Dense Scene Multiple Object Tracking with Box-Plane Matching

TL;DR

This paper introduces a novel Box-Plane Matching method for multiple object tracking in dense scenes, combining detection filtering, appearance feature extraction, and data association to improve accuracy.

Contribution

It proposes a comprehensive framework with three modules—LADM, GAFM, and BPM—that significantly enhances MOT performance in dense environments.

Findings

Achieved 1st place on ACM MM Grand Challenge HiEve 2020 leaderboard.

Effectively filters noisy detections in dense scenes.

Improves data association accuracy using combined motion and appearance similarity.

Abstract

Multiple Object Tracking (MOT) is an important task in computer vision. MOT is still challenging due to the occlusion problem, especially in dense scenes. Following the tracking-by-detection framework, we propose the Box-Plane Matching (BPM) method to improve the MOT performacne in dense scenes. First, we design the Layer-wise Aggregation Discriminative Model (LADM) to filter the noisy detections. Then, to associate remaining detections correctly, we introduce the Global Attention Feature Model (GAFM) to extract appearance feature and use it to calculate the appearance similarity between history tracklets and current detections. Finally, we propose the Box-Plane Matching strategy to achieve data association according to the motion similarity and appearance similarity between tracklets and detections. With the effectiveness of the three modules, our team achieves the 1st place on the Track-1 leaderboard in the ACM MM Grand Challenge HiEve 2020.