Abstract: The road vanishing point detection algorithm is widely used in fields such as intelligent transportation and autonomous driving. Existing road vanishing point detection algorithms based on image texture features are easily affected by buildings such as vertical light poles in the scene, and have the problem of poor robustness in some scenes. In response to this problem, the article proposes an unstructured road vanishing point detection algorithm for complex scenes. First, the method updates the weights of voting points through the angle features of the image texture so that the image texture in the area of the vertical building is weakened; Secondly, the area growth algorithm is used to obtain the road area, and the weight of the voting points in the road area is strengthened, and then Enhanced texture extraction of regions of interest in images. Experiments show that the proposed algorithm can accurately detect the vanishing points of roads in complex scenes and has a certain degree of robustness.

Key words: vanishing point detection, Gabor filter, soft voting in local areas