BULLDOZER, A FREE OPEN SOURCE SCALABLE SOFTWARE FOR DTM EXTRACTION

Abstract

The paper introduces a software called Bulldozer, designed to extract a Digital Terrain Model (DTM) from a Digital Surface Model (DSM) obtained from various sensors. The software is based on a modified version of the multi-scale Drap Cloth principle to process noisy DSMs of any size, employing a tiling strategy and a stability margin to ensure consistent results. A parameter called max object size is introduced to differentiate objects from the ground during the drap cloth process. Gravity steps and ground distance sampling resolution are adjusted based on the input DSM. No-data and noisy values present in the DSM are detected and converted into no-data values to improve the quality of the Cloth simulation. The paper describes a memory-aware parallel execution strategy using both the multiprocessing and the shared memory Python modules. A benchmark dataset has been created to analyze the results and compare them with alternative approaches and reference datasets. Bulldozer offers an extensive Python API. It is open-source and available on PyPi and GitHub. Additionally, a QGIS plugin has been developed.