Application Of Key Technology Based On Video Content Retrieval In Fruit Pest Detection

To solve the problems that the existing agricultural science and technology knowledge video retrieval is not widely used and its accuracy is not high,this paper applies based on the actual needs of farmers the content-based video retrieval technology in the field of agriculture by taking the knowledge video of fruit diseases and insect pests as an example.In a bid to meet the professional and personalized retrieval needs of farmers,shorten their time to obtain information,and finally improve the ability of farmers to get information,this paper studies from the two aspects of shot boundary detection and key frame extraction,performs shot segmentation on the lengthy fruit diseases and insect pests knowledge video,and extracts the key frames in the video.In terms of shot boundary detection,this paper proposes an improved shot boundary detection scheme for fruit diseases and insect pests knowledge video.This scheme is based on double check(initial check and re-check).During the initial check,the change value of the difference between frames in the video sequence is employed as the basis of the frame skipping value in the adaptive hopping method to improve the overall detection efficiency of the algorithm;at the re-check stage,the fusion of color features and texture features is utilized to highlight the main content of the video frame.In this paper,the multi-feature fusion method is adopted to effectively represent the main content of video image.Meanwhile,the adaptive hopping method is used to replace the conventional frame by frame calculation of the inter frame difference and study its change value so as to mitigate the impact of the flash in the video lens and the object motion in the camera/lens on the lens boundary detection and solve the problem that the conventional video lens detection depends on manually setting the threshold.The experiments demonstrate that this algorithm features fast shot segmentation speed and ideal effect(with recall rate greater than 93.5% and precision rate greater than 93.4%),which improves the video shot retrieval efficiency,cuts the time for farmers to obtain information,and improves the utilization rate of farmers’ agricultural science and technology knowledge video.In terms of key frame extraction,taking fruit diseases and insect pests knowledge video as an example,this paper puts forward an optimization algorithm for key frame extraction of agricultural science and technology knowledge video based on Sobel LBP(Sobel edge Local Binary Patterns).Firstly,in such algorithm,the Sobel LBP algorithm is fused with the difference between video frames,and the key frames are preliminarily selected.Thereafter,considering the position interval of the first extracted key frames in the original video,the redundant key frames are ruled out to obtain the secondary optimized key frames.The test results of four knowledge videos of different kinds of fruit diseases and insect pests suggest that the average value of the comprehensive index F1 of the key frame extracted by this algorithm is up to 0.91,the average accuracy is 91.35%,and the average fidelity is 92.18%.By extracting the representative pictures of the video sequence as the key frames,this algorithm narrows the memory space required to save the video,laying a foundation for forming the video summary of agricultural science and technology knowledge and creating video retrieval.