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.