Method of Unsupervised Static Recognition and Dynamic Tracking for Vehicles

Vehicle object tracking is a research hotspot in computer vision. To solve the problem of single object extraction caused by the shadow effect and occlusion between vehicles, this paper presents a vehicle object tracking algorithm suitable for both dynamic and stationary states. First, the improved Canny algorithm is used to obtain the information in a video sequence, and the dynamic region of the object is extracted using the difference between the mean of the video sequence and the object frame. Secondly, the Gaussian mixture model is used for video object segmentation to obtain the foreground image and the background image, and the static object is identified through the intersection operation of the object dynamic region and the foreground image combined with the edge information. Then, chroma information is introduced into a statistical nonparametric model to eliminate the shadow of the foreground image, and the mean-shift tracking algorithm is used for dynamic object tracking of the foreground image after eliminating the shadow. The experimental results show that the proposed tracking algorithm can identify and track vehicles effectively and quickly, providing new ideas for the future development of the sensor field.