VEHICLE DETECTION METHOD USING MULTI-SCALE MODEL

摘要

The present invention relates to the technical field of vehicle detection, in particular to a vehicle detection method using a multi-scale model. The present invention comprises three steps: multi-scale model construction, multi-scale model learning, and vehicle detection. Multi-scale model construction is realized by utilizing more than two different mixed image templates; multi-scale model learning means picking up vehicle images from actual traffic images as training graphics, and studying edge blocks, texture blocks, color blocks, flatness blocks and images likelihood of the mixed image templates; vehicle detection means performing template matching on traffic images by utilizing the mixed image templates so as to detect target vehicles. The present invention has advantages such as adapting to various weather conditions, to a certain degree of vehicle deformation and the like, is particularly capable of accurately detecting vehicles at different distances from a camera, and can be applied to detecting vehicles in a video.