Coarse-coded higher-order neural networks for PSRI object recognition

Abstract

The authors describe a coarse coding technique and present simulation results illustrating its usefulness and its limitations. Simulations show that a third-order neural network can be trained to distinguish between two objects in a 4096x4096 pixel input field independent of transformations in translation, in-plane rotation, and scale in less than ten passes through the training set. Furthermore, the authors empirically determine the limits of the coarse coding technique in the position, scale, and rotation invariant (PSRI) object recognition domain.

Target Recognition Based on Directional Filter Banks and Higher-Order Neural Networks

“…To help reduce the input dimension, the coarse coding [15] method was proposed, which reported good results for binarized inputs. Unfortunately, binary thresholding cannot be applied to SAR or IR imagery effectively without significant loss of information.…”

Translation, rotation, and scale-invariant object recognition

“…High-order networks (HON's) have been utilized recently for invariant recognition [7], [19]. In this type of model, one has to encode the properties of invariance in the values of the synaptic weights.…”

“…These can be classified as: optical techniques [6], [17], boundary-based analysis via Fourier descriptors [9], [12], neural-networks models [1], [5], [7], [18], [19], invariant moments [2], [11], [13], and genetic algorithms [14].…”

“…This is done easily via the discrete Fourier-transform properties but only on uniform contours of the shape. Madaline structures for translation-invariant recognition [1], the self-organized neocognitron [5], and high-order neural networks [7], [18], [19], are examples of IOR neural-based methods. The self-organized neocognitron is a further extension of the cognitron originally proposed by Fukushima in 1975 [5].…”

Lizard Learning Algorithm for Invariant Pattern Recognition

“…By restricting to the contour cells (active pixels) and similarities among the triangles are used to reduce the number of triangles and weight classes during the process of training. The following tabular form gives the complexity problem of HONN with image size [10,11] . Number of weight classes is equal to the number of triangles which is equal to …”