Towards Effective Multi-Moving-Camera Tracking: A New Dataset and Lightweight Link Model

TL;DR

The paper addresses pedestrian tracking across multiple moving cameras by introducing the MMCT dataset and a lightweight appearance-free global link model called Linker to reduce identity switches in SCT. It integrates a color-transfer module to stabilize cross-camera appearance, enabling more reliable inter-camera data association. Empirical results show that Linker improves IDF1 across multiple SCT trackers and that color transfer boosts cross-camera tracking performance, with MMCT offering a challenging, large-scale benchmark for MTMMC. The work demonstrates practical gains toward coordinated multi-camera perception in on-road scenarios and provides a scalable framework adaptable to new SCT methods.

Abstract

Ensuring driving safety for autonomous vehicles has become increasingly crucial, highlighting the need for systematic tracking of on-road pedestrians. Most vehicles are equipped with visual sensors, however, the large-scale visual data has not been well studied yet. Multi-target multi-camera (MTMC) tracking systems are composed of two modules: single-camera tracking (SCT) and inter-camera tracking (ICT). To reliably coordinate between them, MTMC tracking has been a very complicated task, while tracking across multiple moving cameras makes it even more challenging. In this paper, we focus on multi-target multi-moving-camera (MTMMC) tracking, which is attracting increasing attention from the research community. Observing there are few datasets for MTMMC tracking, we collect a new dataset, called Multi-Moving-Camera Track (MMCT), which contains sequences under various driving scenarios. To address the common problems of identity switch easily faced by most existing SCT trackers, especially for moving cameras due to ego-motion between the camera and targets, a lightweight appearance-free global link model, called Linker, is proposed to mitigate the identity switch by associating two disjoint tracklets of the same target into a complete trajectory within the same camera. Incorporated with Linker, existing SCT trackers generally obtain a significant improvement. Moreover, to alleviate the impact of the image style variations caused by different cameras, a color transfer module is effectively incorporated to extract cross-camera consistent appearance features for pedestrian association across moving cameras for ICT, resulting in a much improved MTMMC tracking system, which can constitute a step further towards coordinated mining of multiple moving cameras. The project page is available at https://dhu-mmct.github.io/.

Figures (10)

• Figure 1: The workflow of our MTMMC tracking, composed of three major parts: single camera tracking (SCT), global link model of Linker for refining SCT results, and cross-camera tracklet matching.
• Figure 2: The workflow of the Linker model. LA refers to the linear assignment algorithm. Given the output short tracklets of SCT trackers and the similarity score, a tracklet graph is constructed. After that, A modified Jonker-Volgenant algorithm is leveraged to solve the assignment problem.
• Figure 3: Tracklet embedding association in multi-camera tracking. A color transfer module is employed to extract cross-camera consistent features.
• Figure 4: Different driving cases considered during the data collection. Some possible exampled pedestrian movements in green color are also shown.
• Figure 5: Overview of the MMCT dataset. Each colored trajectory on the left represents a driving segment from a moving camera. Four images on the right show exemplary road scenes.