MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment

Scene understanding and multisource data fusion are critical challenges in autonomous self-driving systems.In particular, optimizing information fusion strategies for three-dimensional Bird’s Eye View (BEV) scene recognition tasks is crucial for accurate perception and decision-making in...

Full description

Saved in:
Bibliographic Details
Main Authors: Xiubin Cao, Yifan Li, Hongwei Li
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10979852/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850191888494100480
author Xiubin Cao
Yifan Li
Hongwei Li
author_facet Xiubin Cao
Yifan Li
Hongwei Li
author_sort Xiubin Cao
collection DOAJ
description Scene understanding and multisource data fusion are critical challenges in autonomous self-driving systems.In particular, optimizing information fusion strategies for three-dimensional Bird’s Eye View (BEV) scene recognition tasks is crucial for accurate perception and decision-making in dynamic environments. This study proposes a novel architecture that integrates multiscale feature extraction and crossmodal structural alignment to enhance the representation and detection capabilities of BEV features. Specifically, we employ a DCN-based block for visual feature extraction, comprising layer normalization (LN), feedforward networks (FFNs), and the Gaussian Error Linear Unit (GELU) activation function, aligned with the Vision Transformer (ViT) paradigm to improve feature modeling. To fully utilize multiscale information, a dedicated multiscale feature fusion block is introduced to extract expressive scene features within the feature space. Furthermore, we leverage LiDAR to generate LIDAR BEV features and propose a feature alignment block to enhance the complementarity between camera and LiDAR BEV features. The proposed architecture effectively supports precise scene recognition and adaptive decision-making in multi-sensor fusion environments, providing robust perception capabilities for autonomous driving in complex scenarios.
format Article
id doaj-art-a933693874734175985eebb154d6dfda
institution OA Journals
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-a933693874734175985eebb154d6dfda2025-08-20T02:14:45ZengIEEEIEEE Access2169-35362025-01-0113757077571710.1109/ACCESS.2025.356532810979852MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and AlignmentXiubin Cao0https://orcid.org/0009-0002-6438-3058Yifan Li1Hongwei Li2https://orcid.org/0000-0001-6231-4126School of Geo-Science and Technology, Zhengzhou University, Zhengzhou, ChinaInstitute for Geophysics and Meteorology, University of Cologne, Cologne, GermanySchool of Geo-Science and Technology, Zhengzhou University, Zhengzhou, ChinaScene understanding and multisource data fusion are critical challenges in autonomous self-driving systems.In particular, optimizing information fusion strategies for three-dimensional Bird’s Eye View (BEV) scene recognition tasks is crucial for accurate perception and decision-making in dynamic environments. This study proposes a novel architecture that integrates multiscale feature extraction and crossmodal structural alignment to enhance the representation and detection capabilities of BEV features. Specifically, we employ a DCN-based block for visual feature extraction, comprising layer normalization (LN), feedforward networks (FFNs), and the Gaussian Error Linear Unit (GELU) activation function, aligned with the Vision Transformer (ViT) paradigm to improve feature modeling. To fully utilize multiscale information, a dedicated multiscale feature fusion block is introduced to extract expressive scene features within the feature space. Furthermore, we leverage LiDAR to generate LIDAR BEV features and propose a feature alignment block to enhance the complementarity between camera and LiDAR BEV features. The proposed architecture effectively supports precise scene recognition and adaptive decision-making in multi-sensor fusion environments, providing robust perception capabilities for autonomous driving in complex scenarios.https://ieeexplore.ieee.org/document/10979852/Autonomous drivingbird’s eye viewmultiscale feature fusionLiDARfeature alignment
spellingShingle Xiubin Cao
Yifan Li
Hongwei Li
MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
IEEE Access
Autonomous driving
bird’s eye view
multiscale feature fusion
LiDAR
feature alignment
title MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
title_full MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
title_fullStr MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
title_full_unstemmed MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
title_short MSFA-BEVNet: Optimization of BEV Scene Recognition Driven by Multiscale Feature Fusion and Alignment
title_sort msfa bevnet optimization of bev scene recognition driven by multiscale feature fusion and alignment
topic Autonomous driving
bird’s eye view
multiscale feature fusion
LiDAR
feature alignment
url https://ieeexplore.ieee.org/document/10979852/
work_keys_str_mv AT xiubincao msfabevnetoptimizationofbevscenerecognitiondrivenbymultiscalefeaturefusionandalignment
AT yifanli msfabevnetoptimizationofbevscenerecognitiondrivenbymultiscalefeaturefusionandalignment
AT hongweili msfabevnetoptimizationofbevscenerecognitiondrivenbymultiscalefeaturefusionandalignment