Instant Visualization and Interaction for Large Point Clouds

Point clouds are a quintessential 3D geometry representation format, and often the first model obtained from reconstructive efforts, such as LIDAR scans. However, the sheer amount of captured detail poses a challenge for on-site quality assurance: defects or missing information may only be noticed much later after the data has been processed for efficient viewing and interaction. The project IVILPC aims for the instant, authentic, interactive, and high-quality visualization of such unprocessed, point-based data sets. To this end, we will exploit the flexibility and performance of cutting-edge GPU architecture features to formulate novel hybrid rendering approaches, taking advantage of both object- and image-order techniques. The envisioned solutions should be applicable to unstructured point clouds for instant visualization of billions of points, and leverage adaptive compression, workload scheduling and usage-driven data management to enable fluent workflows: our project tackles the GPU-accelerated interaction with point clouds, as well as common point cloud display artifacts through corrective real-time methods. IVILPC lays the foundation for interaction with massive point clouds in conventional and immersive environments. Its seminal goal is an efficient data knowledge transfer from sensor to the user, with a wide range of ICT use cases to virtual reality (VR) technology, architecture, the geospatial industry, and cultural heritage.