SLAM Based on Visual Semantics and Laser Fusion

As one of the main sensors of SLAM mapping, LiDAR has been widely studied and used because of its high precision and stable performance. However, in degraded environments where geometric features such as corridors or movement along single-sided walls are scarce, it is difficult for LiDAR to obtain sufficient feature information, resulting in reduced mapping accuracy. Therefore, a SLAM mapping method based on visual semantics and laser fusion is proposed. This method not only realizes SLAM mapping through LiDAR, but also uses visual deep learning algorithm to extract the semantic information of objects, and determines the relative position of the target by combining the depth data of the depth camera. Finally, the target position is mapped to the world coordinate system through coordinate conversion, and the target semantic annotation is realized in the map. The experimental results show that this method not only improves the accuracy of map construction and enriches the semantic information of the map, but also has high reliability of target location, which can meet the actual needs, make up for the shortcomings of LiDAR in the environment with scarce features, and improve the scene understanding and classification ability of the system.