RLM-Tracking: Online Multi-Pedestrian Tracking Supported by Relative Location Mapping

TL;DR

This paper introduces RLM-Tracking, a real-time multi-pedestrian tracking method that improves accuracy by modeling relative location and region density, effectively handling occlusions and reducing ID switches.

Contribution

The paper presents a novel multi-object tracker incorporating Relative Location Mapping and Target Region Density models for enhanced accuracy and efficiency.

Findings

Significant improvement in HOTA and DF1 metrics on MOT17 and MOT20 datasets.

Enhanced tracking accuracy with real-time processing capabilities.

Better handling of occlusion and object density variations.

Abstract

The problem of multi-object tracking is a fundamental computer vision research focus, widely used in public safety, transport, autonomous vehicles, robotics, and other regions involving artificial intelligence. Because of the complexity of natural scenes, object occlusion and semi-occlusion usually occur in fundamental tracking tasks. These can easily lead to ID switching, object loss, detect errors, and misaligned limitation boxes. These conditions have a significant impact on the precision of multi-object tracking. In this paper, we design a new multi-object tracker for the above issues that contains an object \textbf{Relative Location Mapping} (RLM) model and \textbf{Target Region Density} (TRD) model. The new tracker is more sensitive to the differences in position relationships between objects. It can introduce low-score detection frames into different regions in real-time according to the density of object regions in the video. This improves the accuracy of object tracking without consuming extensive arithmetic resources. Our study shows that the proposed model has considerably enhanced the HOTA and DF1 measurements on the MOT17 and MOT20 data sets when applied to the advanced MOT method.