Tomato Leaf Disease Recognition Based On Neural Network

In recent years,the area of tomato has increased,tomato disease has become the main factor limiting the quality and economic benefit of tomato.So it is very important to identify the disease type accurately in the early stage of tomato.At present,the main method to identify tomato diseases in our country is the method of artificial recognition,which can only be diagnosed according to the experience.With the rise of computer vision technology and artificial intelligence,image processing technology has been applied to various aspects,and a large number of people have begun to study how to use computer to identify crop diseases.In the early stage of the disease,most tomato diseases will cause disease spots in the leaves,so the identification of leaf diseases can be diagnosed and prevented in advance,so as to ensure the yield and quality of tomato.In this paper,we've been analyzing the techniques and pattern recognition techniques in the past and abroad,we've been focusing on the recognition of the tomato leaves disease.In this paper,a method of identifying the leaf diseases of tomato by using BP Neural network is studied,which extracts the texture feature and shape feature of leaf disease spot image.In this paper,the main work is as follows:(1)Pretreatment of diseased tomato leaf image: under the influence of various factors such as collecting equipment,light intensity and weeds dust,there will be some noise interference in the collected tomato disease images,so the image should be pretreated.Image equalization and image denoising for tomato disease image.In the image diagnosing,the mean filtering and median filtering are compared.The experimental results show that the value filtering is better.(2)Image segmentation of tomato leaves: image segmentation is the most important part of the whole image recognition process.The segmentation of the disease directly affects the accuracy of image recognition.Compared with the common color space model RGB,HIS and HSV,it was found that the disease spot under the H component of HSV was the most obvious.In this paper,the image of tomato leaf was studied by double peak method,iteration method and OTSU segmentation method.The experiment proved that the method of OTSU was the best method for segmentation.(3)Feature extraction of diseased tomato leaves image: selection of texture feature and shape feature of two extraction methods.The extraction methods of texture feature with gray level co-occurrence matrix,with energy and moment of inertia moment and entropy,correlation,deficit for the characteristic parameters were studied.Shape feature extraction is studied by using area,circle degree,rectangle degree and elongation ratio as characteristic parameters.Three characteristic parameters of different diseases were selected for disease identification.(4)Identification of disease image of tomato leaves: select BP neural network for classification identification.The seven characteristic parameters are selected as input to the neural network,the number of diseases of tomato leaves is the output.By adjusting the parameters of BP neural network,the optimal parameter training network was selected.Using the trained network to identify the image of tomato leaf disease,it was able to identify the disease,and the recognition rate was 93.5%.