Visual feature extraction via PCA-based parameterization of wavelet density functions

This paper describes the use of a novel method of feature extraction for visual sensor processing. An application to crack detection by a sewer maintenance robot equipped with an infra-red camera is described. A spacefrequency distribution ’signature’ is generated via a three step process involving; space-frequency decomposition, density function estimation, and parameter extraction. These steps are achieved, respectively via; the wavelet packet transform, empirical cumulative density functions, and principle components analysis. The ’wavelet signature’ method is shown to be superior to conventional methods as a feature extractor for a logistic regression model in a crack discrim-