Road environment semantic segmentation based on enhanced U-Net network for camera and LiDAR fusion

Accurate segmentation of drivable areas within roads and obstacles around vehicles provides a prerequisite for subsequent driving decisions. Currently, there have been numerous mature studies on the semantic segmentation of road environments based on visual images. However, these methods still struggle to eliminate the impact of ambient light on segmentation results. LiDAR, being independent of lighting conditions, can acquire high-precision depth information. The fusion of these two modal data sources can yield more accurate segmentation results. Nevertheless, effective fusion of these modalities remains a challenge. In this paper, we propose a method for semantic segmentation of road environments based on the fusion of camera and LiDAR multimodal data using an improved U-Net network.