LiDAR-Camera Fusion for Static Map Construction: Leveraging Dynamic Feature Detection and Filtering

Maps are essential for precise robot localization and navigation. However, the presence of dynamic obstacles, as well as obstacles with the ability to move, can alter the map's local structure, rendering it inaccurate. Inaccurate maps hinder navigation by impairing path planning and localization accuracy. To address this issue, we propose a new method for obstacle point cloud filtering based on laser and vision fusion. We use object detection algorithms to detect obstacles (including pedestrians and vehicles) in images. It subsequently extracts the point cloud of detected objects using the relative relationship between the two sensors, employs clustering algorithms for precise extraction, and filters these object point clouds. We propose an innovative method for constructing static maps based on SLAM. This approach involves assembling the map using the acquired static point cloud and utilizing the original point cloud for localization. We conducted relevant experiments in a campus environment. The results demonstrate the effectiveness of our proposed method in filtering obstacles and constructing static maps in outdoor scenes.