SHOP
English (EN) 
Deutsch (DE) 
Visual inspection of cooling boxes for pharmaceutical and food industry
Our customer eutecma GmbH is a leading manufacturer of cooling solutions for the transport of temperature-sensitive products in the food and pharmaceutical sectors. Through its retecma division, the Mannheim-based company is implementing the world's first recycling cycle for modular cooling boxes in sizes from 4 litres to 1588 litres.
These EPS (expanded polystyrene) boxes are equipped with RFID chips that give each component a unique identity including type, weight and number of recycling cycles. Used or no longer required boxes can be returned to special "refreshment centres" where they enter a cycle that guarantees their reuse.
This is where phil-vision comes into play. In one system, a camera system measures the returned boxes, which consist of several modules, compares them with the corresponding 3D models and checks the outer shell for damage. This is all in the interests of sustainability, as the transport containers checked in this way can be reused more than 15 times under optimum conditions, provided that they are not damaged.
What makes the task so challenging is the fact that the boxes do not arrive at the test facility unmixed, but wildly mixed up. However, upstream processes ensure that the control system recognises the current model and can load the appropriate 3D data. The difficulty, however, lies in capturing the different sizes and achieving sufficient resolution on a measurement volume of varying sizes.
After initial experiments with ToF/LIDAR technology, the final decision was made to use stereo cameras. According to project manager Markus Kirstein, a key advantage of stereo cameras in combination with pattern projectors is the high degree of flexibility with regard to the materials to be analysed. When asked about the main challenges for the vision system, Kirstein cites durability, reliability, performance and repeatability.
The sensor head used for the visual inspection consists of five stereo heads with 20MP GigE cameras, which are mounted at a distance of 1550 mm from the transport rollers, covering an area of approximately 200x160 cm and capturing the boxes four times from above at an angle and once straight from above.
This data is used to create a 3D image, which is then compared to the corresponding CAD model. The calculations are performed on an IPC that is connected via ProfiNET.
The cameras are connected via the GigE Vision interface and calibrated and reconstructed using Halcon image processing software. The matching and checking of the point distances to the matched model is carried out in a software component developed in-house.
Another difficulty in the application is distinguishing between main defects pseudo defects and artefacts, as well as possible material deviations. If the material deviates significantly from the CAD, it can be very difficult to find defects.
Checking for smallest damages
The system currently inspects boxes for dents and chips > 5mm. For this reason, high accuracy is extremely important. Defects are initially detected using a defect algorithm as basis for future categorisation and the use of artificial intelligence.
If defects are not detected correctly, defective parts can pass as "OK" or good parts can be categorised as "not OK", which can distort material flows and lead to customer complaints.
