Real-time Scene Reconstruction and Human Detection Fusing 3D LiDAR and Infrared for Rescue Robots

Human target detection is an essential task for rescue robots. Since LiDAR scans are sparse and camera image is sensitive to lighting condition for detection, this paper presents a method fusing LiDAR and Infrared Camera sensors to enhance all-clock detecting performance. At first, a fundamental Lidar odometer method is employed to reconstruct robot’s surrounding environments. Then a plane-to-plane joint calibration strategy is designed for the LiDAR-infrared fusion, and an adaptive threshold filtering method is proposed for infrared segmentation. In addition, a processing framework combined a plurality of point clouds segmentation strategies is employed after LiDAR-camera data fusion, and the human targets in single and sparse LiDAR scan are successfully detected. Experimental results demonstrate the effectiveness of the proposed algorithm.