The Design And Realization Of Yellow Stem Borer Identification System Based On Neural Network

In this paper, yellow stem borer, which is common in preventing and cure rice diseases and insect pest, was studied with modern information technologies, such as neural network and digital image-processing. New theory about identification yellow stern borer is based on technologies of digital image-processing and neural network. The technology of image-processing is made use of extracting features of some images. The packages of Visual C++ neutral network are made use of identifying insect pests. In this paper, not only an overall system design but also the solution scenario has been provided.The following are main research contents and achievements.(1) Pretreatment of yellow stern borer image. The paper put the emphasis in image clipping, color counterchanging and image segmentation and so on. As it hard to distinguish rice pest and ambience using by machine vision, image clipping is controlled by user. Color-counterchanging can mostly wiped off molestation information caused by color and standout the pest itself. After pretreatment image, we choose valve-way for image segmentation. The system works out the recommendation valve by self-motion and user can modify this valve by himself. In this way the best effect two-color image can be get for exacting features of images.(2) Feature extraction of yellow stern borer images. According to the conditions in practice and suggestions given by experts of rice protection, this paper chooses geometry characters and color characters of yellow stern borer. Geometry characters include area and perimeter. Color characters include red area, yellow area and green area.(3) Identification of yellow stern borer by neural network. Compared with classical statistical approaches, neural networks approach to pattern recognition has many advantages, such as self-adaptability, parallel processing, robustness and strong classification ability. So BP neural network, which has three layers, is used to identify the yellow stern borer primarily. The result shows our method is effective and satisfaction.