Generating Models of Building Interiors from 3D Data

Tuesday, May 2, 2:30 PM - 3:00 PM
Summit Track: 
Technical Insights II

We present an automatic approach for the task of reconstructing a 2D floor plan from unstructured point clouds of building interiors. Our approach emphasizes accurate and robust detection of building structural elements, and unlike previous approaches does not require prior knowledge of scanning device poses. The reconstruction task is formulated as a multi-class labeling problem that we approach using energy minimization. We use intuitive priors to define the costs for the energy minimization problem, and rely on accurate wall and opening detection algorithms to ensure robustness. We provide detailed experimental evaluation results, both qualitative and quantitative, against state of the art methods and labeled ground truth data.


Axel Wendt

Head of Robotics Research, Robert Bosch LLC

Dr. Axel Wendt is currently responsible for the Robotics Technologies Research Group at Bosch. His research topics includes collaborative robotic technologies and perception. Prior to his current position, Dr. Wendt was a Senior Research Engineer, Project Manager as well as an Innovation Manager in the field of Advanced Production Automation and Computer Vision Systems. He is an inventor, who is listed on more than 10 patents and patent applications and an author of various publications covering computer vision, photogrammetry and driver assistance systems. In 2007 he received his Ph.D. from Leibniz University in Hannover, Germany for his work on object-based simultaneous multi-sensor orientation. Dr. Wendt belongs to the German Association of Pattern Recognition (DAGM) and is board member of Silicon Valley Robotics.


Join us May 22-24, 2018 in Santa Clara, California.
Book your discounted hotel room now - offer expires April 30!