Accession Number : ADA285202

Title :   Genetic Contour Matching: A Feasibility Study.

Descriptive Note : Final rept.,

Corporate Author : INSTITUTE FOR PERCEPTION RVO-TNO SOESTERBERG (NETHERLANDS)

Personal Author(s) : Toet, A.

Report Date : 01 FEB 1994

Pagination or Media Count : 22

Abstract : This report presents a genetic algorithm that detects, in grey-level images, instances of a parameterized object contour model. Object detection is formulated as an optimization problem. The objective function measures the evidential support for any particular projection of the parametrized object contour model onto the input image. A genetic algorithm is used to find a set of parameters which provide a good (though not necessarily optimal) interpretation of the image in terms of the model. object classification can be performed through comparison of the parameters of the best fit with those of the prototypes in a data base. The genetic contour matching algorithm can therefore be used as a basis for a scale and pose-invariant object recognition system. The results of some preliminary tests demonstrate the feasibility of the approach. Image processing, Automatic target recognition, Pattern recognition, Edge detection, Recognition, Genetic algorithms, Target acquisition, Image processing.

Descriptors :   *CONTOURS, *GENETICS, ACQUISITION, ALGORITHMS, APPROACH, AUTOMATIC, CLASSIFICATION, COMPARISON, DATA BASES, DETECTION, EDGES, FUNCTIONS, IMAGE PROCESSING, INPUT, MATCHING, MODELS, OPTIMIZATION, PARAMETERS, PATTERN RECOGNITION, PROTOTYPES, SCALE, TARGET ACQUISITION, TARGET RECOGNITION, TARGETS, TEST AND EVALUATION, NETHERLANDS, TRANSLATIONS, DUTCH LANGUAGE.

Subject Categories : Biology

Distribution Statement : APPROVED FOR PUBLIC RELEASE