农作物病害危害程度自动测定与分级的研究

Research of Automatic Menstruation and Classification for Degree Disserved of Crop Disease

准确、快速地处理与识别植物病害图像,并自动可靠地对其进行危害程度测定是进行农作物病害防治的关键。为此,研究了利用分形理论对农作物病变叶片自然纹理图像进行了处理,利用BP网络来实现自然纹理图像的分类问题。同时,分析了各种不同的特征在分类器中的性能,提出了一种纹理特征性能的比较方法。结果表明:适当选取分形维数的不同计算方法互相组合,可以提高分类器的性能。

During the process of growth and developing, crop is usually influenced by all kinds of factors, which cause crop disease. Disposing, recognizing plant disease image truly and fleetly, and menstruating on it by classification for degree disserved automatically and reliably is the key of crop disease prevention. So, the problems, that utilize the theory of analyzing figure to dispose crop pathological changes laminate nature texture image, and utilize BP network to class nature texture image, are studied. By analyzing all kinds of different characters which have capability in classified apparatus, the ways that compare texture character capability is presented. The result shows： choosing different count methods of analyzing figure dimension to fit together each other, may improve the capability of classified apparatus.