Study On Wheat Disease Recognition Based On Support Vector Machine

China is the world’s total wheat output, consumption of the largest countries, wheat acreage accounted for 22% of the total area of China’s grain crops, grain output accounted for more than 20% of total output. However, the occurrence of wheat diseases seriously harms the wheat output and quality.The technology of image processing, machine learning and so on is very important for the introduction of wheat disease recognition and diagnosis.In view of the current wheat disease image recognition accuracy is not high the, in this paper, the author uses the superior classification performance of support vector machine(SVM), has conducted the research to the wheat powdery mildew, leaf blight and stripe rust,leaf rust, four types of wheat disease images.Firstly, pretreatment of wheat disease image collection of: wheat diseases image acquisition when is inevitably affected by external noise interference, for feature extraction and classification of impact. In view of this situation, treatment means of the current commonly used pre to effectively integrate, avoid duplication and unnecessary steps in the process of pretreatment, find suitable for wheat disease image preprocessing work flow,ensure the authenticity and reliability of the later work.Secondly, the extraction of the wheat disease image color feature and texture feature: at present crop disease identification in the feature extraction methods varied, and did not fully consider the characteristics of color and texture. This paper combines these features and influence of considering various characteristics of wheat disease image classification, find out a series of suitable for wheat powdery mildew, leaf blight and stripe rust and leaf rust recognition feature classification parameters.Finally, the support vector machine(SVM) on wheat disease image recognition: for the common several kinds of multi class classification support vector machine(SVM) was studied, and compared their advantages and disadvantages by "one to many" method to establish multi class classification support vector machine and the four kinds of wheat diseases in recognition, the accuracy rate reaches 98.33%.