Page 435 - Sensors and Control Systems in Manufacturing

Sensors in Flexible Manufacturing Systems
Once calibrated, the system can then locate objects in its own coordinate 389
system (pixels) and translate the position into work cell coordinates
(inches, millimeters, etc.).

8.5 Robot Guidance with Vision System
Another use of machine vision is in robot guidance—helping a robot
to handle and place parts and providing it with the visual configura-
tion of an assembly after successive tasks. This can involve a series of
identification and location tasks. The camera can be attached to a
mobile arm, making the location task seem somewhat more like nor-
mal vision. However, the camera typically is mounted on a fixed loca-
tion to reduce system complexity.
While each image can give the location of certain features with
respect to the camera, this information must be combined with infor-
mation about the current location and orientation of the camera to
give an absolute location of the object. However, the ability to move
the camera for a second look at an object allows unambiguous loca-
tion of visible features by triangulation. Recognition is a useful tool
for flexible manufacturing systems within a CIM environment. Any
of several parts may be presented to a station where a vision system
determines the type of part and its exact location. While it may be
economical and simpler to send the robot a signal giving the part
type when it arrives, the ability to detect what is actually present at
the assembly point, not just what is supposed to be there, is of real
value for totally flexible manufacturing.

8.5.1 Robot Vision Performing Inspection Tasks
Visual inspection can mean any of a wide variety of tasks, many of
which can be successfully automated. Successful inspection tasks are
those in which a small number of reliable visual cues (features) are to
be checked and a relatively simple procedure is used to make the
required evaluation from those cues.
The differences between human and robot capabilities are most
evident in this kind of task, where the requirement is not simply to
distinguish between good parts and anything else—a hard enough
task—but usually to distinguish between good parts or parts with
harmless blemishes, and bad parts. Nevertheless, when inspection
can be done by robot vision, it can be done very predictably.
Many inspection tasks are well suited to robot vision. A robot-
vision system can dependably determine the presence or absence of
particular items in an assembly (Fig. 8.5), providing accurate infor-
mation on each of them. It can quickly gauge the approximate area of
each item passing before it, as long as the item appears somewhere in
its field of view (Fig. 8.6).