A new point cloud processing method unveiled hidden coastal boulders from deep vegetation

Abstract

Huge coastal boulders are useful to reconstruct the size of past extreme waves such as those associated with tsunamis and storms using inverse-type or forward-type boulder transport models. These models fundamentally require the precise shape of boulders. Traditionally, they have often been assumed to be rectangular or ellipsoidal with three axes measured in the field. However, if the boulder’s shape is complex, this method is unable to represent the actual shape accurately. Therefore, it prevents estimation of the tsunami or storm size reasonably using models. For this reason, boulders have recently been surveyed using 3D scanning techniques such as LiDAR. However, coastal boulders now on land in tropical and subtropical areas such as Japan and Tonga are often covered by deep vegetation, which makes 3D surveys difficult. This report presents new methods to ascertain boulder shapes when they are obscured by vegetation. First, using UAV-type and mobile-type LiDAR, we scanned well-known tsunami boulders in southwestern Japan that had been covered with deep vegetation. Then, we developed a new method to extract only boulders and filter out vegetation from a point cloud. Thereby, we created 3D models of the boulders. We improved the boulder transport model further to assume the 3D boulder model accurately. In addition to coastal boulders, this filtering method is expected to be useful for unveiling any object, such as an archaeological structure, that is hidden in deep vegetation.