Research On Image Recognition Arthmetic Of Rice Blast Spores

Rice is one of the most important grain in our country, has a critical influence on food safety. Either Rice production or its quality suffer a lot from the rice plague. To prevent rice plague, we need to detect plogue in early period and evaluate its disease degree. But It would takes so much time to observe it under microscope. Building on these problems,this paper introduces another method, using digital image processing and pattern recognition, which runs faster and cost less.This paper designs a two-stage classfier based on adaptive threshold algorithm. First introduces the research background and significance, the basic methods of digital image processing and pattern recognition, and then expounds the image recognition technology at home and abroad research status in the field of microscopic image.The first problem to solve is extracting the contour of spores, Generally, the original image of spore in rice plague, will be preprocessed first, including gray scale, histogram equaliztion, in order to enhance constrast ratio of the image. Then an adaptive threshold detect method is applied to gain binaryzation image. After denoise process, with the apdative Canny contour dectect method, we use contour extracting algorithm to dectect all the contours. Then we calculate all the valid feature of the spore by the contour and other information. The features include match ratio with the standard spore contour, minimum containg eclipse area, contour area, contour length, width heigth ratio, rounded degree and so on. These features is the key elements to the following pattern recognitionThis paper introduces two classify method, SVM(Supported Vector Machine) and decision tree classfier, and designs a two-stage classfier,which use decision at the first stage to filterate most of the negative samples and then run futher image processes to gain more valid features for the following SVM classfier. After the second stage,spores and other spots will be classify, in an accuracy of 86%。All the operation programs on Visual Studio 2010 and the result of spore dectection and the number of spores in the image are shown visually on a designed app written in Cpp.