Automated optical inspection of geometrical shapes

The paper presents an automated optical inspection system for geometrical feature measurements. The system can be programmed through friendly user menus for the measurement of any geometrical shape. A series of movable calipers can be set for different rectangular dimensions while radiuses can be used for circular shapes. The accuracy of these measurements is mechanically first set by choosing the appropriate magnification for the lens set mounted on the TV industrial camera. This precision is augmented fifty times by the use of a subpixel location method of the image processing package. Fast and reliable edge detection techniques are introduced which allow for one hundred percent nondestructive inspection of manufactured parts. Inference methods are used in conjunction with the edge locator to obtain robust decisions on alignment results. The system has been implemented on a SIMD parallel vision computer using an object oriented programming environment. It is successfully working on a production line.