采摘机器人振荡果实动态识别  

Dynamic Recognition of Oscillating Fruit for Harvesting Robot

作　　者：吕继东[1] 赵德安[1] 姬伟[1] 陈玉[1] 沈惠良[1] 张颖[1] 

机构地区：[1]江苏大学电气信息工程学院,镇江212013

出　　处：《农业机械学报》2012年第5期173-178,196,共7页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家高技术研究发展计划(863计划)资助项目(2006AA10Z254);高等学校博士学科点专项科研基金资助项目(20093227120013);江苏高校优势学科建设工程资助项目;江苏省普通高校研究生科研创新计划资助项目(CXZZ11_0573)

摘　　要：提出一种采摘机器人在果实振荡状况下的动态识别方法,解决由于果实振荡影响采摘机器人识别定位时间,进而影响采摘速度和效率的问题。首先对所采集的振荡果实图像进行图像分割,将其分为果实和背景两部分;其次引入帧间差分法、水平最小外接矩形法等对分割图像进行振荡果实动态区域的区域标识,然后对其振荡果实进行识别,当图像中有多个振荡果实时,以距离图像中心最近原则确定采摘振荡目标果实。试验结果表明对实际采摘环境下遇到的多数情况,所提算法都能很好地识别出振荡果实,识别时间少于0.5 s。A kind of dynamic recognition method under the fruit oscillation condition for the harvesting robot was researched in order to resolve the problem of fruit oscillation influence to its recognition and orientation,and further influence to the harvest efficiency of harvesting robot.Firstly,the acquired oscillation fruit was segmented into two parts： fruit image and background image;secondly,the inter-frame difference method and the minimum enclosing rectangle method in horizontal direction were introduced to label dynamic region of the oscillation fruit for the segmented image;thirdly,the oscillation fruit was identified and located.The harvesting oscillation fruit was selected based on the principle of the nearest to the image center when the image had a lot of oscillation fruits.The test result showed that the oscillating fruit was nicely recognized and the recognition time was less than 0.5 s for the most cases under the actual harvesting environment.

关 键 词：采摘机器人 振荡 图像分割 动态识别 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]