基于圆心定位的籽粒计数方法研究  

A Counting Method of Grain Based on Localization of Center

作　　者：戴文华 吴翔[2] 裘正军[2] 何勇[2] 

机构地区：[1]嘉兴市农机技术推广站,浙江嘉兴314000 [2]浙江大学生物系统工程与食品科学学院,杭州310058

出　　处：《农机化研究》2015年第7期198-202,共5页Journal of Agricultural Mechanization Research

基　　金："十二五"国家科技支撑计划(863计划)项目(2012AA101903)

摘　　要：作物籽粒数常用于评价农作物的产量和品质。为此,利用机器视觉技术,提出了一种基于圆心定位的籽粒计数方法。针对不粘连、简单粘连和复杂粘连3种情况,各拍摄15张大豆照片,利用Mat Lab进行图像预处理、粒径统计、Freeman链码提取、圆心定位、圆心筛选和聚类等处理后,计算大豆的籽粒数。实验表明:不粘连和简单粘连准确率均为100%,复杂粘连的平均准确率为97.8%;此方法对于圆形籽粒的计数具有很好的效果。The grain number of crops has been used commonly for assesment of yield and quality.Based on the machine vision technology, this study carried out a simple method of localization of centers for counting of grain particles.For three condition of no-adhesion, simple-adhesion and complex-adhesion, we took 15 pictures of soybeans respectively. After consecutive image operation of preprocessing, statistics of granulometry, extraction of Freeman code, calculation of centers, sieving and clustering using Matlab IPT toolbox, we got the total number of soybeans.The obtained reasult was that, both no-adhesion and simple-adhesion got 100%accuracy rate, and the complex-adhesion achieved accuracy rate 97.8%in average.The experiment proved the feasibility of this method in the counting of round like grain particles.

关 键 词：籽粒计数 机器视觉 FREEMAN链码 大豆 

分 类 号：S15[农业科学—土壤学] TP391.41[农业科学—农业基础科学]