Recognition Of Greenhouse Tomato Disease Based On Image Processing Technology

It is difficult to identify unobvious or uncommon diseases by traditional method thatis influence from human factor seriously. Greenhouse tomato diseases that including earlyblight, late blight, and leaf mold were identified by using computer vision technology anddigital image processing techniques in this thesis.This thesis uses three common tomato leaf diseases as the research object that areearly blight, leaf mold, late blight, and some ideal results were obtained on the study ofimage preprocessing, image segmentation, feature extraction, disease image and diseaserecognition.Firstly, a tomato leaf disease image was obtained through median filtering noisereduction. Secondly, through improving otsu threshold segmentation method to get imagesegmentation and accurately extract image lesion, and the result was compared with themethod before. After obtaining image lesion information, making the color spaceconversion of lesion color image, while extracting the color characteristics of the lesionregion and shape feature of the lesion. Finally, Bayesian classifier was designed by usingBayesian discrimination to establish Bayesian discriminated function. And three diseasesincluding150samples (Each group has50models) were divided into two groups, onegroup for training, and another group for inspection. The results showed that recognitionrate of early blight and late blight is92%, the identification rate of leaf mold is96%, andthis method has valid identification and a high recognition rate to tomato diseases wasproved. Then, each feature was removed individually through the experiment, and getdisease recognition rate. After comparative analysis, ultimately identified five specificrepresentative characteristics (color features u, v; shape characteristics; roundness;complexity; elongation) as the optimal characteristics (that recognition rate of early blightand late blight is92%, the identification rate of leaf mold is96%). Finally, a completeimage processing system was formed by writing procedures of the image pre-processing,segmentation, feature extraction, and disease determination with the development tool ofVisual C++6.0in the Windows XP platform.