Fusion of Wireless Signal and Computer Vision for Identification and Tracking

Abstract

In public safety scenarios, target objects identification and tracking is an important application, and two positioning methods including wireless and computer vision are respectively used for applications. In this article, we combine the wireless signal and computer vision, and propose a novel object identification and tracking technology. The positioning method based on computer vision helps to improve the accuracy of positioning, and we can easily distinguish different users according to wireless device information. Based on our proposed trajectory association technology, the visual trajectory is accurately matched to the corresponding wireless trajectory, and the identity of the visual trajectory is confirmed. Combined with the analysis of the position change and appearance change of visual objects, wireless positioning results are fused to correct the affected visual trajectory to improve overall system performance. A tracking system was deployed in the real world. The fusion path is proved to be closer to the real path and 90% of the errors were less than 1m. We have also implemented large-scale simulation experiments to evaluate our approach. The results show that our association algorithm has a high matching success rate and is insensitive to synchronization errors.