Research On Recognition Method Of Strawberry Disease Based On Image Processing

Strawberry leaves are important part of their growth status. In order to monitor the growth status of strawberry, the health state of their leaves should be monitored continuously to determine whether there is a disease with strawberry leaves and guide the growers whether to spray and the amount of spraying. The main difficulties in the identification of the natural light of the disease of strawberry images is that the shadow caused by non-consistent light and overlapped green leaf, so that the target and background are uneasy to be segmented and the disease characteristics are difficult to be selected. To segment and recognize of the strawberry snake disease image, improved watershed segmentation algorithm based on mark and image recognition method of support vector machine(SVM) based on texture statistical feature were proposed in this paper. In this method, firstly, the color threshold was used to extract the strawberry leaf area image; Then gray images in HSV color space and edge gradient image was processed respectively to get foreground and background markers, and markers were calibrated by the forced minimum technique, standard watershed transform was applied to extract strawberry single leaf image; Finally, the normalized gray histogram of the strawberry single leaf was obtained to process extraction and fusion of 8 texture statistic feature, and SVM was applied to identify the strawberry disease.In this paper, the image of strawberry disease could be segmented effectively with the improved watershed algorithm and disease of strawberry single leaf image could be identified through the SVM pattern recognition method. Strawberry disease identification was processed based on image processing method, following aspects of the study were mainly carried out:Firstly, in order to complete the removal of complex background and the extraction of leaf area of strawberry disease image, the operation of green component prominent, image enhancement, threshold segmentation, corrosion and expansion was mainly applied to pre-segment the original image of strawberry;Secondly, the green leaf area image was transformed into gray image in HSV color space and strawberry single leaf image was extracted through improved watershed algorithm, results of segmentation would be used for subsequent disease identification;Thirdly, the normalized gray histogram of the strawberry single leaf was obtained to extract 8 texture statistic features, including: average gray level, standard deviation, three order center moment, smoothness, uniformity, average information, maximum probability of gray level, gray scale;Then, characteristic vectors of strawberry single leaf samples were trained and recognized respectively with support vector machine(SVM), K-nearest neighbor(K Nearest Neighbor, KNN) and Naive Bayes Classifier(NB). By comparing the recognition results of the three methods, the recognition accuracy of SVM is higher than the other two methods and is more suitable for the identification of strawberry diseases;Finally, Matlab was applied to simulate the realization of the system process, and a strawberry disease identification software system was established by MFC and OpenCV in the VS2010.