CAREER: Bridging the Gap in Point Cloud Analysis: Task-Oriented Visualization for Enhanced Human-AI Collaboration

Clemson University, Clemson SC

Investigators

Abstract

As cameras take 2D to capture images, spatial cameras use point clouds to look at 3D images. Point clouds are an increasingly important data format for representing 3D scenes, and are used in a number of contexts from self-driving cars to satellite image analysis. However, methods for helping people make sense of point cloud data lag behind, in terms of both computational methods for analyzing these data and in terms of information visualization techniques designed for point cloud data and related tasks. This project will investigate effective, scalable visualization techniques to support the development of systems where people and artificial intelligence techniques collaborate in point cloud data analysis tasks. The chosen tasks will be grounded in real-world problems posed by forest monitoring applications and search and rescue using autonomous vehicles. The research objectives of this project will be complemented by educational activities targeting middle and high school students and promoting the engagement of underrepresented undergraduate students in research. The project's central innovation is to move beyond approaches to point cloud visualization, such as scatterplot representations, that focus on creating realistic three-dimensional reconstructions. Instead, the goal is to create effective, task-oriented visualizations that prioritize user objectives in point cloud analysis and thus guide both the representations and the algorithms used to analyze point cloud data. The work is organized around three key research thrusts: (1) producing rankings and guidelines for the design of effective task-oriented point cloud visualizations, (2) developing uncertainty visualization techniques to support the visual interpretation of artificial intelligence models operating on point clouds, and (3) deploying effective visualization methods in real-world scenarios for collaborative human-machine point cloud analysis. These scenarios will involve extensive outreach activities focused on promoting task-based point cloud visualization among researchers and domain experts working with point cloud data, as well as a wider package of outreach and educational activities. This project is jointly funded by Secure and Trustworthy Cyberspace (SaTC) and the Established Program to Stimulate Competitive Research (EPSCoR). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.