Implementation of Automatic Focusing Algorithms for a Computer Vision System with Camera Control

Abstract

The POPEYE system is a grey level computer vision system developed at CMU for research and development. The system provides a convenient environment for research by coupling a powerful microproccssor with a large base of support software. The particulars of the system's hardware configuration and software support are given after an explanation of the desires which motivated its fabrication. In addition to providing general computation and display capabilities, the POPEYE system provides open loop manual or software control over the camera parameters of pan, tilt, focus and zoom. This offers many adwntagcs over fixed arrangements, such as the ability to investigate focusing and elementary tracking algorithms. Automatic focusing has been iniplemented before and has been found to be a particularly useful camera accommodation. In these cases, however, the approach was to implement a single algorithm. This work describes the implementation of several standard automatic focusing algorithms on the POPEYE system and provides experimental evaluation and comparison. This leaves the POPEYE system with a valuable enhancement and provides a starting point for the implementation of a production focusing system. There are many possible uses for such a system, including robotic assembly and inspection tasks. One of the applications to which the POPEYE system has been applied is the development of industrial inspection algorithms for the Westinghouse-CMU Factory of the Future Project. Part of this project involved thc inspection of fluorescent lamp mount assemblies. Algorithms for the automated inspection of the assemblies are described which represent the solutions to difficult inspection problems currently beyond the capabilities of commercial vision systems. Suggcstions for the implementation of a production focusing system are given along with suggestions for possible hardware improvements to the POPEYE system.