Study Of Forecast System For Potato Late Blight Based On Spores Image Recognition

With the rapid development of computer technology, computers have been widely used in agriculture. Now, people pay more and more attention to automatic recognition of agricultural diseases, particular, in pathogenic microorganism's diseases. Compare with traditional biology microscopic detection, machine view technology has many advantages: high efficiency, low work intensity and workers no more need strict training to fit for the work. It will improve the speed and accuracy of diseases broadcast.This paper first research on image processing and recognition of Potato Late Blight spores. By comparing with RGB color model images and HSV color model images, putting forward a threshold segmentation method that use S-channel gray image as main image and RGB gray image as assistant image. Then using image smooth method and improved watershed algorithm processes the image to obtain separated and complete spore binary images. After pre-processing, each spore's contour characteristics that area, circle, Hu moment and so on can be exactly extracted. Through require of characteristics, 5 contour characteristics were chosen to compose spores feature vectors. They are area, circle, extent and Hu1, Hu2 moment. A three-layer artificial neural network is established by feature vectors. According to comparing the training speed and convergence of different BP algorithm, Variable Learning Rate BP Algorithm is selected to train BP neural network. Simulation results show that the artificial neural network can recognize and counted the potato late blight spores well.The forecast model is established according to the key weather conditions for the outbreak of Potato Late Blight, combining with Hyer model and integrating flying levels of spore in the air. A short-term prediction for Potato Late Blight can be gained with spore's flying conditions measured by the recognition system combining the weather forecast.Recognition count and disease prediction function described in this paper have been programmed and implemented with VC++, with a view to provide some help in future prevention and control of Potato Late Blight.