Research On Rice Pests Warning System Based On Image Recognition

Hunan is the hometown of hybrid rice and rice is the most important crop. National’s economy and the people’s livelihood is relied on the safety and guarantee of the grain production. The study of this essay is national science and technology support project, the common and key technology research of agricultural IOT’S basis platform. Through the video system deployed in the rice field of Quyuan’s Huizhong grain organization, the growth of rice and insect pests images are collected. By the means of pest recognition and counting from image, the prediction of rice pests is realized by insects density analysis.The main work of this essay includes:1) Disposed of denoising and image equalization of pictures photoed under the natural environment to eliminate the impurity.2) In order to determine the highest accuracy and the most stable method, this paper analyzes the insects recognition effect through region segmentation algoritlim and edge segmentation algorithm. 3) Make a compare of the insect identification and counting effect between realization by different segmentation operators.4) Use improved Canny operator to realize the edge segmentation and the counting of insects image.5) Encode of pest warning system.The innovation of this paper is improvement on the Canny operator. The improvement of the algorithm is mainly in two aspect. One is gray stretch after image smoothly in the Canny. It is to prevent the loss of the edge in the subsequent edge detection, which leads to poor results in experiments. The other one is that the selection of self-adaptive threshold method is added in Canny operator. This is in order to improve the self-adaptive ability to determine adaptive threshold in Canny operator. The application of the improved Canny operator to realize the image to count the number of insects is divided into two steps. Firstly, extract the target image, and then count insects. The experimental results show that, the application of this method can get accurate results in insects recognition and counting the insects number in images with stable environment background.Without regarding of insects classification, only the density of insect in image is considers in this insects warning system, which is a forecasting pest system based on fuzzy recognition.