Subblock-weighted template matching method applied to overlappedvehicle detection
适用于遮挡车辆检测的子块带权模板匹配方法

To solve the occlusion problem in tracking of multiple vehicles, this paper proposed a subblock-weighted template matching method based on histograms of oriented gradients, which first divided the object into blocks and extracted histograms of oriented gradients for each block. By weighting different blocks into different grades using kernel function, this method calculated the similarity between the target model and the searching window based on the subblock-weighted feature matching degree measurement method, hereby, obtained the best matching region with the maximum calculation result. Finally, this paper used this method to tracking of the overlapped vehicle in multi-vehicle tracking process. The experiment result shows that this method embodies higher accuracy and robustness.