Faster Multibeam Sonar Data Cleaning: Evaluation of Editing 3D Point Clouds using Immersive VR

Abstract

Remote sensing technologies routinely generate point cloud datasets with billions of points. While automatic data cleaning algorithms exist, safety-critical applications (such as waterway surveys) still require that data be processed and verified by a human. This presents a significant bottleneck in the pipeline from surveys into navigational maps. The recent proliferation of low-cost, high-quality virtual reality systems presents an opportunity to explore how these technologies might be integrated into the point cloud data processing pipeline. Prior research has shown that stereoscopic viewing, head-tracked perspective, and bimanual interactions can lead to faster 3D task completion times and lower errors compared to traditional monoscopic, mouse-and-keyboard desktop systems. In this paper, we present a human factors study that compares 3D point cloud editing performance between a traditional interface and type types of immersive virtual reality interfaces. Our results showed that for complex datasets, the immersive interfaces generally led to faster task completion times than when using the desktop interface. Participants also reported a strong subjective preference for the immersive interface.