Automatic classification of point clouds obtained with different airborne sensors in UAV

摘要

One of the main objectives when compiling information from drones, is the generation of digital terrain models - DTM, which are what ultimately allow the creation of cartographic products reliably, efficiently, accurately, economically and quickly. In the present work, information obtained in Colombia is used in 3 types of terrain, with different varying conditions and characteristics. The data is taken with 2 different sensors, located in 2 fixed-wing drones (X8 skywalker with Sony Alpha camera a6000 24MP and UAVER Avian P with Sony RX1R II 42 MP). Information processing was performed and point clouds were obtained about the same area for executing a comparative iterative analysis, and obtaining the optimal parameters of: iteration angle, slope and distance of terrain iteration. A semiautomatic classification of point clouds was used, with 4 different treatments proposed by the authors, called classification MACROS and different DTM was generated. To analyze the behavior of the point clouds and check the accuracy, a control of dimensions in the field was made, the land was divided into 12 plots and the difference in elevation was calculated with a cloud of checkpoints, obtained manually. Finally, a completely randomized block experiment model is designed, created and tested.