Image Recognition Of Peanut Brown Spot Based On Support Vector Machine

With the rapid development of information technology,the use of computer technology to solve the problems related to agricultural production has also become a hot spot of current research.In recent years,a large number of researchers at home and abroad have used the computer vision technology to study the identification and prevention of crop diseases and pests.These studies are of great significance to guide farmers to control crop diseases and improve crop yield.On the basis of previous studies,this study investigated the recognition of peanut brown spot disease by using image processing technology and support vector machine.In this paper,the main work is as follows:(1)An improved image segmentation algorithm based on non-green features was proposed.The improved algorithm firstly divides the image of peanut lesions by the image segmentation algorithm based on the non-green feature,and gets the initial spot area.Then,the difference between the G component and the mean value of the R component in the lesion region obtained after the initial segmentation was calculated.Finally,through the comparison of the difference between the size and a threshold to decide whether the two segmentation of lesion.The test results show that the average segmentation accuracy of the improved segmentation algorithm to the image of peanut brown spot was 97.22%.After the improvement method is processed,spot image segmentation rate is 5.28%higher than the original method.The segmentation result is 12.88%higher than that of the traditional C-V model,and the average running time is only 24.64%of the latter.(2)Put forward a segmentation method of spot image based on improved C-V model.First of all,using the method of image segmentation based on the non-green feature to divide the image and get the initial spot area.Then,marking the lesion area and calculate the area,remove the connected area which is less than the threshold.Then,using the 3×3 sliding window to traversal lesion image.to calculated the difference between the R value and the G value of the pixels in the window.Find the region with the largest difference between the R value and the G value in the image.Record the center pixel coordinates of the sliding window at this time.Let this coordinate as the initialization position of the C-V model evolution contour for lesion image segmentation.The test results show that.The average accuracy of this segmentation algorithm for the image segmentation of peanut brown spot is 94.42%.Through this approach,the lesion segmentation accuracy is 10.08%higher than the original method.The average running time is much less than that of the traditional C-V model.(3)32 peanut brown spot lesion feature parameters were extracted,including shape,color and texture feature space.These features are used as input items of SVM to identify the image of peanut brown spot disease.Then,each characteristic parameter is optimized by stepwise discriminant method.Finally,13 valid characteristic parameters were selected.The optimal recognition rate was obtained by using these 13 characteristic parameters to identify the samples of peanut brown spot.The optimal recognition rate of the 170 training samples was 96.47%,and the optimal recognition rate of the 60 test samples was 95%.