An Architectural Framework for Object Recognition and Selective Attention
Researchers
Project Description
Visual search and recognition are often hindered by irrelevant information in the form of background clutters and noise. The ability for humans to selectively focus on a portion of the input scene that is of interest appears to be particularly useful when handling such difficulties. In this project we aim to develop a novel framework for a biologically plausible model of visual perception with attentional mechanisms. Specifically, we propose an automatic 2D shape based object recognition system suitable for both static and moving targets, capable of performing invariant transformations under a variety of conditions. The model is able to detect, locate, recognise any familiar object, may it be rotated, shifted in position, distorted in shape, or embedded in a visually cluttered environment automatically.
Support: University of Adelaide.
Publications
- E. Chong, C.-C. Lim and P. Lozo. Modelling of a Neural Motion Detection Filter for Attentional Modulation. In International Workshop on Image Analysis and Information Fusion, IAIF'97, Adelaide, November, 1997, pp311-321.
- E. Chong, C.-C. Lim, N. Atsikbasis and P. Lozo. Design of a 2-D Neural motion Detection Filter. In Proceedings of the IEEE Region 10 Annual Conference, TENCON'97, Brisbane, December, 1997.
- E. Chong, C.-C. Lim and P. Lozo. Neural Model of Visual Selective Attention for Automatic Translation Invariant Object Recognition in Cluttered Images. In Proceedings of the Third International Conference on Knowledge-Based Intelligent Information Engineering Systems, KES'99, Adelaide, August, 1999.
- E. Chong, and C.-C. Lim. Elementary motion detection with selective attention. In Proceedings of the Third International Conference on Knowledge-Based Intelligent Information Engineering Systems, KES'99, Adelaide, August, 1999.
Research Links