An Improved Canny–Hough Algorithm for Lane Detection

Abstract

Lane detection is a key perception task in autonomous driving. However, traditional lane detection methods based on edge extraction and Hough transform often struggle to balance detection accuracy and real-time performance under complex driving conditions. To address this issue, this study proposes a lightweight lane detection framework based on an improved Canny–Hough algorithm. The proposed method is designed as a coordinated pipeline that enhances edge extraction, adaptive thresholding, and constrained Hough voting in a unified manner. Specifically, a 3×3 multi-direction gradient operator, interpolation-based non-maximum suppression, and Otsu-based adaptive thresholding are introduced to improve the quality and stability of lane-edge extraction. In addition, region-of-interest masking and polar-angle restriction are incorporated into the Hough transform to reduce non-lane interference and computational redundancy. Experimental results on the TuSimple dataset show that the proposed method achieves a recognition rate of 91.72% and an average detection time of 32.45 ms. Compared with the traditional Hough-based method, the recognition rate is improved by 13.30 percentage points (16.95% relative to the baseline), while the detection time is reduced by 39.11%. Ablation and comparative experiments further verify that the proposed framework achieves a favorable trade-off between detection accuracy and efficiency. The proposed method provides an effective and lightweight solution for lane detection, with promising applicability under the tested TuSimple-based forward-view setting in CPU-based and resource-constrained deployment scenarios.