基于图像分割的玉米粒数高通量估算方法
作者:张颐康 李翠丽 李源
来源:《湖北农业科学》2016年第12期
摘要:设计了一种高通量玉米粒计数的方法,首先针对玉米穗分割对边界进行跟踪,根据方向变化幅度寻找“凹点”,然后根据边界的方向对“凹点”两两连接进行图像分割;其次提出了改进的支持向量机分类算法的玉米粒识别,除了颜色信息外还将图像中的梯度信息加入分类的自变量中,避免了玉米粒的连通现象;最后通过面积对玉米粒粒数进行估算。结果表明,该方法精确度较高,准确率达到了96.4%;实现了高通量处理,解决了随机摆放的玉米穗粘连分割与高通量条件下图像模糊玉米子粒难以识别的问题。
关键词:图像分割;高通量;最大类间方差法;边界跟踪算法
中图分类号:S2;TP339 文献标识码:A 文章编号:0439-8114(2016)12-3185-04
DOI:10.ki.issn0439-8114.2016.12.047
Abstract:A kind of high flux estimation method was propod. Firstly, track the boun
dary and arch for the concave by the range ability of direction. Then join every two concave bad on the direction in order to gment the corncob. Secondly, u the support vector machine to identify corn kernels bad on gradient and color. At last, estimate the number of corn kernels according to area. Experiments show that the estimation accuracy is higher, with 96.4% accuracy. Compared with existing methods, this method simplifies the kernels counting device, realizes the high flux processing, solves the problem of gmenting corncob which is placed randomly and identifying corn kernels which is vague becau of the high flux conditions.