Abstract:

Application of the techniques of image analysis and pattern recognition to the sorted diagnose of Tilletia disease is of significance to improving the automation of detecting the diseases of entry-exit plants.After an analysis of the shapes and texture features of the spore images of Tilletia caries (DC.) Tul.,Tilletia indica Mitra and Tilletia controversa Kühn,this paper selects six typical features including major axis,minor axis,minor axis of equivalent ellipse,area,perimeter and moment of inertia to describe the spore region.Next,a classifier for wheat disease based on the BP neural network consisting of 6 input vectors and 4 output vectors is designed for the classification of the three diseases' images.The primary tests,in which the accuracy of recognition accuracy towards 33 samples is up to 81.8%,shows high accuracy and capability of the classification of the three diseases.

Key words: BP neural network;tilletia disease;classification