Adaptive autonomous positioning of a robot vision system: Application to quality control on production lines

• An adaptive robot vision system detects obstacles along the camera line of sight.
• The robot vision system automatically repositions the camera to re-establish a correct target imaging.
• The method employs an FFT to determine obstacle rotation and to compute the new position of the robot.
• Perspective distortions are corrected using the knowledge of 3D working space of the robot.
• The method is demonstrated for on-line quality control along appliances production lines.

This paper presents an adaptive strategy for automatic camera placement in a 3-dimensional space during a robotized vision based quality control. The approach proposed improves the overall efficiency of the system, allowing a correct image acquisition, even in cases where an obstacle along the camera line-of-sight hides the object to be inspected, not making it possible to perform the inspection by template matching. In particular, a strategy for automatically avoiding a possible obstacle is defined.The paper focuses on a class of obstacles whose position on the scene can be detected by applying a Fast Fourier Transform (FFT) algorithm to the image; these kind of obstacles can rotate randomly. The phase information from the FFT is then used to define the movement necessary to clear the camera line-of-sight and acquire an image of the target object. Such image will result distorted by perspective. However, both the image acquired before and the one acquired after the camera repositioning, together with the knowledge of the relative displacement, are used to partially reconstruct the three-dimensional structure of the object inspected. Corresponding points in the two images are located and perspective distortions resulting from the movement of the camera are corrected. The correction improves the object recognition task since it makes it possible to use a single template on images acquired from very different positions.The whole method has been tested for on-line quality control of washing machine parts, which represents a typical case of quality control on a production line.