基于BP网络的植物病害彩色图像的分割技术  

Segmentation technology of color image about plant disease based on BP neural network

作　　者：齐龙[1] 马旭[1] 张小超 

机构地区：[1]吉林大学生物与农业工程学院,长春130022 [2]中国农业机械化研究院,北京100083

出　　处：《吉林大学学报（工学版）》2006年第S2期126-129,共4页Journal of Jilin University:Engineering and Technology Edition

基　　金："863"国家高技术研究发展计划项目(2003AA209012)

摘　　要：通过颜色空间的转换,把RGB颜色空间的植物病害图像转换到HSI和YCbCr颜色空间,由于HSI和YCbCr空间的H,Cb,Cr分量与图像的亮度无关,可以有效地抑制噪声,因此选取H,Cb,Cr分量为特征值作为BP人工神经网络的输入值。网络经过训练后,对265幅病害图像进行分割,其中251幅被正确分割,分割的准确率达到94.7%,满足了诊断植物病害的前期处理要求。Because the H,Cb,Cr components of the HSI and YCbCr color spaces do not depend upon the light intensity of the image,the RGB color space of the plant disease images was transformed to the HSI and YCbCr color spaces to restrain the noise in the image.A 3-layer BP artificial neural network was set up to segment the color images about the plant diseases.The characteristic values of the H,Cb,Cr components were selected as the input of the network.After the successful training of the network,265 plant disease images were segmented and 251 images were done correctly,the correctness rate was 94.7%,satisfying the request from the preparation for diagnosing the plant diseases.

关 键 词：农业工程 彩色图像分割 BP神经网络 植物病害 混合颜色空间 

分 类 号：S432[农业科学—植物病理学]