DETECTION OF FABRICATION HOLES WITH SIGNATURE ANALYSIS

Abstract

The article deals with the problem of determining the shape and parameters of holes (shape, location (coordinates), size, etc.) with the method of image signature analysis. This subject is particularly relevant to the mechanical engineering, where it is necessary to control the quality of fabrication holes in real-time. The feedforward artificial neural network is used as the classifier. The application of the said approach allows of obtaining an acceptable quality of the image classification (holes) on the basis of the training samples and solving the problem of detecting holes under noise conditions. The design of artificial neural network, training and testing were carried out in Matlab. The filtering of images and obtained signatures was conducted to achieve the noise immunity of the classifier. As a result of computer experiment in the Matlab environment the given system showed the possibility of identification of fabricaion holes with acceptable accuracy.