Tomato Disease Recognition Based On Image Processing

Tomato is one of the most in demand for vegetables in the world.Its unique taste and function are popular at home and abroad.The demand of fresh tomatoes and tomato products makes the scale of tomato planting expand gradually.The increasing demand for tomatoes with the characteristics of high yield,less disease,storage tolerance and good taste has become a two-way pursuit between growers and consumers.The occurrence of common diseases has become an important factor affecting tomato yield.Large-scale planting is a heavy task for disease control.Traditional diagnosis methods can easily lead to a long diagnosis period and misdiagnosed diseases and so on.With the development of computer technology and the application of image processing technology,the two technologies are gradually transported at home and abroad.In the field of crop disease identification,good results have been obtained in the field of crop disease identification and analysis.Usually morbid tomato plants will show physiological structure,shape characteristics of the changes.Therefore,the image of tomato leaf disease is taken as the research object and genetic algorithm and intelligent classification model are used to segment and recognize the disease leaf,which is of great significance to the improvement of tomato yield and quality as well as the identification of crop diseases.The main work of this paper is as follows:(1)the images of tomato diseases are taken from crop Research Network and Agricultural Economic Forum.The common diseases of eggplant were early blight,late blight and gray mildew.Through the study of tomato disease symptoms,the difference of RGB color components of leaves in space and the brightness and hue of the disease site were studied.The shape and so on can be used as the basis for the research.(2)Three kinds of tomato leaf images were pre-treated: enhancement,sharpening,mathematical morphology and so on.Use in Segmentation And 2D Algorithm combined with algorithm for blade processing,2-D The maximum inter-class variance method is used to obtain the segmentation threshold of the image threshold at the maximum difference,while the genetic algorithm provides the parameters with the highest correlation for iterativeoptimization,and the maximum value of the population fitness under the threshold difference function is the optimal threshold solution,which is called the optimal threshold solution,and the maximum value of the population fitness under the threshold difference function is the optimal threshold solution.At the same time,with the watershed method,GA,2D The algorithm is compared.(3)The color parameter extraction of tomato disease spot region: the color quantity of non-binarization image in different color space is different,to a certain extent,the color component contains various information,the disease color feature can be used as the basis of feature extraction.In addition,because the color space(hue,saturation,brightness)is similar to the principle of human eye color recognition,to convert the image to color space,12 statistical features are extracted as color feature parameters.In view of the difference in shape of tomato disease characteristics,the experimental samples used in this paper have obvious differences in shape characteristics,which are manifested in the size,perimeter and axis length of the disease spot.,roundness,etc.,the final shape feature parameters extraction includes 11 statistical features;Texture features are difficult to be observed by naked eyes.The experience of reference researchers and experts usually use gray-level co-occurrence matrix to extract texture features such as energy,correlation,contrast,entropy,and so on.The mean and variance of the four features in the texture of the image are selected to total 8feature parameters.(4)The feature vectors are screened by principal component analysis(PCA),and the support vector machine(SVM)classification model based on genetic algorithm and the support vector machine swarm intelligent classification model based on particle swarm optimization(PSO)are added.In terms of parameter determination of genetic algorithm and parameter determination of particle swarm optimization function is introduced to improve the ability of parameter optimization.The classification effect of the three diseases in the same classifier was verified by several experiments in the model with different parameters.The three experimental models retain the experimental group whose recognition rate is more than90% as an index to evaluate the performance of the classification model.The recognition rate of the classification model for gray mold is 93.87.The average classification rate of target experimental group was calculated,and the recognition rate of early blight disease was more than 90% in the three classification models.