一种玉米雄穗图像分割方法  

A Corn Tassel Image Segmentation Method

作　　者：茅正冲 刘永娟 MAO Zheng-chong;LIU Yong-juan(Key Laboratory of Advanced Process Control for Light Industry,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学轻工过程先进控制教育部重点实验室

出　　处：《控制工程》2018年第10期1892-1897,共6页Control Engineering of China

基　　金：江苏省自然科学基金(NO.BK20131107);国家自然科学基金资助项目(No.60973095)

摘　　要：针对玉米田间环境复杂、光照不均、干扰不确定等问题,提出一种基于改进K-means聚类和二维Otsu的玉米雄穗图像分割方法。首先对玉米雄穗图像的R、G、B通道进行处理,减少玉米叶的干扰,突出玉米雄穗;其次利用改进的K-means对玉米雄穗的Cr图进行聚类分割,利用二维Otsu对玉米雄穗的超红图像进行分割;最后,为了去除土地的干扰,从Cr分割结果图中去除超红分割结果,取剩余的Cr分割结果作为玉米雄穗的分割结果。为了验证该算法的有效性,选用了不同生长环境的玉米雄穗图像,分别利用提出的方法、二维Otsu进行了比较实验。实验结果表明,该方法有很好的抗干扰性,对生长环境有很强的鲁棒性。A com tassel segmentation method based on modified K-means and two-dimensional Otsu is proposed for the problems of complex corn filed environment, uneven illumination, interference uncertainty and so on. Firstly, R, G- B channel is processed respectively to reduce the interference of corn leaves and highlight corn tassel. Secondly, modified K-means is used to cluster and segment Cr images of corn tassel, two-dimensional Otsu is used to segment the super red image of corn tassel. At last, the super red image segmentation result is removed from Cr image segmentation result, the rest of Cr image segmentation result is taken as corn tassel segmentation image in order to remove the interference of soil. In order to verify the effectiveness of the proposed algorxtnm, me corn ta-, ,,,-,,- -, make comparative experiment, the proposed method, two-dimensional Otsu is respectively used in the comparative experiment. Experimental results show that this method has good immunity and it is robust to the growth environment.

关 键 词：玉米雄穗 图像分割 改进K-MEANS 二维OTSU 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]