Application of Artificial Neural Networks for the Recognition of Car Number Plates
The problem of recognition of car number plates by artifitial neural networks is studied. The recognition system being created and consisting of personal computer with WEB camera and special software is presented. The composition of the recognition software module (preprocessing, extraction of the image of car number plate, image segmentation, recognition of individual symbols blocks) is briefly discussed. By the presentation of structures and training rules an analysis of Kohonen Maps and Multilayer Perceptrons for the symbol recognition is done. Results of experiments from the recognition of numerals from car number plates are provided. The estimation of the size of networks, parameters involved in their training rules is done. Ill. 4, bibl. 7 (in Lithuanian; summaries in Lithuanian, English, Russian).