CAVIAR: Research Objectives

This central research question of CAVIAR (Can rich local image descriptions from foveal and other image sensors, selected by a hierarchal visual attention process and guided and processed using task, scene, function and object contextual knowledge improve image-based recognition processes?) leads to several specific research objectives:

1. To investigate if using log-polar instead of Cartesian image data gives image features with more effective descriptiveness.
2. To investigate how contextual knowledge and priming can focus collection of feature data through a selective attention mechanism, in terms of both the quality and the order of feature extraction.
3. To investigate if feature grouping mechanisms can improve the recognition rate, and how contextual knowledge can improve the grouping process.
4. To investigate how contextual knowledge can be used to focus invocation of object, situation and context models from a database.
5. To investigate alternative appearance-based representations, similarity metrics and matching strategies for objects, situations and contexts in order to increase recognition accuracy.
6. To investigate alternative knowledge-based control strategies in order to increase recognition speeds and accuracy.
7. To implement the results of these investigations and integrate the components into a complete contextual recognition system.
8. To test the object and situation recognition system on the tasks of city centre surveillance and customer behaviour analysis.

Date of last update: August 30, 2002
Back to CAVIAR home page.