Classification of mechanical parts using an optical-digital system and the Jacobi-Fourier moments

In this work, we consider the use of circular moments for invariant classification of images which have been blurred by motion. The test images of the objects under consideration have been acquired when they are vibrating. For this task an experimental setup is implemented to generate vibrations. A comparative analysis using several circular moment sets is presented; the studied sets are Zernike, Pseudo-Jacobi-Fourier, Orthogonal Mellin-Fourier, shifted Chebyshev-Fourier, and radial harmonic Fourier. The classification method is tested using images of mechanical parts which have intrinsically little differences between them, as screws with millimetric or standard threads. Experimental results and the optical setup used are presented.