Adaptive Polymorphic Fusion-Based Fast-Tracking Algorithm in Substations
Wenbin Shi, Jingsheng Lei, Xingli Gan, Zhongguang Yang · Mobile Information Systems · 2021
Tracking multiple objects in a substation remains a challenging problem since pedestrians often overlap together and are occluded by infrastructures such as high-tension poles. In this paper, we propose an adaptive polymorphic fusion-based fast-tracking algorithm to address the problem. We first leverage the fast segmentation algorithm to obtain the fine masks of pedestrians and then combine the motion and performance information of pedestrians to realize the fast-tracking in substations. Our model is evaluated on the widely used MOT19 dataset and real-substation scenarios. Experimental results demonstrate that our model outperforms state-of-the-art models with a significant improvement in the MOT19 dataset and occlusion cases in substations.