Large-Scale Point Cloud Based Volume Analysis Acceleration Based on UAV Images

Abstract

Small-sized drones, either having classical fixed-wing design or the so rapidly widespread rotorcraft design, are now revolutionizing the complete work process of aerial surveys. In the paper, we will discuss the possibilities of the quality characteristics of large-scale point cloud processing, already introduced in our previous work, and primarily the reduction of the execution time of the large-point cloud processing presented in our previous work. Based on the work process developed by us, we survey the work area, typically a quarry, using UAVs, and then we create a high resolution 3D point cloud out of the pictures. Then, the time series photos are fitted to each other using the ICP algorithm and we conduct the volume analysis. In the present case, the ICP algorithm and volume calculation were very time-consuming operations, as every 3D model was made up of ~20 million points. The two leading graphics processing unit producers of the market have made the resources of the hardware available for general-purpose calculations. In this way, we can see good opportunity in the implementation of these time-consuming calculations into the target hardware.