Research On Recongnition Of Apple Fruit Disease Based On Low Resolution Image

The apple industry is the main industry of Shaanxi province, which is also one of the six special advantage industries in our national economy. But apple diseases cause a great loss to fruit grower every year. The urgent problem is how to implement the remote intelligent identification of apple diseases. This paper takes three kinds of wide-ranging and heavy-harm diseases as research objects, including apple ring rot, apple anthracnose and new apple ring rot, studies the segmentation method of apple diseases image with low resolution and complex background and the method of feature extraction and feature selection of the disease spots, constructs the disease recognition model based on support vector machine (SVM) , develops apple diseases recognition system. This paper provides the technical support for remote apple diseases recognition timely.The main content and conclusion of this paper are as follows:(1) Considering the image characteristics obtaining from natural scene, image preprocessing method are studied. Median filter of gray image is used in H S V channel image of color image to get more clear image, which better removes the noise and preserve the edge, finite compare adaptive histogram is used to enhance the image, and the image after enhancement becomes soft and has sharp edge.(2) The image segmentation method with low resolution and complex background is studied. The segmentation results of K-means clustering method based on visual attention model and Level Set method are compared. Based on fast marching, results have shown that Level Set method based on fast marching can separate the disease spots from the original image directly.(3) The disease spots feature extraction and feature selection methods are studied. Considering the image characteristics, 2 color features, 4 normal parameters'mean and standard deviation of Gray Level Co-occurrence Matrix as texture features and 7 Hu invariant moments as shape features sum to 21 parameters are extracted. 18 parameters are selected as main features of this study based on characteristics analysis.(4) The disease image recognition method is studied. This paper selects Support Vector Machines Classification Methods based on comparative analysis of three classifier methods such as similarity measure, gray correlation analysis and support vector machine. The constructed recognition model is trained by using linear kernel function, polynomial kernel function and radial basis kernel function as kernel function of support vector machine with different training samples and testing samples. The results show that the recognition ratio is up to 90% when using 48 images(16 images every kind) as training sample, 30 images(10 images every kind)as testing sample, and linear kernel function as kernel function. This result can meet the requirement of apple diseases remotely intelligent identification.(5) This paper develops a prototype of apple fruits'diseases image with low resolution recognition using Matlab (R2010a). The results show that this system can recognize the apple disease accurately and rapidly, and which is effective for apple diseases recognition.