基于机器视觉的玉米种粒破损识别方法研究
|
摘要
针对玉米种粒在收获、脱粒、贮藏时因各种因素造成损伤和人工选种耗时耗力的问题,提出了一种基于机器视觉的玉米种粒破损检测方法。首先,利用图像获取装置得到单粒玉米种粒图像,通过差影法确定图像噪声种类,采用中值滤波方法对图像进行降噪;其次,标记图像边界,运用灰度阈值法完成玉米种粒图像分割。根据玉米种粒的形态特征分别提取玉米种粒的周长、面积、周长面积比、长轴长、短轴长、长宽比6个几何特征和矩形度、圆形度、紧凑度、7个Hu不变矩10个形状特征,共16个特征。完整玉米种粒和破损玉米种粒图像各50幅作为训练样本,将提取的16个特征分量作为输入量,对支持向量机(Support Vector Machine,SVM)进行训练,输出量为1、0,分别代表"合格""不合格",训练完成后获得玉米种粒的SVM识别模型;另取完整的玉米种粒和破损的玉米种粒图像各50幅作为测试样本,对训练好的SVM模型进行测试。结果表明:该检测方法对玉米种粒破损识别准确率达95%以上,识别100幅玉米种粒图像的时间为1.27s。研究结果为玉米种粒的实时破损检测提供了参考。
Aim at the problems that corn seeds are damaged by various factors during harvest,threshing. And storage.Selecting is time-consuming and exhausting. Put forward a corn seed damage test method based on the machine vision.The first step is using the image acquisition device to grain the single particle corn seed image. Using the subtraction method to determine the type of the image noise and Median filtering method is used to denoise the image. And then marking the image boundaries and using fixed threshold method to split corn seed image segmentation. According to the morphological characteristics of corn seeds extract 6 geometric characters including perimeter,proportion,perimeter-proportion ratio,long axis length,minor axis length,length-width ratio,and another 10 morphological characters including rectangular degree,circularity,compactness 7 Hu invariant moment,a total of 16 characters. The extracted 16 characters components are used as inputs,50 whole corn seed and 50 corn seeds damaged were used as training samples to train SVM,output is "qualified","unqualified"two types and an identification model of corn seeds was obtained.Another complete corn seeds and broken corn seeds each containing 50 grains,were used as test samples to test the trained SVM model. The test result shows that the accurate rate of corn seed breakage inspection was above 95%,the detection time is about 1. 27 seconds. The results provide a reference for the real-time breakage inspection of corn seeds.
Aim at the problems that corn seeds are damaged by various factors during harvest,threshing. And storage.Selecting is time-consuming and exhausting. Put forward a corn seed damage test method based on the machine vision.The first step is using the image acquisition device to grain the single particle corn seed image. Using the subtraction method to determine the type of the image noise and Median filtering method is used to denoise the image. And then marking the image boundaries and using fixed threshold method to split corn seed image segmentation. According to the morphological characteristics of corn seeds extract 6 geometric characters including perimeter,proportion,perimeter-proportion ratio,long axis length,minor axis length,length-width ratio,and another 10 morphological characters including rectangular degree,circularity,compactness 7 Hu invariant moment,a total of 16 characters. The extracted 16 characters components are used as inputs,50 whole corn seed and 50 corn seeds damaged were used as training samples to train SVM,output is "qualified","unqualified"two types and an identification model of corn seeds was obtained.Another complete corn seeds and broken corn seeds each containing 50 grains,were used as test samples to test the trained SVM model. The test result shows that the accurate rate of corn seed breakage inspection was above 95%,the detection time is about 1. 27 seconds. The results provide a reference for the real-time breakage inspection of corn seeds.
引文
[1]宁吉喆.中国统计年鉴-2016[M].北京:中国统计出版社,2016.
[2]应义斌.机器视觉技术在农产品品质自动识别中的应用(I)[J].农业工程学报,2000,16(1):103-108.
[3]凌云.基于机器视觉的谷物外观品质检测技术研究[D].北京:中国农业大学,2004.
[4]何东健.数字图像处理[M].西安:西安电子科技大学出版社,2003.
[5]贾伟宽,赵德安,刘晓洋,等.机器人采摘苹果果实的K-means和GA-RBF-LMS神经网络识别[J].农业工程学报,2015(18):175-183.
[6]孙群,王庆,薛卫青,等.无损检测技术在种子质量检验上的应用研究进展[J].中国农业大学学报,2012,17(3):1-6.
[7]郑小东,王杰.机器视觉在玉米籽粒品质检测中的应用研究[J].中国粮油学报,2013,28(4):124-128.
[8]赵敏.基于机器视觉的玉米品质检测[D].长春:吉林大学2012.
[9]王侨,陈兵旗,朱德利.基于机器视觉的定向播种用玉米种粒精选装置研究[J].农业工程学报,2017,48(2):27-37.
[10]Vapnik V N.An overview of statistical learning theory[J].IEEE Transactions on Neural Networks,1999,10(5):988-999.
[11]Vapnik V N.The nature of statistical learning theory[M].New York:Spring,1995.
[12]梁燕.SVM分离器的扩展机器应用研究[D].长沙:湖南大学,2008.
[13]甘志杰.基于Hu矩和支持向量机的静态手势识别及应用[D].青岛:青岛科技大学,2008.
[14]董立峰.基于Hu矩和支持向量机的静态手势识别及应用[D].武汉:武汉理工大学,2012.
[2]应义斌.机器视觉技术在农产品品质自动识别中的应用(I)[J].农业工程学报,2000,16(1):103-108.
[3]凌云.基于机器视觉的谷物外观品质检测技术研究[D].北京:中国农业大学,2004.
[4]何东健.数字图像处理[M].西安:西安电子科技大学出版社,2003.
[5]贾伟宽,赵德安,刘晓洋,等.机器人采摘苹果果实的K-means和GA-RBF-LMS神经网络识别[J].农业工程学报,2015(18):175-183.
[6]孙群,王庆,薛卫青,等.无损检测技术在种子质量检验上的应用研究进展[J].中国农业大学学报,2012,17(3):1-6.
[7]郑小东,王杰.机器视觉在玉米籽粒品质检测中的应用研究[J].中国粮油学报,2013,28(4):124-128.
[8]赵敏.基于机器视觉的玉米品质检测[D].长春:吉林大学2012.
[9]王侨,陈兵旗,朱德利.基于机器视觉的定向播种用玉米种粒精选装置研究[J].农业工程学报,2017,48(2):27-37.
[10]Vapnik V N.An overview of statistical learning theory[J].IEEE Transactions on Neural Networks,1999,10(5):988-999.
[11]Vapnik V N.The nature of statistical learning theory[M].New York:Spring,1995.
[12]梁燕.SVM分离器的扩展机器应用研究[D].长沙:湖南大学,2008.
[13]甘志杰.基于Hu矩和支持向量机的静态手势识别及应用[D].青岛:青岛科技大学,2008.
[14]董立峰.基于Hu矩和支持向量机的静态手势识别及应用[D].武汉:武汉理工大学,2012.