Wavelets transform in artificial vision inspection of threading

We present an application of multiresolution analysis with orthonormal wavelets of 1D signal to quality control by artificial vision. The purpose of the control is to check the thread of a polyethylene bottle stopper. Using a section picture of the stopper provided by a linear CCD camera, we calculate the wavelet coefficients of the first three levels of resolution. The energy densities of these coefficients calculated on a given area, provide three discriminant parameters which permit to distinguish correctly between the two classes (defectless, defective) according to a classifying method which can run without supervision after a period of training