基于G-R分量与K-means的马铃薯病虫害图像分割

IMAGE SEGMENTATION OF POTATO PESTS AND DISEASES BASED ON G-R COMPONENT and K-MEANS

针对复杂背景下马铃薯病虫害图像色彩复杂多样的问题,提出了一种基于G-R分量与K-means的图像分割方法。该方法首先提取G-R分量灰度化图像,然后对图像进行中值滤波和形态学变换操作以去除部分背景,最后将处理过后的图像转换到L*a*b*色彩空间并提取ab分量进行K-means聚类分割。本文利用该方法将暗头豆芫菁及芫菁、马铃薯重花叶病毒病及马铃薯黄痿病病斑从原彩色图像中准确提取了出来。该方法能够较为准确、完整地将目标病虫害从彩色图像中提取出来,在马铃薯病虫害治理方面有较好的应用价值。

An image segmentation method based on the G-R component and K-means was proposed to solve the problem of complex and diversified colors of potato pests and diseases images under complex backgrounds.First, the G-R component was extracted.Then the morphological transformation of the median filtered image was performed to remove part of the background.Finally, the images were converted to the L*a*b* color space, and the ab component was extracted, the K-means algorithm was used to segment the image.In this paper, the Epicauta obscurocephala Reitter, blister beetle, and disease spot of the heavy mosaic virus disease and Verticillium wilt of potato were accurately extracted from the original color image by using this method.It could be concluded that the method could accurately and completely extract the target pests and diseases from the color images, and had good application value in potato pest management.