机器视觉在采摘机器人识别与定位中的应用  

Machine Vision Stay Picking Robot Identification and Positioning Application of Virtual Reality

作　　者：焦迎雪 董海涛[2] 武文革[3] JIAO Ying-xue;DONG Hai-tao;WU Wen-ge(Shanxi Railway Vocational and Technical College,Shanxi Taiyuan 030013,China;Shanxi Electromechanical Vocational and Technical College,Shanxi Changzhi 046011,China;North University of China,Shanxi Taiyuan 030051,China)

机构地区：[1]山西铁道职业技术学院,山西太原030013 [2]山西机电职业技术学院,山西长治046011 [3]中北大学,山西太原030051

出　　处：《机械设计与制造》2024年第2期280-285,共6页Machinery Design & Manufacture

基　　金：山西省教育科学“十三五”规划“1331工程”研究专项课题(ZX—18125);国家自然科学基金(51875533)。

摘　　要：针对采摘机器人的运行环境复杂,采摘效率无法满足实际生产需求。这里在采摘机器人体系结构的基础上,提出了一种基于机器视觉的夜间识别与定位方法。使用基于粒子群优化的独立成分分析方法来降低夜苹果图像中的噪声,然后使用PCNN分割方法对图像进行分割并通过边缘检测等提取目标轮廓,最后通过改进的三点定圆法对目标果实进行定位。通过仿真验证了该方法的可行性。结果表明,该方法在夜间遮挡小于50%时识别率为94.3%,遮挡大于50%时识别率为89.05%,可以有效提高识别和定位的准确性。为机器人识别和定位技术的发展提供了一定的参考。Due to the complex operation environment of picking robot,picking efficiency cannot meet the actual production requirements.Based on the architecture of picking robot,night recognition and location method based on machine vision is proposed.The independent component analysis method based on particle swarm optimization is used to reduce the noise in the night apple image,then the PCNN segmentation method is used to segment the image and the target contour is extracted by edge detection,etc.Finally,the target fruit is improved by the improved three-point circle method Positioning.The feasibility of this method was verified by simulation.The results show that the recognition rate is 94.3%when the occlusion is less than 50%and89.05%when the occlusion is greater than 50%,which can effectively improve the accuracy of recognition and location.It provides a reference for the development of robot recognition and positioning technology.

关 键 词：机械视觉 采摘机械人 识别与定位 独立成分分析 三点定圆法 

分 类 号：TH16[机械工程—机械制造及自动化] S24[农业科学—农业电气化与自动化]