An equalized global graph model-based approach for multi-camera object tracking

Abstract— Non-overlapping multi-camera visual object tracking typically consists of two steps: single camera object tracking and inter-camera object tracking. Most of tracking methods focus on single camera object tracking, which happens in the same scene, while for real surveillance scenes, inter-camera object tracking is needed and single camera tracking methods cannot work effectively. In this paper, we try to improve the overall multi-camera object tracking performance by a global graph model with an improved similarity metric. Our method treats the similarities of single camera tracking and inter-camera tracking differently and obtains the optimization in a global graph model. The results show that our method can work better even in the condition of poor single camera object tracking. These problems are inevitable as long as the multi-camera object tracking is solved in two steps. We address these problems by integrating the two separate modules and jointly optimising them. We develop a global multi-camera object tracking approach. It integrates two steps together via an equalized global graph model to avoid these “inevitable” problems and aims to improve the overall performance of multi-camera object tracking. To integrate these two data associations, the straightforward idea is to establish a new data association which takes initial observations as inputs and outputs the final trajectories directly. However, a new problem arises, i. e. how to measure the similarity between two observations in the new graph. Some similarities are from the observations which belong to the same camera, and others are from those belong to different cameras. < final year projects >