D]:[硕士].合肥:中科院合肥智能机械研究所,2005.
[29][美]桑肯等.图像处理分析与机器视觉(第二版X英文版) [M].北京:人民邮电出版社,2002.
[30]Ostu N.Diserlminant and least square threshold seleetion[J].In:Proc 4llCPR,1978:592-596.
[31]D.M.Woebbecke, G.E.Meyer, et al. Color indices for weed identification under various soil,residue, and lighting conditions[J]. Transactions of the ASAE. 1995,38(1):259-269.
[32]Gpmza;ez R C,Woods R E. Digital image processing[M]. Addison-wesley Publishing Co.Inc,1992.
[33]孙明等.基于计算机视觉的萝卜幼苗白动识别技术[J].农业机械学报2002,33(5):75-77.
[34]吕朝辉等.用BP神经网络进行秧苗图像分割[J].农业工程学报,2002,117(3):146-148
[35]何斌.数字图像处理[M].北京:人民邮电出版社,2001。
[36]章毓晋.图像工程(上册)图像处理与分析[M].北京:清华大学出版社,1991.
[37]章毓晋.图像工程(下册)图像处理与分析[M].北京:清华大学出版社,1999
[38]朱志刚.数字图像处理[M].北京:电子工业出版社,1998
[39]杨淑莹.图像模式识别[M].北京:清华大学出版社,2006
[40]李金屏,吴波,王英姿.基于纹理分析的图像形貌特征提取[J].济南大学学报(自然科学版),2004.18(3):217-221.
[41]李杰.基于支持向量机的遗传算法的纹理识别[J].四川大学学报(工程科学版),2005.37(4):104-108.
[42]张宏林.图像模式识别工程实践[M].北京:人民邮电出版社,2003.
[43]冈萨雷斯.数字图像处理.第二版中文版[M].北京:电子工业出版社
[44]边肇祺,张学工.模式识别[M].第二版.北京:清华大学出版社,1999.
[45]Vapnik V N. The Nature of Statistical Learning Theory[M]. New York:Springer Verlag,1995.
[46]刘华富.支持向量机Mercer核的若干性质[J].北京联合大学学报,2005,(01):19-23
[47]林升梁,刘志基于RBF核函数的支持向量机参数选择[J].浙江工业大学学报,2007,(02):58-62
[48]任东.基于支持向量机的植物病害识别研究[D] :[博士].长春:吉林农业大学,2007
[49]李国正,王猛.支持向量机导论[M].北京:电子工业出版社,2004.
[2]洪涛.农产品现代物流需进入快车道[J].中国物流与采购,2006, (7):26-30.
[3]安冈善文.图像处理技术在环境中的应用[C].电气学会杂志特集, 1985,455-458.
[4]Russ J C. The Image Processing Handbook[M].CRC Press,1995.
[5]H.D.Cheng, X.H.Jiang, Y.Sun, et al. Color image segmentation on: advances and prospect[J]. Pattern Recognition,2001,34(7):2259-2281.
[6]Zayas I, Pomeranz Y, Lai F S. Discrimination between arthur and arkan wheat by mage analysis[J]. Cereal Chemistry,1985.62(2):478-480.
[7]Yutaka SASAKI,Tsuguo OKAMOTO,Kenji IMOU,Toru TOR.Automatic Diagnosis of Plant Disease[J].Journal of JSAM.1999.61(2):119-126.
[8]Yutaka SASAKI, Masato SUZUKI. Construction of the Automatic Diagnosis System of Plant Disease Using Genetic Programming Which Paid Its Attention to Variety[C]. ASAE Meeting Presentation(paper No.031049),2003.
[9]El-Helly M,El-Beltagy S.,and Rafea A. Image analysis based interface for diagnostic expert systems[A].Proceedings of the Winter international synposium on information and Communication Technologies[C].Trinity College Dublin,2004,1-6.
[10]Sammany, M, Mohammed El-Beltagy. Optimizing Neural Networks Architecture and Parameters Using Genetic Algorithms for diagnosing Plant Diseases[A].Proceeding of and International Computer Engineering Conference[C],IEEE(Egypt section),2006.
[11]毛罕平,徐贵力,李萍萍.番茄缺素叶片的图像特征提取和优化选择研究[J].农业工程学报,2003,19(2):133-136.
[12]毛罕平,徐贵力,李萍萍.基于计算机视觉的番茄营养元素亏缺的识别[J].农业机械学报,2003,34(2):73-75.
[13]毛罕平,吴雪梅,李萍萍.基于计算机视觉的番茄缺素神经网络识别[J].农业工程学报,2005,21(8):106-109.
[14]吴雪梅,毛罕平.计算机视觉描述缺素番茄叶片颜色变化的研究[J].农机化研究,2004,(5):7-90.
[15]徐贵力,毛罕平,李萍萍.缺素叶片彩色图像颜色特征提取的研究[J].农业工程学报,2002,18(4):150-154.
[16]田有文,张长水,李成华.支持向量机在植物病斑形状识别中的应用研究[J].农业工程学报,2004,20(3):134-136.
[15]田有文,张长水,李成华.基于支持向量机和色度矩的植物病害识别研究[J].农业机械学报,2004,35(3):95-98.
[16]田有文,王滨,唐晓明.基于纹理特征和支持向量机的玉米病害的识别[J].沈阳农业大学学报,2005,36(6):730-732.
[17]田有文,李成华.基于图像处理的日光温室黄瓜病害识别的研究[J].农机化研究,2006,(2):151-153,160.
[18]田有文,李天来,李成华等.基于支持向量机的葡萄病害图像识别方法[J].农业工程学报,2007,23(6):175-180.
[19]王双喜,董晓志,王旭.温室植物病害数字化处理中图像增强方法的研究[J].内蒙古农业大学学报,2007,28(3):15-18.
[20]张静,王双喜,董晓志等.基于温室植物叶片纹理的病害图像处理及特征值提取方法的研究[J].沈阳农业大学学报,2006,37(3):282-285.
[21]张静,王双喜.温室植物病害图像处理技术中图像分割方法的研究[J].内蒙古农业大学学报,2007,28(3):19-21.
[22]王克如.基于图像识别的作物病虫草害诊断研究[D]:[博士].北京:中国农业科学院作物科学研究所,2005.
[23]赵玉霞,王克如,白中英等.贝叶斯方法在玉米叶部病害图像识别中的应用[J].计算机工程与应用,2007,43(5):193-195.
[24]赵玉霞,王克如,白中英等.基于图像识别的玉米叶部病害诊断研究[J].中国农业科学,2007,40(4):698-703.
[25]齐龙.基于图像处理的作物病害诊断及叶片形态参数测量技术的研究[D].吉林:吉林大学生物与农业工程学院,2006.
[26]陈学佺.数字图像分析[M].合肥:中国科学技术大学,2004.
[27]阮秋琦.数字图像处理学[M].北京:电子工业出版社,2001.
[28]王晓峰.植物叶片图像自动识别系统的研究与实现[D]:[硕士].合肥:中科院合肥智能机械研究所,2005.
[29][美]桑肯等.图像处理分析与机器视觉(第二版X英文版) [M].北京:人民邮电出版社,2002.
[30]Ostu N.Diserlminant and least square threshold seleetion[J].In:Proc 4llCPR,1978:592-596.
[31]D.M.Woebbecke, G.E.Meyer, et al. Color indices for weed identification under various soil,residue, and lighting conditions[J]. Transactions of the ASAE. 1995,38(1):259-269.
[32]Gpmza;ez R C,Woods R E. Digital image processing[M]. Addison-wesley Publishing Co.Inc,1992.
[33]孙明等.基于计算机视觉的萝卜幼苗白动识别技术[J].农业机械学报2002,33(5):75-77.
[34]吕朝辉等.用BP神经网络进行秧苗图像分割[J].农业工程学报,2002,117(3):146-148
[35]何斌.数字图像处理[M].北京:人民邮电出版社,2001。
[36]章毓晋.图像工程(上册)图像处理与分析[M].北京:清华大学出版社,1991.
[37]章毓晋.图像工程(下册)图像处理与分析[M].北京:清华大学出版社,1999
[38]朱志刚.数字图像处理[M].北京:电子工业出版社,1998
[39]杨淑莹.图像模式识别[M].北京:清华大学出版社,2006
[40]李金屏,吴波,王英姿.基于纹理分析的图像形貌特征提取[J].济南大学学报(自然科学版),2004.18(3):217-221.
[41]李杰.基于支持向量机的遗传算法的纹理识别[J].四川大学学报(工程科学版),2005.37(4):104-108.
[42]张宏林.图像模式识别工程实践[M].北京:人民邮电出版社,2003.
[43]冈萨雷斯.数字图像处理.第二版中文版[M].北京:电子工业出版社
[44]边肇祺,张学工.模式识别[M].第二版.北京:清华大学出版社,1999.
[45]Vapnik V N. The Nature of Statistical Learning Theory[M]. New York:Springer Verlag,1995.
[46]刘华富.支持向量机Mercer核的若干性质[J].北京联合大学学报,2005,(01):19-23
[47]林升梁,刘志基于RBF核函数的支持向量机参数选择[J].浙江工业大学学报,2007,(02):58-62
[48]任东.基于支持向量机的植物病害识别研究[D] :[博士].长春:吉林农业大学,2007
[49]李国正,王猛.支持向量机导论[M].北京:电子工业出版社,2004.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |