INEA develops packaging lines for various industries. In coin wrapping, the pieces are taken from a pallet or a specific pose, oriented and then deposited in a stencil.
While the precision of the parts in their pose is sufficient for their gripping by the robot, this is not the case on the pallet since their position is random. A camera, placed above the pallet, allows in this case to determine the position of the parts. A second camera then makes it possible to calculate the orientation and the precise position of the part on the gripper in order to deposit it in a chablon.
Qualimatest, as a supplier, was able to develop a solution based on machine vision to meet the requirements. The technical criteria have been clearly defined in a specification as in any specific development. In addition to the requirements for accuracy, reliability and cycle time, it also meant reducing engineering costs. High flexibility and an easy-to-use graphical user interface are now basic requirements for all vision systems.
The wide variety of coins, however, has imposed a high requirement for flexible system settings. It must indeed be possible for a user to easily define a new coin. The system must therefore cover all kinds of possible parts for future use.
The size of the pallets and the reflective parts imposed high demands on lighting. The illumination was to illuminate an area of 440x330mm evenly and to minimize parasitic reflections.
Another difficulty was to reach the predicted accuracy taking into account the cycle time. The critical element was the second camera since image capture and processing directly influence the system's cycle time. An algorithm has been developed for calculating the position of the part and its orientation with a measurement accuracy of 0.3 ° for the angle and 0.01mm for the center position, for a computation time of a few milliseconds.
The vision solution of Qualimatest is based on the hardware platform of QMT Vision Inspector. The choice of cameras (1392x1040Pixel) was essentially defined by requirements on the resolution and the ability to accept an external trigger.
The entire system has been developed as a module. The camera 1 captures an image (440x330mm) from the board and recognizes the center of the coins (Ø16-40mm). The camera 2 takes an image (45x60mm) of the coin seized by the robot, its orientation and its position on the gripper are then calculated. A standard computer was used with two image acquisition cards. The data transfer with the robot is done by TCP / IP.
While for the calculation of the orientation of the coin a DOAL lighting could be used, a special custom lighting had to be developed to ensure a uniform illumination of the tray.
Qualimatest integrated the application into its QMTView GUI and developed an algorithm for calculating the position and orientation of parts for the camera 2 in order to reduce the calculation time. The calibration of the system is done using specific tool and markings on the screen. The system can be controlled at any time and without additional help, and if necessary calibrated again.
The development of the vision system was achieved in a short time thanks to the QMT Vision Inspector platform. The cycle time required for the entire system of 2.8 seconds per coin could be achieved inter alia by the performance of the calculation algorithm, the calculation of the orientation and the position of the coin requires less than 30ms. The data transfer between the robot and the vision system was done in TCP / IP.