FastRSR: Efficient and Accurate Road Surface Reconstruction from Bird's Eye View

2025-04-13

AI Summary

To address information loss and sparse feature representation caused by perspective transformation in bird’s-eye-view (BEV) road surface reconstruction (RSR), as well as the inherent trade-off between accuracy and efficiency in stereo matching, this paper proposes DAP-Net. First, a Depth-Aware Projection (DAP) module is introduced to alleviate BEV feature sparsity. Second, Spatial Attention Enhancement (SAE) and Confidence-based Attention Generation (CAG) modules are jointly employed to improve both robustness and computational efficiency of stereo matching. Finally, BEV feature aggregation is performed to enhance geometric consistency. Evaluated on the RSRD dataset, DAP-Net reduces absolute elevation error by over 6.0% compared to the best monocular method, and achieves ≥3.0× speedup over mainstream stereo-based approaches, establishing new state-of-the-art performance for BEV road surface reconstruction.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Large-scale security measurements

Abstract

Road Surface Reconstruction (RSR) is crucial for autonomous driving, enabling the understanding of road surface conditions. Recently, RSR from the Bird's Eye View (BEV) has gained attention for its potential to enhance performance. However, existing methods for transforming perspective views to BEV face challenges such as information loss and representation sparsity. Moreover, stereo matching in BEV is limited by the need to balance accuracy with inference speed. To address these challenges, we propose two efficient and accurate BEV-based RSR models: FastRSR-mono and FastRSR-stereo. Specifically, we first introduce Depth-Aware Projection (DAP), an efficient view transformation strategy designed to mitigate information loss and sparsity by querying depth and image features to aggregate BEV data within specific road surface regions using a pre-computed look-up table. To optimize accuracy and speed in stereo matching, we design the Spatial Attention Enhancement (SAE) and Confidence Attention Generation (CAG) modules. SAE adaptively highlights important regions, while CAG focuses on high-confidence predictions and filters out irrelevant information. FastRSR achieves state-of-the-art performance, exceeding monocular competitors by over 6.0% in elevation absolute error and providing at least a 3.0x speedup by stereo methods on the RSRD dataset. The source code will be released.

Problem

• Addresses information loss in BEV transformation for road reconstruction
• Improves accuracy-speed balance in BEV stereo matching
• Enhances road surface condition understanding for autonomous driving

Innovation

• Depth-Aware Projection for efficient BEV transformation
• Spatial Attention Enhancement to highlight key regions
• Confidence Attention Generation for filtering irrelevant data