A Robust Multi-Modal Tracking Method for Challenging Weather Conditions

Multi-object tracking (MOT) is a critical task in autonomous driving, aiming to identify the locations and identities of objects in data collected by cameras or LiDAR sensors. However, adverse weather conditions (rain, fog, low light) degrade the quality of sensor data, thereby affecting the accuracy and stability of multi-object tracking. To address this issue, this study proposes an adverse weather simulation algorithm to generate a large amount of training data under challenging weather conditions and develops a multi-modal association tracking strategy to achieve stable multi-object tracking in such environments. Experimental results demonstrate that the proposed method achieves superior performance on the KITTI MOT dataset, with HOTA (78.36<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), MOTA (87.64%), and MOTP (85.72%). Furthermore, it exhibits robust tracking performance under various adverse weather conditions. This study provides an effective solution to the challenges of multi-object tracking in adverse weather for autonomous driving.