Curve Lane detection is an indispensable part of auto-sensing technology, and plays a key role in auto-driving. The traditional curve lane detection methods are based on Hough transform and YOLOv5 Algorithm. Based on Hough transform method, the detection effect is not ideal when the lane line is discontinuous and short, that is, the dotted line is prone to detection errors. Yolov5 Algorithm is not effective for continuous lane detection or long lane detection. In this paper, deep learning based on the combination of Hough transform and Yolov5 Algorithm is studied, which can effectively detect the solid line and dotted line, and can accurately detect the curve. The experimental results show that the curvature radius has a good detection effect in the case of large or small curvature radius, which is of great significance for reducing the probability of traffic accidents on the curve section.