Multi-Target Multi-Camera Tracking (MTMCT) achieves continuous cross-camera tracking and spatiotemporal trajectory reconstruction through collaborative multi-camera sensing. In tunnel scenes, abrupt illumination changes and low contrast markedly degrade appearance features, impairing cross-camera association; dense and short-term complete occlusions further induce trajectory breaks and identity inconsistency. To address this, we construct a new dataset tailored to tunnel conditions to support joint optimization of the object detection network and core MTMCT tracking algorithms, and propose a tunnel-specialized spatiotemporal constraint to match the scene’s distinctive characteristics, improving tracking accuracy and reliability under complex dynamics. Extensive experiments and ablation studies on this dataset show that the enhanced MTMCT delivers significant gains in robustness and performance, better handling the complexities of tunnel scenes and establishing a new baseline for multi-target tracking in constrained environments.