Recursive event model improves multi-camera tracking across gaps
ReWorld-Track: A Recursive Event World Model for Language-Guided Multi-Camera Tracking
Computer Vision and Pattern Recognition
Summary
Tracking a person or object across multiple camera views is hard when the target disappears for a while because guesses about their identity can be wrong. The authors propose a model called ReWorld-Track that keeps track of uncertainty about where the target might be and how they look as they move between cameras. This helps to make better predictions about which camera will see the target next and when, reducing mistakes that happen when the system is too sure about early guesses. The method improves identity tracking and timing accuracy on challenging multi-camera benchmarks.
What this means in practice
- •For security monitoring teams: Improve continuous identity tracking of people moving through interconnected camera networks despite occlusions or gaps in view.
- •For smart city infrastructure developers: Enhance prediction of pedestrian or vehicle movements across city cameras using language cues along with visual data.
Authors
Haoyang Wu, Shoudong Han, Chaoyue Li, Sijia Chen, Zhenyang Xie, Wang sihan
Abstract
Language-guided multi-camera tracking must preserve a target identity across unobserved gaps, where similar candidates and uncertain returns can make early associations unreliable. A wrong match can corrupt the history used to predict later observations and propagate identity errors across subsequent camera handoffs. We propose ReWorld-Track, a recursive event world model that carries association uncertainty into future predictions. Candidate matches and continued waiting define alternative target states, whose posterior probabilities are used to update a persistent recurrent belief. This representation preserves uncertainty about alternative trajectories through successive observations. This belief predicts the next camera, arrival time, and entry region, while appearance and language evidence guide association. By training across successive handoffs, the model learns to retain uncertainty that remains useful for later predictions and identity decisions. ReWorld-Track achieves HOTA scores of 65.19 on CityFlowV2 and 45.36 on MTMMC, with improved identity continuity across repeated handoffs. On MTMMC, its structured posterior update gains 0.50 HOTA points over a similarly sized generic updater and 0.94 points over fixed-moment soft association, raising next-camera accuracy from 86.03% to 87.41% and reducing median arrival-time error from 0.78 s to 0.71 s for subsequent target returns.