A lane detection based on machine vision for lane departure warning system

The vision-based lane departure warning system aids individuals in accurately assessing lane departure even amidst distractions, consequently diminishing the likelihood of traffic accidents arising from such deviations. The advancements in deep learning within the field of image processing have rendered deep learning-based image processing techniques significantly more robust when compared to traditional approaches. In recent years, a plethora of practical lane inspection algorithms based on deep learning have been proposed, showcasing their effectiveness in the field. By leveraging the lane lines detected through deep learning, the integration of a lane departure warning module has the potential to enhance the stability and precision of the lane departure warning system. This paper proposed a fast lane detection method based on deep learning. To enhance the algorithm's focus on the target, the detection algorithm incorporates the Convolutional Block Attention Module (CBAM). Diverging from conventional lane departure warning approaches, our proposed lane departure warning system utilizes the angle formed between the lane lines and the horizontal axis within the image to accurately identify instances of lane departure. Initially, we extracted the coordinates of the left and right lane lines adjacent to the vehicle body, and subsequently synthesized a straight line utilizing the least squares method. Subsequently, the deflection angle of the vehicle was estimated based on the included angle between the lane line and the horizontal axis within the image. Ultimately, this algorithm determines the distance between the vehicle body and the lane line, enabling accurate lane departure warning. In this paper, we propose an experimental verification to validate the effectiveness of the approach.