Department of Computer Science and Engineering
University of Washington

Object Recognition

Object recognition is the subfield of computer vision whose goal is to recognize objects from image data and, often, to estimate the positions and orientations of the recognized objects in the 3D world. The images to be analyzed may be 2D gray-scale or color images or 3D range data images. Applications are many and include industrial machine vision, medical image analysis, and content-based image retrieval. Object reconstruction refers to the construction of 3D object models from image or range data.

In the 3D domain, we are working on the representation, recognition, and reconstruction of objects from range data. We are developing symbolic 3D object models that can be used in several different applications, including modeling the organs of the human body and modeling man-made objects for use in augmented reality environments. Our 3D object reconstruction system begins with range/color views obtained from a 4-camera active stereo systems, cleans and registers the data, produces a 3D mesh using a space-carving technique, and can render the objects using a view-dependent texturing algorithm. We are also working on algorithms to match symbolic models to mesh data, using a knowledge-directed approach. This work is part of a collaboration with researchers in the Department of Biostructure to fully model the human body, including symbolic, spatial, and relational aspects, and involves a number of different aspects of computer science: database systems, artificial intelligence, computer graphics, and computer vision.

Research Projects

button 3D Object Reconstruction
button 3D Object Recognition from Meshes
button Object Recognition for Content-Based Retrieval
button Model-Based Industrial Object Recognition