Based On The Segmentation And Recognition Of GGCM Plant Lesions

This paper puts forward a kind of research method of plant disease region segmentation and recognition of gray gradient co-occurrence matrix based on scholars research on plant disease recognition in the domestic and foreign. The production process of Cotton Verticillium Wilt in common as the main research object of the extent of disease classification and recognition of the disease, not only to better prevent and control the disease, but also to save manpower and financial resources,while improving crop yields and reducing the environment pollution of pesticides.Cotton leaf collection with verticillium wilt, using grayscale, binary, wavelet denoising, morphological processing techniques for image preprocessing, image segmentation using Otsu method on the basis of preprocessing. Image segmentation process by following three steps:(1)color space transform, cotton disease leaf images were converted from RGB color space to HSV and YCbCr color space.(2) use of Otsu algorithm respectively to B component(blue component), S component(chroma saturation), Cr component(red chrominance) separately thresholding, the B component of the image be inverted after the division operation, B negated image after image segmentation and S components of a binary image after fusion was then obtained morphological image processing cotton leaves, so cotton leaves and background areas separated.(3) The resulting image and the Cr component of cotton leaves image segmentation phase after the operation, then the morphological opening operation, thereby obtaining an image of a lesion area, to complete the entire process of dividing the image of disease. Secondly, the use of GGCM and color moments were to extract the texture and color characteristics of disease areas, but also to calculate the area of cotton and cotton leaf diseases area of the region, and then find the relative area of the extracted three types of feature fusion together, thereby obtaining identification features cotton disease region.The paper also leaves on the extraction of cotton disease region all the feature vectors arepreferred,select some of the more notable characteristic parameters, this approach can effectively reduce the number of dimensions to be extracted features, to extract all feature vectors from 25 dimensional reduction to 8 dimensions,not only reduces the time required for training but also improves the efficiency of BP neural network. The research on the prevention of Verticillium governance provides an effective and advanced technical support, but also conducive to the prevention and treatment of cotton pests.