基于图像识别的苹果等级分级研究
|
摘要
针对传统的苹果人工分级方法存在检测不全面、分级效率较低等问题,以红富士苹果为对象,研究了利用机器视觉实现对苹果等级进行分级的方法。搭建了图像采集系统;运用中值滤波方法去除图像噪声,并创造性地提出分离彩色图像HSL模型中的S通道分量作为后续图像处理的源图像,结合Otsu算法实现了自动阈值分割进行轮廓提取。一方面选择色调值H通道分量的直方图数据作为苹果颜色分级的特征参数,通过支持向量机对苹果进行等级判定,判定准确率为89%。另一方面选择亮度L通道分量的能量、熵和逆差矩作为特征参数,利用神经网络对苹果进行有无缺陷判定,判定准确率为95.5%。
Aiming at the problems of incomplete detection and low classification efficiency of traditional algorithm grading methods of apple,the method of grading apples by machine vision was studied with Red Fuji apple as the object. An image acquisition system is built. The median filtering method is used to remove image noise,and the Schannel component in HSL model of color image is creatively separated as the source image of subsequent image processing. The contour extraction is realized by automatic threshold segmentation combined with Otsu algorithm. On the one hand,the histogram data of hue value H channel component is selected as the characteristic parameter of apple color grading,and the classification accuracy of apple is 89% by using support vector machine. On the other hand,taking the energy,entropy and deficit moment of the luminance L channel component as the feature parameters,the neural network is used to judge whether the apple is defective or not,and the accuracy of the determination is 95.5%.
Aiming at the problems of incomplete detection and low classification efficiency of traditional algorithm grading methods of apple,the method of grading apples by machine vision was studied with Red Fuji apple as the object. An image acquisition system is built. The median filtering method is used to remove image noise,and the Schannel component in HSL model of color image is creatively separated as the source image of subsequent image processing. The contour extraction is realized by automatic threshold segmentation combined with Otsu algorithm. On the one hand,the histogram data of hue value H channel component is selected as the characteristic parameter of apple color grading,and the classification accuracy of apple is 89% by using support vector machine. On the other hand,taking the energy,entropy and deficit moment of the luminance L channel component as the feature parameters,the neural network is used to judge whether the apple is defective or not,and the accuracy of the determination is 95.5%.
引文
[1] Mizushima A,Lu R.An image segmentation method for apple sorting and grading using support vector machine and Otsu’s method[J].Computers and Electronics in Agriculture,2013,94:29-37.
[2] Sofu M M,Er O,Kayacan M C,et al.Design of an automatic apple sorting system using machine vision[J]. Computers and Electronics in Agriculture,2016,127:395-405.
[3] Yossy E H,Pranata J,Wijaya T,et al. Mango fruit sortation system using neural network and computer vision[J].Procedia Computer Science,2017,116:596-603.
[4]何微.基于外部特征参数的番茄分级方法研究[D].武汉:华中农业大学,2013.
[5]付鹏.基于机器视觉的苹果检测与识别关键技术研究[D].咸阳:西北农林科技大学,2012.
[6]朱垚,骆利华.近红外光谱技术的水果成分检测数学模型分析[J].激光杂志,2018,39(8):109-112.
[7]乔陆,陈静.基于SOPC水果分级检测系统研究[J].食品与机械,2016,32(8):95-97,201.
[8]辛华健.计算机视觉在芒果品质检测中的应用研究[J].农机化研究,2019,41(9):190-193.
[9]黄星奕,林建荣,赵杰文.苹果果梗和缺陷的识别技术研究[J].江苏大学学报:自然科学版,2004,25(3):193-195.
[2] Sofu M M,Er O,Kayacan M C,et al.Design of an automatic apple sorting system using machine vision[J]. Computers and Electronics in Agriculture,2016,127:395-405.
[3] Yossy E H,Pranata J,Wijaya T,et al. Mango fruit sortation system using neural network and computer vision[J].Procedia Computer Science,2017,116:596-603.
[4]何微.基于外部特征参数的番茄分级方法研究[D].武汉:华中农业大学,2013.
[5]付鹏.基于机器视觉的苹果检测与识别关键技术研究[D].咸阳:西北农林科技大学,2012.
[6]朱垚,骆利华.近红外光谱技术的水果成分检测数学模型分析[J].激光杂志,2018,39(8):109-112.
[7]乔陆,陈静.基于SOPC水果分级检测系统研究[J].食品与机械,2016,32(8):95-97,201.
[8]辛华健.计算机视觉在芒果品质检测中的应用研究[J].农机化研究,2019,41(9):190-193.
[9]黄星奕,林建荣,赵杰文.苹果果梗和缺陷的识别技术研究[J].江苏大学学报:自然科学版,2004,25(3):193-195.