玉米小斑病病害程度图像检测

Image Detecting for Hazard Levels of Corn Spots

为了实时获取作物病害程度信息,研究了一种基于RGB彩色模型的玉米小斑病图像的分割方法,并利用分割结果求得的玉米冠层危害程度来计算整株玉米的发病程度。由于图像中背景复杂,将叶片与病斑同时分离出来的可能性小,故该方法分为两步:首先从获取的RGB图像中提取R、G、B分量,利用2G-R-B图像采用迭代法自动选取阈值将玉米叶片从背景中分割出来;然后根据R-G图像将病斑从叶片上分离出来。30幅图像中玉米叶片、病斑基本上能提取出来,但没有黄化的少量侵染点无法有效分割。

In order to instantly acquire the information of hazard levels of plant diseases and pests, a method of segmentation of corn spot image based on RGB color model was proposed. The hazard levels of the whole plant could be calculated through hazard levels of the corn canopy acquired by the results of image segmentation. During the segmentation of crop leaf and disease spots, the method was separated into two steps： （1） R, G, B three components were gotten from RGB color images which were caught by numeral camera, and the corn leaves were segmented from the background of the 2G - R - B images by choosing threshold automatically via iterative method. （2） The disease spots were segmented from leaves based on R - G images. The results showed that the corn leaf blades and the sickness spots of the 30 images could be picked up except a few non-etiolation infection courts.