An Automated Process to Filter UAS-Based Point Clouds

Abstract

Digital Terrain Models (DTMs), which represent the topography of the bare Earth surface, are widely used in many geomatics applications. In parallel to the emergence of sophisticated Unmanned Aerial Systems (UASs) in recent years, they are produced from point clouds generated through aerial images taken from digital imaging systems mounted on UASs. The first and most important step of DTM production is to remove the points of the above-ground objects such as trees, buildings, bridges, etc. A great variety of point cloud filtering strategies have been developed so far. However, due to the irregularities in the topography of the Earth's surface, all proposed approaches employ several user-defined parameters, which makes point cloud filtering dependent on the parameter values defined. Since complex topographies make it very hard to define some protocols to estimate the best parameter values, users usually have to try a large number of parameter values for optimal filtering performance, which is neither practical nor time-efficient. Hence, this study proposed to use the metaheuristic Whale Optimization Algorithm (WOA) to estimate the parameters of a simple morphology-based (SMRF) point cloud filtering algorithm to improve its performance, automating the filtering process. The performance of the proposed filtering methodology was compared not only against that of the standard SMRF algorithm but also against those of popular filtering algorithms Cloth Simulation Filtering (CSF) and Progressive TIN Densification (PTIN). The results showed that the proposed filtering methodology outperformed the PTIN and standard SMRF algorithms and presented a comparable performance with the CSF algorithm, which is one of the most robust point cloud filtering algorithms proposed to date. It can also be concluded that metaheuristic optimization algorithms can be used to automate the point cloud filtering process, minimizing the filtering errors caused by user intervention.