A Road Surveillance Video Analysis Algorithm Based on YOLO and LDA

Video analysis is a classic problem in the field of computer vision. In the specific scenario of road video surveillance, the main problem of video analysis is to find out the movement patterns of vehicles and pedestrians and classify events. Based on YOLO and LDA algorithms, this paper proposes an algorithm that uses text classification algorithm for video analysis. The algorithm first uses YOLO for multi-target tracking to extract video features, then innovatively quantizes the video feature data of multi-object tracking into text data, uses the LDA topic model for classification, and finally maps the classification results back to the video content. The core of the algorithm lies in the clever fusion of YOLO and LDA, and proposes an effective quantification mechanism. Experiments show that the proposed algorithm can effectively classify video content and complete the task of video analysis in the scenario of road surveillance video.