Rotation-invariant neural pattern recognition system estimating a rotation angle

Abstract

A rotation-invariant neural pattern recognition system, which can recognize a rotated pattern and estimate its rotation angle, is considered. It is well-known that humans sometimes recognize a rotated form by means of mental rotation. The occurrence of mental rotation can be explained in terms of the theory of information types. Therefore, we first examine the applicability of the theory to a rotation-invariant neural pattern recognition system. Next, we present a rotation-invariant neural network which can estimate a rotation angle. The neural network consists of a preprocessing network to detect the edge features of input patterns and a trainable multilayered network. Furthermore, a rotation-invariant neural pattern recognition system which includes the rotation-invariant neural network is proposed. This system is constructed on the basis of the above-mentioned theory. Finally, it is shown that, by means of computer simulations of a binary pattern and a coin recognition problem, the system is able to recognize rotated patterns and estimate their rotation angle.

Generalisation, decision making, and embodiment effects in mental rotation: A neurorobotic architecture tested with a humanoid robot

“…Most of these models tend to focus on the rotation of objects having different rotation axis and shapes to support object recognition, and have been developed using artificial neural network design (Kulkarni, Yap, & Byars, 1990;Fukumi et al, 1992;Fukumi, Omatu, & Nishikawa, 1997;Rowley, Baluja, & Kanade, 1998;Sasama et al, 2009;Inui & Ashizawa, 2011). For example, Sasama et al (2009) proposed an error back-propagation neural network model that takes two images as input and produces one binary vector as an output that encodes the angular disparity between the two input images together with a response answer (match/mismatch).…”

A rotation-invariant neocognitron

“…Moreover, it is expected that the new model will be able to recognize deformed-and-rotated patterns, which are not recognized correctly in Ref. 5, since the model inherits all functions of the standard neocognitron.…”

Reconhecimento De Padr�es Invariante A Rota��o Utilizando Uma Rede Neural Morfol�gica N�o Supervisionada

“…Dentre as técnicas mais utilizadas para detecção de imagens invariantes a rotação encontram-se as expansões circulares harmônicas [2], as transformadas Fourier-Mellin [3] e Radon [4], momentos geométricos [5] e abordagens baseadas em Redes Neurais Artificiais (RNAs) [6][7][8]. Tais técnicas geralmente requerem que o padrão a ser detectado seja segmentado previamente, de maneira a ser separado do contexto.…”

“…

Abstract [6][7][8]. Tais técnicas geralmente requerem que o padrão a ser detectado seja segmentado previamente, de maneira a ser separado do contexto.

O modelo de detecção invariante a rotação apresentado neste trabalho baseia-se no conceito de rotação e replicação de pesos de neurônios em diversos ângulos.

…”