The Identification And Diagnosis Of The Flying Moths And Harm Region Of The Rice Leaf Folders On Rice Canopy Based On Image Processing

Rice leaf folder is a kind of rice pest. At present, the forecasting methods of rice leaf folder in China are manual driving moths in the rice paddy and the survey of roll leaf rate which is facing problems such as low efficiency, hard labour, and human error. The result of video tracking and counting of rice leaf folder moths by Li Yang showed that 72.2% of moth targets can be tracked successfully and 75.9% of non targets are mistaken as flying targets. The error detection rate of moth targets is relatively high. Based on the results of Li Yang, this paper continues to study automatic tracking and counting of flying moths in manual driving moths method and automatic judgment of harm level by image processing. The main content, method and result of the paper are as follows:(1) It researches on tracking and counting of rice leaf folder moths by the Moving Object Auto-tracking Technology under complicated background. The flying moths are driven by a bamboo pole stirring the rice plants. There are many moving targets in video picture including rice leaf folder moths and swing leaves rice. Firstly,this paper extracts moving targets in the video by three-image difference. Then, RGB color space model is adopted to filter the background and remove the background noises. Next, the current video frame is processed through the binaryzation of color threshold value to show the motion target region. Finally, color features of target area are extracted for identifying the moths target and pseudos target by a SVM classifier.Due to the sustained flight of moth in the video frame, Particle filtering and Kalman filtering tracking technology are adopted in this paper to track the moving targets which are detected in every frame. Tracking and counting method of Rice leaf folder moths includes the initialization of target location, the prediction of target position and the updating of the parameters of filter. The results show that 77.1 % of moth targets can be tracked successfully, but 34.5% of non targets are mistaken as flying moth targets simultaneously.(2) By combining the saliency segmentation algorithm based on global contrastwith the super-color space segmentation algorithm, the harm region by rice leaf folder is segmented and the harm level by rice leaf folder is calculated by the proportion of damage area and the whole global area in image. Based on positive rate in ROC theory, the segmentation results are evaluated by the proportion of pixel number of correctly divided to foreground area and the pixel number of global foreground area.The results prove that the positive rate is 85.8%.This paper studies the automatic tracking and counting method of rice leaf folder moths and harm region segmentation method of rice leaf folder. The results provide a reference for intelligent prediction and diagnosis technique of rice leaf folder based on video picture.