Unmanned aerial vehicles (UAVs) and Unmanned ground vehicles (UGV s) have complementary sensing characteristics, and their cooperative sensing can effectively solve the problems of target loss due to view occlusion or distance in multi-target tracking tasks. In this paper, we propose a track by detection framework based on a multi-source data fusion approach. In the detection phase, a modified ImvoxelN et network with attention mechanism is proposed to fuse the multi-view image data of unmanned vehicles and UAVs to estimate the position and size of the target. In the tracking stage, the method involves correlation processing of both 2D and 3D detection data. It utilizes an interactive multi-mode unscented Kalman filter (IMM-UKF) algorithm that incorporates the Minimum-update Successive Shortest Path (muSSP) method to enhance the accuracy and real-time performance of dynamic target tracking. The experimental results demonstrate the effectiveness of the proposed approach in significantly improving the performance of multi-target tracking tasks.
UGV-UAV Cooperative 3D Multi-Object Tracking Based on Multi-Source Data Fusion
2023-10-13
950652 byte
Conference paper
Electronic Resource
English
Multiple camera fusion for multi-object tracking
IEEE | 2001
|Multiple Camera Fusion for Multi-Object Tracking
British Library Conference Proceedings | 2001
|