Real-time optical Hough transform and morphological inspection techniques

A real-time optical Hough transform inspection processor is described and applied to the industrial case study of the inspection of cigarette packages. The inspections are performed in real-time using a color Epson liquid crystal television as the input spatial light modulator. A new selective, real-time edge-enhancement technique is used to enhance edges only at the selected orientations which will Hough transform to peaks (this maximizes the signal to noise ratio in the Hough transform output). Several slices of the Hough transform are generated in parallel with a new computer generated hologram utilizing 1D modulated error diffusion. We benchmark our Hough transform processor with the inspection of 100 sample cigarette packages. Correct classification of all the packages (with defects that we intended to inspect for) was attained. We also discuss the confidence with which we can apply our laboratory sample set results to the industrial installation of our inspection processor.; Extensions of the Hough transform are also described which generalize the Hough space to recognize shapes other than straight lines (e.g. circles, ellipses). We propose a new high speed acousto-optic architecture to implement these generalized Hough transforms. We also describe how the exact position of input lines can be extracted from Hough transform peak data. Typically, only the contour on which an input line resides is extracted from Hough data. Digital simulations illustrating both generalized Hough transformations and determination of input line position from Hough data are presented.; Basic morphological transformations and their real-time optical implementation are discussed. We describe how morphological transformations are useful in inspection and show digital simulations and optical results of an inspection application (string detection in chopped tobacco) where morphological transformations are used.