Research Of Vegetable Leaf-Eating Pests Based On Image Analysis

In this paper, the traditional methods of pest identification existed in low efficiency, low accuracy, and time-consuming, so I attempt to find a new calculation vegetable pests automatic identification methods .The study use leaves of "Jing Feng Number One" cabbage as the research object, computer image-based image processing, using of automatic identification technology to research and search the holes of leafs'information made by pests. The purpose is that through the holes'information accurate and efficient determination of the appropriate type of pest and pests reverse to achieve recognition.The main contents of this paper are:(1) Using the average of background intensity as the threshold for binary image segmentation:Through observing the existing image binarization segmentation methods and studying the background of leaves pictures, I found there are significant differences between the images that is proposed using the mean of background gray image binarization threshold value . In the gray image background I will select n points and calculate the n points's average gray values of P, while the unit of P/255(Threshold between 0 and 1).Finally using MATLAB'library function im2bw Segmentation of images. Through the experiment, finding that it has a good segmentation effect, and leaves can be completely separated from the background , at same times, simplify the degree of image processing, reduce the loss of volume image data on image processing.(2)Using of ultra-blue method for binary image segmentation: An RGB color space image,the image pixel color values in all three primary colors, R,G,B value of the size of the three sub-combination.The purpose of Image graying is to obtain the pixels of image R, G, B components into the same value. Based on this point, in order to make the image after gray level of the simplest to use a global threshold value with the background and leaves separated, tested, ultimately establish a super-blue of the segmentation. We found that the leaves of cabbages images are all green leaves with white background, so here using a color of RGB color space as a threshold value, get the following transformation matrix. Using the blue color of The RGB color space information translate the image to the grounds of the gray image,at same time using the function Threshold=graythresh(I) extract the threshold of image,finally through function im2bw complete the image binarization.(3) The paper proposed the reverse identification method through the pests ate leaves to determine the pests. After image pre-processing of the pest ate leaves, we can automatically extract 7 spherical shape feature values of the pest ate leaves: the image roundness degree, complexity degree, spherical degree, etc., and then build the BP neural network model for recognition. The results show that the method can get a very good result at recognizing the pest species by the geometric feature of pest ate leaves, and give a scientific evaluation for the leaf's harm degree.In summary, using of computing to carry out the reverse leaves pest identification method is feasible, and the recognition efficiency of this method, the accuracy of the method are higher than in the past.this study provides a scientific method for the realization of vegetable pest identification and computer science degree evaluation, and enable path, giving the timely for pest controlling, at same time insuring the vegetable production.