Road environment perception algorithm in dark scene

In order to solve the technical problem that environment perception can be efficiently realized in a dark road scene, the invention provides a road environment perception algorithm in a dark scene, light supplement enhancement is carried out on a part lacking illumination in an image through a lightweight HDR dark light image enhancement method, a plurality of photos with different exposures are superposed, and a light-weight HDR dark light image enhancement method is provided. An image with wider brightness and color range is obtained, the quality of a picture shot in a dark environment is improved, and the improved picture is input into a YOLOv8 algorithm added with SPD-Conv and PSA polarized attention mechanisms. Three tasks of road target detection, drivable region segmentation and lane line segmentation are respectively carried out through a road target detection head, a drivable region segmentation head and a lane line segmentation head which are connected to a neck bottom layer C2f in parallel. 本发明所解决的