Study On Crop Leaf Disease Extraction System Based On Computer Image Processing

China is a large agricultural country since ancient times, agriculture is the mainstay ofthe economy, the development of agriculture related to the vital interests of people. However,natural disasters,especially crop diseases frequently occur, seriously hampering thedevelopment of agricultural economy. Therefore, how to effectively control crop diseases,improve the agricultural environment, improve the quality of agricultural products andincrease agricultural production, is an important issue in today’s agricultural research. In thispaper, aiming at the agricultural automation, we research and development the crop leafdisease extraction system based on computer image processing.In this paper, according to the research results at home and abroad, we select cucumber,cotton, corn and other crops as sample and analysis the image of their leaf diseases, withparticular emphasis on the key technologies such as image filtering and segmentation, andthen establish the crop leaf disease extraction system. The system is based on computer imageprocessing, combined with the actual needs of the market and agricultural producers, aimingat extract the crop diseases image accurately and quickly, it provides a scientific basis for theprevention and treatment of crop diseases.(1) According to the Intensity and color guidelines,we propose two filtering algorithm(IVMF and CDVMF), to ensure that we identify the noise pixels first, and then filtering thepixels. The two filtering algorithm can solve the problem of large amount of computation,edge blur and others for traditional vector median filtering algorithm. By contrast experiment,the effect of the filtering algorithm CDVMF is the best and the filtering algorithm IVMF isthe fastest one. On this basis, in order to further improve the filtering efficiency, we usepseudo-vector median value instead of vector median, creating two vector pseudo-medianfiltering algorithm(BVPMF and CDVPMF) based on luminance and chrominance guidelines.Experimental results show that the pseudo-vector median filtering algorithm efficiency nearly10%.(2) We propose a new segmentation algorithm-color information fusion segmentationalgorithm.It combines the brightness information and color information to solve the problem of hole, piebald and fuzzy boundary by traditional image segmentation methods.(3) In order to ensure adjacent pixels with similar colors ignite at the same time, wepropose an improved parallel PCNN network model, it consists of the inner PCNN and outerPCNN nested structure, select the minimum color contrast of the target region and thebackground region as the iteration termination condition. For general image segmentation, inorder to obtain better segmentation results, the threshold attenuation step γ and connectioncoefficient β is selected in the interval (0.01,0.06),(0.01,0.08).