3D image processing in online inspection

In this paper a method for 3D online inspection system is presented. The system utilizes 3D information, obtained from a vision system, for qualifying objects running on a conveyor belt. The accuracy of the system is +/- 0,5 % (in height) and further studies will be made to even increase the systems accuracy. The system is based on a frame grabber, an area scan camera, a laser light source and a PC. The laser light source projects a structured line of light in the surface of the object to be measured. A CCD matrix camera will then continuously grab images from the surface and the frame grabber then preprocesses the output image. Each output image results in one continuous line profile. Number of profile lines is only limited by the main memory of the PC, which gets the 'raw' profile line as input. The software in the PC will then transform the 'raw' profile line to global coordinates, which can be used in further canalization. Analyzation is done in several phases. The object's height, shape and the irregularities are measured directly from the 3D information. Several different methods are used to make the decision whether the object is acceptable or not.