中文摘要：

为了根据田间图像自动判断玉米抽雄期,提出了1种玉米雄穗分割方法.首先将红绿蓝（RGB）图像转换到YCbCr空间,对C6、CV分量图进行增强处理;再利用训练好的Fisher分类器对每个像素的CKO值进行分类,初步分割出玉米雄穗;然后利用颜色指数超蓝因子（ExB）对RGB图像进行灰度化处理,利用改进的Kmeans聚类对灰度图像进行聚类;最后结合Fisher分类结果和聚类结果确定玉米雄穗像素.实验结果表明采用该文方法识别玉米雄穗,正常环境下的错分率和查全率分别为0.177%和0.831%,干旱环境下的错分率和查全率分别为0.141%和0.811%,该文方法对玉米生长环境具有很好的鲁棒性.

英文摘要：

mA corn tassel segmentation algorithm is proposed for com field large area images to determine com tasseling stage automatically. Firstly, a red green blue（ RGB） image is converted to the YCbCr space, Cb and Cr component images are enhanced respectively;then a Fisher classifier is trained to classify the Cb and Cr value of each pixel and com tassels are segmented preliminary ; next,a new color index excess blue index（ExB）is used to gray the RGB image,and the gray image is clustered by an improved Kmeans; lastly, the Fisher classification results and clustering results are combined to determine final com tassel pixels. Experimental results show that this algorithm can identify corn tassels effectively, fault rate and recall rate of a normal environment are 0. 177% a n d 0. 831% respectively,fault rate and recall rate of a drought environment are 0. 141% a n d 0. 811% respectively, this algorithm is robust for maize growth environments.