2022年12月Dec.2022第46卷第6期Vol.46,No.6热带农业工热带农业工程TROPICALAGRICULTURALENCINEERING基于有序轮廓链的葫芦科砧木苗生长点检测①查杨②陈武赖一波吴尧方家吉黄铮喻擎苍③（浙江理工大学浙江杭州310018）摘要为更好地提高葫芦科幼苗嫁接设备的自动化程度，解决葫芦科幼苗插接法嫁接过程中生长点定位的问题，实现葫芦科幼苗插接法的全自动化，本文提出了基于有序轮廓链的葫芦科砧木苗生长点检测方法，通过目标像素的有序轮廓链，获取嫁接苗边缘信息，提取边缘曲率特征，或是提取目标区域单像素骨架，并以此为基础实现嫁接苗生长点定位。结果显示，俯视图的生长点检测其误差最小为1个像素值，最大为5个像素值，而侧视图的生长点定位结果误差最小为1个像素值，最大为4个像素值，因此，本算法可以准确的定位砧木苗生长点坐标并有效避免葫芦科幼苗嫁接过程中因生长点坐标偏差导致的嫁接失败的问题，大幅度提高嫁接成功率。关键词机器视觉；自动嫁接；生长点提取；轮廓检测中图分类号S224DetectionofGrowthPointonCucurbitaceousRootstockSeedlingsBasedonOrderedContourChainZHAYangCHENWuLAIYiboWUYaoFANGJiajiHUANGZhengYUQingcang(ZhejiangSci-techUniversity,Hangzhou,Zhejiang310018)AbstractInordertoimprovethedegreeofautomationofthecucurbitseedlinggraftingequipment,solvetheproblemoflocatingthegrowingpointduringthegraftingprocessofcucurbitaseedlingpluggingmethodandaccomplishthefullautomationofcucurbitaceousseedlingpluggingmethod,thispaperproposesameth‐odfordetectingofgrowthpointoncucurbitaceousrootstockseedlingsbasedonorderedcontourchain,whichobtainstheedgeinformationofthegraftedseedlingandthenextractstheedgecurvaturefeatureorthesingle-pixelskeletonofthetargetareathroughtheorderedcontourchainofthetargetpixel.Basedonthismethod,thegrowthpointofthegraftedseedlingsislocated.Theresultsshowthattheerrorofthegrowthpointdetectioninthetopviewhasaminimumerrorof1pixelandamaximumerrorof5pixels,whiletheerrorofthegrowingpointlocationinthesideviewistheminimumvalueof1pixelandthemaxi‐mumvalueof4pixels.Therefore,thealgorithmcanaccuratelylocatethegrowthpointcoordinatesofroot‐stockseedlingsandeffectivelyavoidtheproblemofgraftingfailurecausedbythedeviationoftheg...
