A Fast Modeling Method of 3D Mapping Based on Point Cloud Data

With the development of smart cars, map accuracy requirements continue to increase. Traditional maps can carry limited information from artificial site mapping, which is time-consuming and labor-intensive. Therefore, it is important to create 3D reconstructions quickly and efficiently, and to add more portable information. We propose a 3D map based on real-time reconstruction. An advantage of point cloud data is that it is simple to collect, allowing for real-time collection and reconstruction. Here, we analyze modeling point cloud data through data preprocessing, matching of point cloud data, point cloud integration and segmentation operation, the curve and curved surface reconstruction, and establishing a good model for surface smoothing. The experimental environment adopts KinectFusion accelerated by GPU, which can produce dense 3D reconstructions in real time and display 3D maps quickly and conveniently with Unity. For this time-consuming problem, we use octrees instead of KD trees to optimize map reconstruction.