基于圆形模型的骏枣果梗检测方法  

Detection Method of Junzao Stems Based on Circle Model

作　　者：肖爱玲[1,2] 李伟[2] 李传峰[1] 

机构地区：[1]塔里木大学机械电气化工程学院,新疆阿拉尔843300 [2]中国农业大学工学院,北京100083

出　　处：《农机化研究》2013年第1期152-155,共4页Journal of Agricultural Mechanization Research

基　　金：国家"十一五"科技支撑计划项目(2010BAD05B02);塔里木大学校长基金项目(TDZKSS08006)

摘　　要：果梗有无的判别是骏枣分级系统中的一项重要指标,针对传统的膨胀与腐蚀操作不能达到果梗识别的要求,为此提出了基于圆形模型的骏枣果梗检测方法。通过图像预处理获取二值图像;构建一圆形模型,以目标图像的形心为圆心,以等效椭圆长半轴乘1个调整系数为半径,该圆形模型可以覆盖骏枣果肉部分图像,剩下图像部分主要为果梗;由于图像可能还含有噪声,为此进行中值滤波处理,最后即可提取果梗图像部分。对120帧图像进行检测试验,结果表明:每个图像的平均处理时间小于500ms,本次试验果梗识别准确率93%,基本满足红枣分级系统精度的要求。Whether Junzao had stems or not is an important index in its grading system.In order to detect it,a method of catching junzao stems was put forward based on image morphology for this reason.Getted binary image through image preprocessing,and then negated the binary image;build a circular model with centroid to central,equivalent elliptic long half shaft by adjustment coefficient for radius;in order to catch stem,the circular model must cover pulp part of images;the analytical results from 120 junzao images show that the detecting speed of this method requires 500ms to determine an image and the detecting accuracy is 93%.It can basically meet the requirement of the accuracy of detection of chinese red dates grading system.

关 键 词：骏枣 图像形态学 圆形模型 果梗 检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]