Fast and scalable 3D texture-minimal object detection

This work develops a method for detecting objects such as hand tools that have very few inner texture with a fast and scalable approach. 

The method can work with a RGB video stream and handles clutter well, however it can also be combined with an RGB-D camera for faster operation.

There are several aspects that make this method work one of them is the notion of edge paths, which scan the edge image in an efficient way, another is the notion of making this search tractable since without that, for a busy edge image the search explodes in a combinatorial manner. We address this isse by having fixed search paths in a tree-like representation making the method scalable and fast.

A video showing online learning and multiple-object detection is available here.

• Dima Damen, Pished Bunnun, Andrew Calway, Walterio Mayol-Cuevas, Real-time Learning and Detection of 3D Texture-less Objects: A Scalable Approach. British Machine Vision Conference (BMVC), 2012. PDF. Video. Winner best poster paper.

• Dima Damen, Andrew Gee, Andrew Calway, Walterio Mayol-Cuevas, Detecting and Localising Multiple 3D Objects: A Fast and Scalable Approach. IROS Workshop on Active Semantic Perception and Object Search in the Real World (ASP-AVS-11). September 2011. PDF.