Automatisation of sparse cloud cleaning in Agisoft Metashape Professional (ver. 2.x)

The U.S. Geological Survey (USGS) has recently published a guideline for processing coastal imagery acquired by unmanned aerial vehicles (UAVs) based on the widely used Agisoft Metashape Professional software (USGS Open-File Report 2021-1039; Over et al. 2021). The guideline aims to improve the quality of photogrammetric reconstructions by iteratively removing low-quality tie points based on different cleaning parameters. However, the improvement procedure is iteratively performed and requires permanent attendance of the operator. Furthermore, the different cleaning steps are executed on a trial-and-error basis, adding up to the overall attentiveness required.

To minimise the time expenditure necessary for conducting the cleaning procedure and to provide a frame for the reproducibility of photogrammetric product derivations, we have compiled a python script to automate the tie point cloud optimisation as detailed in the USGS report. The graphical user interface of the script allows non-expert users to adjust important cleaning parameters, such as maximum reconstruction uncertainty, minimum projection accuracy and/or maximum reprojection error thresholds, and number of iterations to be performed. Furthermore, main tie point cloud quality measures can be directly assessed. We will demonstrate that the time required to clean tie point clouds (using a computer equipped with a 3.60 GHz processor, 64 GB Ram, and NVIDIA Quadro M4000) can be significantly reduced, such that cleaning of a (unprocessed) ~8 million tie point cloud is achieved unattended in about 30 minutes (default cleaning threshold values used), with ~7 million tie points being automatically removed while significantly increasing point cloud quality measures.