The authors explore alternatives that reduce the number of network weights while maintaining geometric invariant properties for recognizing patterns in real-time processing applications. This study is limited to translation and rotation invariance. The primary interest is in examining the properties of various feature spaces for higher-order neural networks (HONNs), in correlated and uncorrelated noise, such as the effect of various types of input features, feature size and number of feature pixels, and effect of scene size. The robustness of HONN training is considered in terms of target detectability. The experimental setup consists of a 15*20 pixel scene possibly containing a 3*10 target. Each trial used 500 training scenes plus 500 testing scenes. Results indicate that HONNs yield similar geometric invariant target recognition properties to classical template matching. However, the HONNs require an order of magnitude less computer processing time compared with template matching. Results also indicate that HONNs could be considered for real-time target recognition applications.>

An approach is presented to the problem of target detection under conditions of low signal-to-noise ratio. A feature space is constructed from local correlation properties of a model one-dimensional image. A multilayer back-propagation network is trained using these features. A target is embedded in noise, either uncorrelated or pairwise correlated. The error rate for target detection is higher for correlated than for uncorrelated noise when using only a pairwise-correlated feature space. Adding triplet correlated features has no effect on the error rate for the case of uncorrelated noise, but for the pairwise correlated noise the additional features reduce the error rate to that of the uncorrelated noise