摘要
针对现有马铃薯商品薯薯型检测方法精度低、效率差,难以满足机械化与自动化分选要求的问题,基于机器视觉技术开展马铃薯薯型检测方法研究。以3种薯型(球形或类球形、椭球形和畸形)马铃薯为研究对象,首先搭建马铃薯图像采集系统采集3种薯型的图像,然后对马铃薯图像进行预处理,获取马铃薯轮廓图像;再提取马铃薯图像的10个不变矩作为特征参数,采用量子行为粒子群(QPSO)算法优化支持向量机(SVM)分类器,最终实现薯型的自动检测。结果表明,QPSO-SVM模型中球形或类球形马铃薯、椭球形马铃薯和畸形马铃薯的薯型检测准确率分别为97.0%、94.0%和91.0%,平均检测准确率达95.2%,该模型可用于马铃薯薯型的快速检测。
Aiming at the problems of low accuracy and poor efficiency of existing potato shape detection methods for commercial potatoes,which made it difficult to meet the requirements of mechanized and automated sorting,a method for potato shape detection was investigated based on machine vision technology.Three potato shapes(spherical or subspherical,ellipsoidal,and irregular)were taken as research objects.First,a potato image acquisition system was built to collect images of the three shapes.Then,the images were preprocessed to obtain potato contour images.Subsequently,ten invariant moments of the potato images were extracted as feature parameters,and thesupport vector machine classifier was optimized using the quantum-behaved particle swarm optimization algorithm,ultimately achieving automatic detection of potato shape.The results showed that the detection accuracies for spherical or subspherical,ellipsoidal,and irregular potato shapes in the QPSO-SVM model were 97.0%,94.0%,and 91.0%,respectively,with an average detection accuracy of 95.2%.This model could be used for the rapid detection of potato shape.
作者
万鹏
熊成新
蔡杰
刘艳芳
吴晓龙
WAN Peng;XIONG Cheng-xin;CAI Jie;LIU Yan-fang;WU Xiao-long(College of Engineering,Huazhong Agricultural University,Wuhan 430070,China;Ministry of Agriculture and Rural Affairs Key Laboratory of Agricultural Equipment in Mid-Lower Yangtze River,Wuhan 430070,China;Hubei Academy of Scientific and Technical Information,Wuhan 430071,China)
出处
《湖北农业科学》
2025年第12期212-217,共6页
Hubei Agricultural Sciences
基金
湖北省重点研发计划项目(2023BBB062)。
关键词
量子行为粒子群
支持向量机
马铃薯
商品薯
薯型
quantum-behaved particle swarm optimization
support vector machine
potato
commercial potato
potato shape
