3D Sensing & Recognition for Transparent Object (ARD/311)

Transparent objects present unique challenges for traditional sensing and recognition systems due to their lack of distinct features and the refractive properties. This characteristic poses difficulties across multiple industries, including scanning for Building Information Modeling (BIM), perception and navigation of service robots, autonomous mobile robots (AMR) and human-robot collaboration, assembly automation, as well as picking and placing automation. Currently, there is no existing three-dimensional sensing and recognition solution specifically designed for transparent objects. The project aims to develop and implement three kinds of 3D sensing and recognition technologies for transparent objects, including 3D scanning, 6D pose estimation, and 3D surface reconstruction. This project mainly utilizes consumer-level sensing devices, particularly cameras, to accomplish the tasks of 3D sensing and recognition, distinguishing it from conventional 3D sensing & recognition methods that rely on specialized equipment. The outcome of this project will overcome the limitations of existing approaches and enable cost-effective and accessible 3D sensing and recognition solutions for transparent objects, empowering various industries to benefit from enhanced perception capabilities and drive advancements in their respective fields.