Improved transcenter vehicle tracking algorithm based on attention mechanism feature fusion

Qiang Yue

doi:10.1117/12.2689615

7 September 2023 Improved transcenter vehicle tracking algorithm based on attention mechanism feature fusion

Qiang Yue

Proceedings Volume 12790, Eighth International Conference on Electromechanical Control Technology and Transportation (ICECTT 2023); 127904O (2023) https://doi.org/10.1117/12.2689615
Event: 8th International Conference on Electromechanical Control Technology and Transportation (ICECTT 2023), 2023, Hangzhou, China

Abstract

Vehicle target tracking based on surveillance video has important application value for road running state perception and abnormal event detection, and has always been a research hotspot in the field of intelligent transportation. Since the advent of Transfomer, it has achieved superior performance in various tasks of computer vision. The multi-target tracking framework based on Transfomer has also received extensive attention. TransCenter algorithm, as a multi-target tracking network based on Transfomer, has achieved good results in pedestrian tracking tasks. However, in the top-down feature fusion process, the semantic information of high-level features is lost. Combined with the attention mechanism, we propose a multi-scale feature fusion strategy based on attention mechanism, which effectively utilizes the semantic information and detail information in multi-scale features. The effectiveness of the method is verified on the UA-DETRAC dataset. Experiments show that the vehicle tracking effect of this method is better than the benchmark model TransCenter

Citation Download Citation

Qiang Yue "Improved transcenter vehicle tracking algorithm based on attention mechanism feature fusion", Proc. SPIE 12790, Eighth International Conference on Electromechanical Control Technology and Transportation (ICECTT 2023), 127904O (7 September 2023); https://doi.org/10.1117/12.2689615

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available