Enhancing multi-sensor classification robustness under noise via adaptive attention fusion

Vehicle classification is an important part of Intelligent Transportation Systems (ITS). While vision-based sensors struggle in harsh weather, non-vision sensors like PIR and LiDAR offer robust additional data. However, conventional fixed-weight fusion strategies often perform poorly when a single sensor becomes noisy. To address this, we propose an Attention-based Adaptive Sensor Fusion (AASF) model that can dynamically adjust fusion weights using a lightweight attention mechanism. The dataset includes 3,963 samples labeled for four main vehicle types: Buses, Cars, Motorbikes, and Trucks. According to experimental results, the AASF model achieved the highest observed accuracy of 96.3%, compared with 95.8% for CNN fusion and 94.4% for fixed-weight fusion. Under clean conditions, its performance was statistically comparable to CNN fusion, while its main advantage was observed under increasing sensor noise. When compared to data from individual sensors, the combined features from the suggested model show a distinct cluster structure and improved separability between vehicle classes through t-SNE visualization. The work specifically shows that the adaptive attention mechanism retains greater robustness against Gaussian noise, enabling the model to automatically modify weights to prioritize sensor channels that are less impacted by noise. This result confirms the potential of sensor aggregation in improving the efficiency of intelligent transportation systems (ITS).

Authors

Institutions

Publication Details

Journal
Journal of Science and Transport Technology
Published
2026-11-16
DOI
https://doi.org/10.58845/jstt.utt.2026.en.6.5.364-377
Primary Topic
Autonomous Vehicle Technology and Safety
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Enhancing multi-sensor classification robustness under noise via adaptive attention fusion

Bui Hai Dang, Vũ Toàn Thắng, Van Quang Vu, Chu Tien Dat et al.
Journal of Science and Transport Technology
Autonomous Vehicle Technology and Safety
article

Enhancing multi-sensor classification robustness under noise via adaptive attention fusion

Bui Hai Dang, Vũ Toàn Thắng, Van Quang Vu, Chu Tien Dat, Nguyen Duc Son, Nguyễn Văn Tuân, Nguyen Minh Tri
article en

Abstract

Vehicle classification is an important part of Intelligent Transportation Systems (ITS). While vision-based sensors struggle in harsh weather, non-vision sensors like PIR and LiDAR offer robust additional data. However, conventional fixed-weight fusion strategies often perform poorly when a single sensor becomes noisy. To address this, we propose an Attention-based Adaptive Sensor Fusion (AASF) model that can dynamically adjust fusion weights using a lightweight attention mechanism. The dataset includes 3,963 samples labeled for four main vehicle types: Buses, Cars, Motorbikes, and Trucks. According to experimental results, the AASF model achieved the highest observed accuracy of 96.3%, compared with 95.8% for CNN fusion and 94.4% for fixed-weight fusion. Under clean conditions, its performance was statistically comparable to CNN fusion, while its main advantage was observed under increasing sensor noise. When compared to data from individual sensors, the combined features from the suggested model show a distinct cluster structure and improved separability between vehicle classes through t-SNE visualization. The work specifically shows that the adaptive attention mechanism retains greater robustness against Gaussian noise, enabling the model to automatically modify weights to prioritize sensor channels that are less impacted by noise. This result confirms the potential of sensor aggregation in improving the efficiency of intelligent transportation systems (ITS).

Journal of Science and Transport Technology
University Of Transport Technology (VN), Trường Đại học Khoa học và Công nghệ Hà Nội (VN), Hanoi University of Science and Technology (VN)
Autonomous Vehicle Technology and Safety
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon