1. Science
  2. Publications
  3. Information Processing Systems
  4. 2(148)'2017
  5. LETTER IMAGES RECOGNITION BASED ON PROJECTIONS AND MULTILAYER PERCEPTRON

LETTER IMAGES RECOGNITION BASED ON PROJECTIONS AND MULTILAYER PERCEPTRON

O.V. Gorokhovatskyi, O.V. Teslenko
Annotations languages:

Method of image recognition using the projections of an image as a method of structural information preserve with the further training of multilayer perceptron as an instrument of generalization and recognition was proposed. Features of the combination of projection with subsequent perceptron learning process were investigated, two approaches to use the projection feature were proposed. Experimental modeling shown the effectiveness of proposed method to solve the practical problems of letter images recognition retrieved from scanned text
Keywords: image recognition, multilayer perceptron, learning, projection, Radon transform, features, network structure.