Eliminating Uneven Illumination for Accurate Lane Line Detection: A Dark Channel Prior Algorithm Approach 2023-01-5039

Accurately detecting lane lines remains a challenging task, especially with low-quality cameras due to the complex environment such as haze, uneven lighting, and shadows of actual roads. Despite numerous studies, lane line detection algorithms are still required to be improved for practical applications. In this work, we propose a new lane detection method that incorporates the brightness estimation concept of the single-scale retinex (SSR) algorithm into the dark channel prior (DCP) algorithm for image preprocessing. The improved DCP algorithm is used to estimate the atmospheric light intensity and remove haze noise, while simultaneously enhancing image contrast to reduce the difficulty of lane detection, especially under uneven lighting conditions, then followed by perspective transformation and HSV color space (hue, saturation, value) conversion, and finally, lane line recognition and tracking are performed using sliding windows and histogram statistics. Experimental results demonstrate that our method achieves high accuracy and robustness in lane line detection, especially under uneven lighting conditions, compared to existing methods. The proposed method can effectively handle complex situations such as haze, uneven lighting, and shadows, making it feasible for practical applications.