计算机视觉技术在核桃壳、仁及分心木识别中的应用研究
|
摘要
[目的]为了研究快速准确识别核桃壳、仁及分心木的方法,[方法]本研究以太谷"清香"核桃为研究对象,搭建计算机视觉系统获取核桃壳、仁及分心木的图像信息,提取各样本的12个颜色特征值(RGB和HSI各分量的均值与方差),以及利用灰度共生矩阵(gray level co-occurrence matrix,GLCM)分别提取各样本能量、熵、惯性矩、相关性、逆差矩4个方向(0°,45°,90°,135°)共20个纹理特征参数。以颜色特征值、纹理特征值、颜色-纹理特征值融合作为输入建立3种最小二乘支持向量机(LS-SVM)模型(Y-LS-SVM、W-LS-SVM、Y-W-LS-SVM),并对各预测集样本进行判别。[结果]结果表明,Y-LS-SVM模型对核桃仁、分心木及核桃壳的判别准确率分别为91.1%、82.4%、96.6%;W-LS-SVM模型对核桃仁、分心木及核桃壳的判别准确率分别为100%、85.3%、84.7%;Y-W-LS-SVM模型对核桃仁、分心木及核桃壳的判别准确率分别为93.3%、97.1%、100%。Y-W-LS-SVM模型的判别效果最好。[结论]本研究表明,基于计算机视觉技术能够很好地对核桃壳、仁及分心木进行识别,为核桃壳、仁及分心木在线分选技术提供理论基础。
[Objective]Present study was aimed at developing a computer vision technology based method for rapid and accurate identification of walnut shell, kernel, and distraction wood. [Methods] Local walnut variety "Qingxiang" was selected as research material. The computer vision system was built up to obtain the image information of walnut shell, kernel and distraction wood. A total of 20 texture feature parameters including the energy, entropy, moment of inertia moment, relevance, and deficit four direction(0 °, 45 °, 90 °, 135 °) were extracted using 12 color eigenvalues(the mean and variance of each component of RGB and HSI of each sample) coupled with gray level co-occurrence matrix(GLCM),. Three least squares support vector machine(LS-SVM) models, named as Y-LS-SVM, W-LS-SVM, and Y-W-LS-SVM, were established based on color feature values, texture feature values, and color-texture feature values as inputs, and each prediction of a set of samples was discriminated. [Results] The results showed that the accuracy of Y-LS-SVM model for walnut kernel, distraction wood, and walnut shell was 91.1%, 82.4%, 96.6%, respectively; the accuracy of W-LS-SVM model for walnut kernel, distraction wood, and walnut shell was 100%, 85.3%,84.7%, respectively; and the accuracy of Y-W-LS-SVM model for walnut kernel, distraction wood and walnut shell was 93.3%, 97.1%, 100%, respectively. Therefore, the Y-W-LS-SVM model possessed the best discrimination effect. [Conclusion] This study indicated that the computer vision technology is able to successfully identify walnut shell, kernel and distraction wood, and the results provide theoretical basis for online sorting technology of walnut shell, kernel and distraction wood.
[Objective]Present study was aimed at developing a computer vision technology based method for rapid and accurate identification of walnut shell, kernel, and distraction wood. [Methods] Local walnut variety "Qingxiang" was selected as research material. The computer vision system was built up to obtain the image information of walnut shell, kernel and distraction wood. A total of 20 texture feature parameters including the energy, entropy, moment of inertia moment, relevance, and deficit four direction(0 °, 45 °, 90 °, 135 °) were extracted using 12 color eigenvalues(the mean and variance of each component of RGB and HSI of each sample) coupled with gray level co-occurrence matrix(GLCM),. Three least squares support vector machine(LS-SVM) models, named as Y-LS-SVM, W-LS-SVM, and Y-W-LS-SVM, were established based on color feature values, texture feature values, and color-texture feature values as inputs, and each prediction of a set of samples was discriminated. [Results] The results showed that the accuracy of Y-LS-SVM model for walnut kernel, distraction wood, and walnut shell was 91.1%, 82.4%, 96.6%, respectively; the accuracy of W-LS-SVM model for walnut kernel, distraction wood, and walnut shell was 100%, 85.3%,84.7%, respectively; and the accuracy of Y-W-LS-SVM model for walnut kernel, distraction wood and walnut shell was 93.3%, 97.1%, 100%, respectively. Therefore, the Y-W-LS-SVM model possessed the best discrimination effect. [Conclusion] This study indicated that the computer vision technology is able to successfully identify walnut shell, kernel and distraction wood, and the results provide theoretical basis for online sorting technology of walnut shell, kernel and distraction wood.
引文
[1]黄黎慧,黄群,孙术国,等.核桃的营养保健功能与开发利用[J].粮食科技与经济,2009,34(4):48-50.
[2]高海生,朱凤妹,李润丰.我国核桃加工产业的生产现状与发展趋势[J].经济林研究,2008,26(3):119-126.
[3]颜秉忠,王晓玲.基于计算机视觉技术大枣品质检测分级的研究[J].农机化研究,2018,40(8):232-235,268.
[4]胡孟晗,董庆利,刘宝林,等.基于计算机视觉的香蕉贮藏过程中颜色和纹理监测[J].农业机械学报,2013,44(8):180-184.
[5]申爱敏,霍晓静,王文娣,等.基于机器视觉的核桃仁大小自动分级技术[J].江苏农业科学,2016,44(9):383-385.
[6]赵海波,周向红.基于计算机视觉的番茄催熟与正常熟识别[J].农业工程学报,2011,27(2):355-359.
[7]Macho S, Rius A, Callao M P. Monitoring ethylene content in heterophasic copolymers by near-infrared spectroscopy: Standardisation of the calibration model [J]. Analytica Chimica Acta, 2001, 445(2): 213-220.
[8]贺东霞.数字图像去噪算法的研究与应用[D].延安:延安大学,2015.
[9]余洪.基于计算机视觉技术的茶叶品质分级研究[D].南昌:江西农业大学,2017.
[10]李晓斌.运用图像处理技术在线监测真空冻干果蔬含水率[D].太谷:山西农业大学,2013.
[11]潘磊庆,王振杰,孙柯,等.基于计算机视觉的稻谷霉变程度检测[J].农业工程学报,2017,33(3):272-280.
[12]孙俊,张跃春,毛罕平,等.基于计算机视觉的土壤镉胁迫生菜叶片污染响应分析[J].农业机械学报,2018,49(3):166-172.
[13]Zhao X, Burks T F, Qin J, et al. Digital microscopic imaging for citrus peel disease classification using color texture features[J]. Applied Engineering in Agriculture, 2009, 25(5):769-776.
[14]杨丹,赵海滨,龙哲,等. MATLAB图像处理实例详解[M].北京:清华大学出版社,2013:321-325.
[15]王斌,张淑娟,赵聪慧.壶瓶枣轻微损伤可见/近红外光谱动态判别模型研究[J].现代食品科技,2015,31(4): 258- 263.
[2]高海生,朱凤妹,李润丰.我国核桃加工产业的生产现状与发展趋势[J].经济林研究,2008,26(3):119-126.
[3]颜秉忠,王晓玲.基于计算机视觉技术大枣品质检测分级的研究[J].农机化研究,2018,40(8):232-235,268.
[4]胡孟晗,董庆利,刘宝林,等.基于计算机视觉的香蕉贮藏过程中颜色和纹理监测[J].农业机械学报,2013,44(8):180-184.
[5]申爱敏,霍晓静,王文娣,等.基于机器视觉的核桃仁大小自动分级技术[J].江苏农业科学,2016,44(9):383-385.
[6]赵海波,周向红.基于计算机视觉的番茄催熟与正常熟识别[J].农业工程学报,2011,27(2):355-359.
[7]Macho S, Rius A, Callao M P. Monitoring ethylene content in heterophasic copolymers by near-infrared spectroscopy: Standardisation of the calibration model [J]. Analytica Chimica Acta, 2001, 445(2): 213-220.
[8]贺东霞.数字图像去噪算法的研究与应用[D].延安:延安大学,2015.
[9]余洪.基于计算机视觉技术的茶叶品质分级研究[D].南昌:江西农业大学,2017.
[10]李晓斌.运用图像处理技术在线监测真空冻干果蔬含水率[D].太谷:山西农业大学,2013.
[11]潘磊庆,王振杰,孙柯,等.基于计算机视觉的稻谷霉变程度检测[J].农业工程学报,2017,33(3):272-280.
[12]孙俊,张跃春,毛罕平,等.基于计算机视觉的土壤镉胁迫生菜叶片污染响应分析[J].农业机械学报,2018,49(3):166-172.
[13]Zhao X, Burks T F, Qin J, et al. Digital microscopic imaging for citrus peel disease classification using color texture features[J]. Applied Engineering in Agriculture, 2009, 25(5):769-776.
[14]杨丹,赵海滨,龙哲,等. MATLAB图像处理实例详解[M].北京:清华大学出版社,2013:321-325.
[15]王斌,张淑娟,赵聪慧.壶瓶枣轻微损伤可见/近红外光谱动态判别模型研究[J].现代食品科技,2015,31(4): 258- 263.