QPAFusion: Query-Guided Radar Refinement and Prior-Guided Adaptive Fusion for Marine Object Detection

Accurate marine target detection is essential for maritime situational awareness and safe navigation by unmanned surface vehicles. However, uneven radar returns and uncertain spatial correspondence complicate radar–camera fusion, while input-level integration does not explicitly adapt subsequent feature processing to local radar evidence. We propose QPAFusion, which refines multi-scale features through query-guided local radar sampling and adaptively aggregates feature branches using radar priors. On the WaterScenes dataset, QPAFusion achieves 64.5%mAP50−−95 and 90.9%mAP50, exceeding the reported three-frame REVP fusion baseline by 2.0 and 2.1 percentage points, respectively, with mAP50 gains of 6.4 and 3.1 percentage points on the adverse-lighting and adverse-weather subsets. These results support combining local-radar refinement with adaptive feature aggregation for object detection on the water surface, with larger accuracy gains on the evaluated adverse-condition subsets.

Authors

Institutions

Publication Details

Journal
Remote Sensing
Published
2026-11-18
DOI
https://doi.org/10.3390/rs18223965
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

QPAFusion: Query-Guided Radar Refinement and Prior-Guided Adaptive Fusion for Marine Object Detection

Jinhe Su, Wentao Yang, Zongyue Wang, Yihan Li
3 links
Remote Sensing
Advanced Neural Network Applications
article

QPAFusion: Query-Guided Radar Refinement and Prior-Guided Adaptive Fusion for Marine Object Detection

Jinhe Su, Wentao Yang, Zongyue Wang, Yihan Li
article en

Abstract

Accurate marine target detection is essential for maritime situational awareness and safe navigation by unmanned surface vehicles. However, uneven radar returns and uncertain spatial correspondence complicate radar–camera fusion, while input-level integration does not explicitly adapt subsequent feature processing to local radar evidence. We propose QPAFusion, which refines multi-scale features through query-guided local radar sampling and adaptively aggregates feature branches using radar priors. On the WaterScenes dataset, QPAFusion achieves 64.5%mAP50−−95 and 90.9%mAP50, exceeding the reported three-frame REVP fusion baseline by 2.0 and 2.1 percentage points, respectively, with mAP50 gains of 6.4 and 3.1 percentage points on the adverse-lighting and adverse-weather subsets. These results support combining local-radar refinement with adaptive feature aggregation for object detection on the water surface, with larger accuracy gains on the evaluated adverse-condition subsets.

Remote SensingVol. 18(22)
Jimei University (CN)
Advanced Neural Network Applications
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