College of Science and Engineering
Department or Program Affiliation
Computer Science Department
There are numerous solutions to simple object recognition problems when the machine is operating under strict environmental conditions (such as lighting). Object recognition in real-world environments poses greater difficulty however. Ideally mobile robots will function in real-world environments without the aid of fiduciary identifiers. More robust methods are therefore needed to perform object recognition reliably. A combined approach of multiple techniques improves recognition results. Active vision and peripheral-foveal vision—systems that are designed to improve the information gathered for the purposes of object recognition—are examined. In addition to active vision and peripheral-foveal vision, five object recognition methods that either make use of some form of active vision or could leverage active vision and/or peripheral-foveal vision systems are also investigated: affine-invariant image patches, perceptual organization, 3D morphable models (3DMMs), active viewpoint, and adaptive color segmentation. The current state-of-the-art in these areas of vision research and observations on areas of future research are presented. Examples of state-of-theart methods employed in other vision applications that have not been used for object recognition are also mentioned. Lastly, the future direction of the research field is hypothesized.
Subject – LCSH
Computer vision, Pattern recognition systems, Image processing--Digital techniques
Western Washington University
Copying of this document in whole or in part is allowable only for scholarly purposes. It is understood, however, that any copying or publication of this document for commercial purposes, or for financial gain, shall not be allowed without the author’s written permission.
Gerard, Rusty, "Combined object recognition approaches for mobile robotics" (2008). Computer Science Graduate Student Publications. Paper 6.