3D Point Cloud Reconstruction Based on Stereo Camera and RTK-GPS with High Accuracy under Sequential Correction

In the field of three-dimensional (3D) point cloud reconstruction, a commonly used algorithm is called Real-Time-Appearance-Based-Mapping. Although this algorithm performs well in outputting a point cloud with high accuracy in real time, its processing efficiency is expected to decline when Jetson Nano is installed on drones, as additional memory is required to calculate loop closures. To increase the processing speed and accuracy of 3D point cloud reconstruction based on Jetson Nano, we propose a new algorithm for drone flight, that does not rely on overlap. The proposed method combines the point cloud and real-time kinematic Global Positioning System, and its effectiveness is demonstrated by experiments in a real environment.