23 June 2000 Detector-based visual identification of known objects
Author Affiliations +
Regarding input image as a function defined on sensor frame, the structural identification of a known object is considered as extraction of domains of sensor frame responsible for all object details. The low- level processing tools are formed by a set of detectors. The detector is introduced as a generalization of any feature detection destined to recognize a certain class of patterns. As an operation of intermediate- level processing, the detector filling is introduced. It cuts a domain of sensor frame in response to input parameters of the detector. Then, a new object representation (TSG-model) is introduced. It consists of three components: the one responsible for the object structure, the one for construction of a detail domain by domains of its sub-details, and the one reducing identification of a detail if other details are already identified. The image processing for structural identification of an object is reduced to a typical AI search. It executes operations of detector filling, interacting with the TSG-model of the object. The paper outlines a new technology that can be applied in many areas. The presented approach is flexible with respect to the kind of input sensor, say, usual gray-scale camera , stereo image system, range image system, etc.
© (2000) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Georgii Khachaturov, Georgii Khachaturov, } "Detector-based visual identification of known objects", Proc. SPIE 4023, Enhanced and Synthetic Vision 2000, (23 June 2000); doi: 10.1117/12.389350; https://doi.org/10.1117/12.389350

Back to Top