Method

Multi-object tracking algorithm based on multi-level data association [MLA-MOT]


Submitted on 3 Dec. 2024 07:19 by
Victor Huang (西北工业大学)

Running time:0.1 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
For detection, we fuse image and point cloud data
to avoid the complexity of training point cloud-
based neural networks. Specifically, we use a YOLO
algorithm with an attention mechanism to detect
targets in images, and then extract their 3D
location features using the spatial relationships of
the sensors. In tracking, we design a multi-level
data association algorithm that groups data
associations based on the completeness of the 3D and
2D location information of the targets. This
approach helps reduce the impact of target occlusion
on tracking accuracy. Our algorithm enables low-cost
robots to track dynamic targets in real-time in the
environment.
Parameters:
alpha=0.4,iou_thres=0.45,score_thr=0.5
Latex Bibtex:

Detailed Results

From all 29 test sequences, our benchmark computes the commonly used tracking metrics CLEARMOT, MT/PT/ML, identity switches, and fragmentations [1,2]. The tables below show all of these metrics.


Benchmark MOTA MOTP MODA MODP
CAR 64.98 % 81.69 % 65.06 % 86.72 %
PEDESTRIAN 30.39 % 77.20 % 30.54 % 95.01 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 70.31 % 96.89 % 81.48 % 26441 850 11167 7.64 % 28446 648
PEDESTRIAN 33.90 % 91.65 % 49.50 % 7880 718 15362 6.45 % 8853 185

Benchmark MT PT ML IDS FRAG
CAR 28.92 % 42.77 % 28.31 % 28 339
PEDESTRIAN 11.68 % 27.84 % 60.48 % 35 354

This table as LaTeX

[1] K. Bernardin, R. Stiefelhagen: Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. JIVP 2008.
[2] Y. Li, C. Huang, R. Nevatia: Learning to associate: HybridBoosted multi-target tracker for crowded scene. CVPR 2009.


eXTReMe Tracker