基于模板匹配的多模式车辆跟踪算法

Algorithm for Multimode Vehicle Tracking Based on Template Match

动态目标识别与跟踪是计算机视觉研究的热点问题,为实现汽车辅助驾驶系统中复杂背景下多目标的稳定跟踪,提出了一种基于灰度图像模板匹配的多模式车辆跟踪算法。根据不同模式下车辆灰度图像的特点,分别采用目标边缘特征更新模板、切割分块模板和逆转模板实现了车辆在正常行驶、遮挡及边界条件下的稳定跟踪。算法应用于道路多个车辆目标跟踪提高了目标形变和遮挡情况下的跟踪鲁棒性,采用改进的金字塔加速算法提高了跟踪速度。

The recognition and tracking of dynamic target is the hot topic of the computer vision research. In order to realize the stably tracking of multi-objectives under complex background in driver assistant system, a multimode vehicle tracking algorithm was put forward based on the match of gradation image template. According to the feature of vehicle gradation image under different mode, the target verge feature＇s update template, the cut piecemeal template and the reverse template were adopted, and the stably tracking of vehicle was achieved under the condition that the vehicle is in normal running, occlusion and boundary condition. The algorithm is applied in the tracking of multi-objectives, and the tracking robustness is enhanced in the situation of target deformation and occlusion, and the track speed is improved with the improved pyramid acceleration algorithm.