PEJC: A Stereo Matching Algorithm for Bee Swarm Targets

To address the matching ambiguity in 3D multi-object tracking when handling large-scale bee swarms, this paper proposes PEJC (Pose-Epipolar Joint Constraint), a novel stereo matching algorithm. PEJC leverages the rotated bounding boxes of detected targets in 2D images to estimate their 3D poses, integrating pose constraints with epipolar constraints to compute matching costs. This joint approach significantly enhances the discriminability of matching candidates. Experimental results demonstrate that PEJC effectively improves tracking accuracy for 3D bee swarm targets in dense scenarios.