Multi-Camera Object Fusion Tracking Model for Autonomous Driving.

In order to solve the problems such as vehicle trajectory interruption and frequent ID switches caused by the limited perception range of a single camera in autonomous driving, a real-time multi-camera object fusion tracking algorithm is proposed, which combines multiple cameras' detected objects, and automatically extends the autonomous vehicle's perception range to 230 meters, providing a larger perception range and more stable and continuous historical trajectory input of surrounding vehicles for the downstream trajectory prediction module in autonomous driving. The experimental results on the public and real-vehicle datasets show that our method can effectively realize the real-time fusion tracking of multiple cameras' multiple objects under the perspective of the autonomous driving vehicle, and achieve high accuracy and robustness under the premise of satisfying the real-time requirements of autonomous driving.