A Rule-Based Pattern Matching System for the Recognition of Three-Dimensional Line Drawn Objects: A Foundation for Video Tracking,
Abstract
Object recognition is an important subtask in image understanding, moving object tracking, and scene analysis. When identifying an object in a scene, it is essential that the same object is recognized as being so from different view angles. Also, in cases where the object is occluded or the image is noisy, the recognition is expected to function under the existence of uncertainty. In this report, a system for object recognition, with emphasis on view angle independence, is studied. The system uses simple line drawn objects as the input image. An algorithm to extract important information from the image is developed. A rule-based pattern matching scheme is used to recognize the object in the image. In each decision made, a confidence factor is associated to indicate the system's certainty in making this decision. This study shows that the rule-based pattern matching system is a useful and flexible framework for object recognition and scene analysis. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1986
- Accession Number
- ADA170701
Entities
People
- Y. S. Fong
Organizations
- Clarkson University