A Lane Detection Method Based on Multi-Indicator Scoring Mechanism

Due to the complexity of driving surroundings, current advanced driving assistance system struggle to reliably extract lane lines. Unlike most methods that select or filter potential lane lines step-by-step according to different features, this paper proposes a new detection method that creates a scoring mechanism based on lane line features, which combines multiple lane line features for comprehensive judgment, providing a new idea to detect lane lines using conventional computer vision methods. First, the region of interest (ROI) is defined and a priority pixel method is used to binarize the image after converting to grayscale. The gradient image of the binary image is then extracted using the Sobel operator and converted to an 8-bit (uint8) data format for further processing. Contour detection is then performed on the binary image, and a scoring mechanism is used to select lane contours. After that, performing an AND operation between the mask containing only the pixels within the contours of the lane line and the gradient image yields the points used to fit the lane lines. Finally, the lane lines are fitted using the least squares method. The experimental results indicate that the method proposed can significantly improve the accuracy of detection.