Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Automatisches Optimieren von Objekt-Erkennungssystemen

Automatic configuration of test system for object identification - has camera input to processor performing learning of object characteristics for use in identification process.
: Wagner, T.; Bauer, N.; Plankensteiner, P.

Frontpage ()

DE 1996-19612465 A: 19960328
DE 1996-19612465 A: 19960328
DE 19612465 C2: 19981224
Patent, Electronic Publication
Fraunhofer IIS ()

The classification system has a video camera used to monitor different objects (1a,b,c) that move past the field of vision. Image data is entered into a data logger (60) coupled to a processor (61). The system operates with a learning phase in which data is received by functional block (30) coupled to the classifier (20). This provides reference data to be used by the evaluation unit (50). The objects can differentiate on the basis of the surface condition and distinguishes between intact surfaces, scratches and marks. The capacity of the configuration is continually evaluated by sampling. The configuration of the pattern recognition system is modified with the help of optimising strategies (50) e.g. linear or non linear optimising, genetic algorithms or genetic programming. USE/ADVANTAGE - Esp. for industrial test system for surface area analysis, analysis of complex real view, person identification. Automated process for high productivity.