Unfallverhuetung und Funktionskontrolle bei Industrierobotern durch automatische Bildverarbeitung.

The goal of this research is the development of an automatic video surveillance system for industrial workroom environments. The objective of this system is the prevention of collisions between an industrial robot and human workers as well as the monitoring of the robot's motion. Contrary to conventional safety measures that do not process any motion information, and therefore require restrictive regulations for the operation of robots, the described video processor detects and monitors moving objects in the workroom from digital images taken by a stationary camera. The processor consists of two parallel sub-units, each specialized for a certain class of moving objects. The first is based on the statistical analysis of original and difference pictures, and is applied to objects with unknown features. In the second sub-processor, scene-invariant features of the moving machinery are extracted and tracked. Both sub-processors can cope with varying conditions of illumination. Motion models derived from the history of observation are employed for the prediction of the trajectories of the moving objects, and thus permit collision forecasting and limited tracking capability in case of object occlusion. (IITB)