Multi-part pedestrian tracking algorithm with fuzzy decision for partial occlusion
Lin Mao, Yehao Xu, Fan Cheng, Rubo Zhang · 2018
For the obstacle of the traffic sign and facilities causing different degrees of occlusion to up-right person tracking system on-board visual equipment, a multi-part pedestrian tracking algorithm taking advantage of multi particle filters is proposed for solving the partial occlusion tracking failure problems that resulting in a wrong bounding box of the whole body. A novel spring elastic-load decision was introduced to coordinate and calibrate the actual position of each person's part, in order to decrease the error risk only using a single particle tracker and increase the visual object tracking robust under the partial occluded conditions. With the simulation results, the algorithm achieves better performance compared to only single tracking filter in partial occlusion of pedestrian, and weakens the wrong tracking effect in complex street background, so that it can provide a support for driving assistance visual object obstacle avoidance applications.