Optimization and Application of Multi-Target Tracking Algorithm for Indoor Service Robots Based on Enhanced TransTrack

With the popularity of indoor service robots in public and private environments, the demand for accurate and real-time multi-object tracking (MOT) systems has become increasingly urgent. This study explores the development and optimization of an MOT algorithm for indoor service robots based on a modified TransTrack model. To address challenges such as occlusion, low light, and real-time performance, this study makes several improvements to the original TransTrack architecture, including a lightweight backbone network, an occlusion handling module, and an attention mechanism pruning strategy, which improves accuracy and computational efficiency to a certain extent. The system is trained and evaluated on the Indoor-MOT dataset, and some preprocessing techniques are applied to make it adaptable to some complex indoor environments. Based on standard MOT metrics such as MOTA, IDF1, FPS, and HOTA, the improved system outperforms DeepSORT, FairMOT, and the original TransTrack in terms of accuracy and speed. In spite of its limitations in tracking fast-moving targets, the system still has certain application prospects in the field of indoor robotics.