基于图像形态学的骏枣果梗检测方法  

Detection Method of Junzao Stems Based on Image Morphology Processing

作　　者：肖爱玲[1,2] 李伟[2] 张俊雄[2] 宋鹏[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》2012年第12期140-143,共4页Journal of Agricultural Mechanization Research

基　　金：国家科技支撑支疆项目(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。为此,提出了基于图像形态学的一种快速检测骏枣果梗方法。通过图像预处理获取骏枣二值图像,取反二值图像,对其进行多次膨胀、腐蚀和再膨胀处理,用二值取反图像减去该图像,得到只含有果梗和噪声的图像,再中值滤波处理就可得到只含有骏枣果梗成分的图像。对118帧图像进行检测试验,结果表明:检测速度平均为0.45s/个,果梗识别准确率为92.7%,无果梗误判率为0。该算法鲁棒性更强,基本满足骏枣分级系统精度的要求。The stems detection is an important index in junzao grading system,so a rapid detection method on junzao＇s stems is put forward based on image morphology.got the binary image through the image preprocessing,negated the binary image,then dilated and corroded it many times,at last,dilated the image to the same size as the binary image.In order to remove the influence of the non-stem area,used the negation binary image minus this image get the image only including stem and noises.median filtered the image to get the image including stem area.The analytical results from 120 Junzao images show that detection speed of this method is 0.45 s/a and the segmentation accuracy rate is 92.7% and false Rate of non-stems junzao is 0.The algorithm is more strong robustness which can basically meet the accuracy requirement of Chinese red dates grading system.

关 键 词：骏枣 图像形态学 果梗 检测 

分 类 号：S126[农业科学—农业基础科学] TP391.41[自动化与计算机技术—计算机应用技术]