This project involves developing a camera-based inspection vision system for item classification and inspection capable of examining objects from very close shooting distance. This enables the surface of the object to be scrutinized very closely. We use a commercially available xy-robot along with a color camera mounted on the robot.
The following constraints pertain:
1.The items are bigger than the viewable area of the camera and hence several images must be patched together in order to reconstruct the item’s complete image.
2. The camera must be moved appropriately in order to capture these multiple images. Hence appropriate commands must be sent to the robot holding the camera.
3. A parametric representation that is rotationally and translationally invariant must be generated for each new item.
4. The model of each new item must be learned.
5. A decision-making criteria must be employed in order to inspect about a new item.