GFSR: Geometric Fidelity and Spatial Refinement for Reliable Lane Detection

2026-05-22

AI Summary

Existing lane detection methods struggle to accurately fit lanes with high curvature or complex topologies in challenging scenarios due to the decoupling between classification confidence and geometric quality, as well as weakened associations among sampled points. To address these limitations, this work proposes the GFSR framework, which introduces a novel LaneIoU-guided confidence calibration to establish a Collaborative Reliability Indicator (CRI) and designs an Adaptive Gated Location Refinement (AGLR) mechanism to strengthen inter-point correlations. By integrating multi-stage regression heads with an optimized non-maximum suppression strategy, the method simultaneously maintains high classification confidence and enhances geometric fidelity. The proposed approach achieves state-of-the-art performance, yielding 81.46% F1@50 and 65.01% F1@75 on CULane, and 87.35% F1@50 on CurveLanes.

Abstract

Lane detection stands as a crucial perception task in autonomous driving and advanced driver assistance systems. However, existing methods still degrade in complex real scenarios due to two major limitations. First, classification confidence only characterizes the categorical existence of lane candidates and has no strong correlation with geometric quality. If threshold filtering and NMS are conducted merely based on this confidence, the model tends to retain lane priors with high confidence while eliminating those with lower confidence but superior geometric representation. Secondly, existing regression modules weaken correlations among sampling points, hindering fine-grained optimization of distant, high-curvature and complex-topology lanes and causing underfitting. To address these issues, we propose Geometric Fidelity and Spatial Refinement (GFSR), a framework consisting of LaneIoU-guided Confidence Calibration (LCC) and Adaptive Gated Location Refinement (AGLR). Specifically, LCC adopts LaneIoU as soft supervision to explicitly estimate geometric fidelity of lane priors, which is further fused with classification confidence to construct the collaborative reliability index (CRI). This index guides threshold filtering and NMS, effectively retaining lane priors with high classification confidence and favorable geometric quality. Meanwhile, cooperating with regression heads in each refinement stage, AGLR predicts sampling point lateral offsets and adopts a gating mechanism to adaptively regulate correction magnitude, strengthen inter-point correlations and boost model adaptability as well as robustness toward complex lane scenarios. Extensive experiments on CULane and CurveLanes demonstrate that our GFSR achieves state-of-the-art performance on CULane, with F1@50 and F1@75 scores of 81.46% and 65.01%, and reaches 87.35% F1@50 on CurveLanes.

Problem

• lane detection
• geometric fidelity
• spatial refinement
• confidence calibration
• point correlation

Innovation

• Geometric Fidelity
• Spatial Refinement
• LaneIoU-guided Confidence Calibration
• Adaptive Gated Location Refinement
• Lane Detection