Image classification for the quality control of watches

Article type: Research Article

Authors: Moreira, M. | Fiesler, E.^; | Pante, G.

Affiliations: IDIAP, C.P. 592, CH-1920 Martigny, Switzerland | GPIL, rue du Grand Verger 6, CH-1920 Martigny, Switzerland

Note: [] Present address: Advanced Signal & Image Processing Laboratory, Intelligent Optical System Inc., 2520 West 237th Street, Torrance, CA 90505-5217, USA.

Abstract: A method is presented for the automatic time detection of watches, where the hands are classified by a neural network. In order to reduce the overall cost of data collection, strict limits were imposed on the data collection time. This constraint severely limits the available amount of images, and poses the challenge to solve the hand recognition problem with a minimum amount of training and test data. Two neural network approaches are presented, together with their performance results which show an excellent final recognition rate.

Journal: Journal of Intelligent and Fuzzy Systems, vol. 7, no. 2, pp. 151-158, 1999