This paper proposes an imaging sensor and LiDAR-based SLAM calculation framework which also considers the loop closure detection. First, the system adopts deep learning object detection method and semantic segmentation approach to detect and eliminate the moving objects in dynamic scenes. Second, the system utilizes a combination of 3D LiDAR data and 2D imaging sensor data to estimate the dense point clouds in scenes. Third, the system performs the mapping calculation for the entire region and considers the similarity calculation criteria to perform the loop closure detection under various constraint conditions. The preliminary experimental results have indicated the correctness and effectiveness of our method, and the positioning accuracy can be improved, while the mapping time can be shortened to some extents.