Title :
A simple invariant neural network for 2-D image recognition
Author_Institution :
Tech. Res. Center, Cairo, Egypt
Abstract :
A simple invariant neural network has been proposed. The network has invariance against scale and, rotation changes, in addition to the inherent shift of starting point on the image contour. This invariance comes from the new use of the MT-transform as a feature vector in a pre-processing stage. Thus, a complete invariance has been achieved, without any complexity in the network. Testing of the network, shows about a 100% recognition rate
Keywords :
backpropagation; edge detection; feature extraction; image recognition; multilayer perceptrons; transforms; 2D image recognition; MT-transform; backpropagation algorithm; feature extraction; feature vector; image contour; image recognition rate; invariant neural network; multilayer preceptron; network testing; preprocessing stage; rotation changes invariance; scale changes invariance; Algorithm design and analysis; Biological neural networks; Cities and towns; Classification tree analysis; Data mining; Image recognition; Neural networks; Neurons; Object recognition; Testing;
Conference_Titel :
Radio Science Conference, 1996. NRSC '96., Thirteenth National
Conference_Location :
Cairo
Print_ISBN :
0-7803-3656-9
DOI :
10.1109/NRSC.1996.551116
