| Automated vision based inspection systems often require a segmentation procedure to separate objects from background regions. Adaptive segmentation reduces the effects of a noisy environment due to uneven lighting conditions, eliminates unnecessary information from an image, and helps identify objects. The segmentation algorithm developed was based on a Ph.D. thesis entitled "Rapid Adaptive Segmentation of Images Using Successive Cost Processing" by Dr. Arturo Rodriguez. This approach to segmentation breaks the image into small regions and tracks slowly varying lighting changes by estimating the local statistics of each region.; The entire inspection project involved integrating a laser ranging system developed by General Dynamics with a Unix based Sun Sparc station, a PC controlled robot, and two CCD cameras to create an automated inspection/finishing system. Further improvements in performance can be obtained by implementing the digital image processing software in a parallel processing environment. |
