The basic function of robot vision system in robots is to identify an object and determine its location (position and orientation). The vision system must be capable of handling multiple views to deal with the multiple stable states. For this purpose, the system has to be fast and work in parallel with the robot system.
Further the system must be able to work in an industrial environment including factory illumination and be insensitive to normal light variation. A gray scale system is required so that numerous different shades of gray can be assigned to each point of the image and the vision system’s performance is not affected by the object’s form, colour or surface texture.
The vision hardware consists of a camera (usually charge-coupled device CCD camera) placed above the scene, and a TV monitor, an image pre-processor and a conventional micro-computer housed in a control cabinet.
CCD camera has no geometric distortion, a high signal-to-noise ratio is obtained because of absence of deflection coil, thus insensitive to disturbances from electro-magnetic fields. The video signal from the camera enters the pre- processor, is digitised and loaded into an image memory.
ADVERTISEMENTS:
The image is then filtered and translated into a gradient image. In gradient processing, the algorithm looks for differences in the level of the pixel and is thus insensitive to light variation because the gradient level remains same irrespective of lighting across the scene. The gradients are then classified in order to create contours. The whole process is concluded with a structuring of the resulting contour image.
The pre-processor thus produces a list of all the features in the image which is then fed to a general purpose micro-computer where the object (s) are identified and the position and orientation determined. The micro-computer is also used to supervise the pre-processor, to handle the robot vision system communication, to perform error supervision and to handle the self-diagnostic routines.
Programming of an object is performed by placing the object beneath the camera and processing the image. The vision system will then display the object’s contour (inner and outer) on the TV monitor. The operator then points out the object’s significant features (characteristic forms).
A cursor, which is moved by using the joystick on the programming unit is utilised for this purpose. The acquired information is stored in the computer memory and is available for use whenever the same object is to be identified.
ADVERTISEMENTS:
The object is then moved a couple of times in the scene, and image is processed in between each move. Thus the system collects statistical data about the object and forms decision areas. The statistical data aids in the certainty factor with respect to part identification. In this way number of different objects, each having different views of all possible stable positions, are programmed and stored on the floppy disk unit.
Robotic vision systems in robots are now available which allow robots to pick randomly oriented parts out of a bin, orient them for further processing, thus enabling handling of delicate parts that can’t be fed using conventional means such as vibrating bowls or belt feeders.