blogpost

Increasing efficiency by optical sorting

Date:

Duration:

You can no longer register for this webinar.

Learn more about optical sorting

What is optical sorting?

Optical sorting is the sorting of objects in a material flow, according to certain properties (size, shape, colour, texture, …). Specific cameras and lighting are used to inspect the material. An evacuation system (compressed air, valves,…) is then used to remove objects with certain properties from the material flow. Because the sorting is in a continuous flow of material, the times are usually very short (milliseconds) which makes it challenging to edit the images in real time and take the right actions.

 

Why optical sorting?

You want to be able to separate certain objects in a material flow. The properties of those objects can be very similar, which means that sorting cannot be done with other techniques. Some products can be mechanically sorted by size, or you can use magnetism to remove ferro-magnetic materials from the flow, for example. However, sometimes the only option is a clear difference through optical properties such as color, reflectivity or transparency. Optical sorting is recommended in this case. You also have to look at optics from ultraviolet to infra red. With optical properties you can usually see more than we can see with the unaided eye.

 

Which technology can you use for optical sorting?

With optical sensors, the light actually determines how the objects are detected. Light has 3 properties:
• Intensity
• Wavelength
• Polarization

  1. By playing with those 3 properties you can obtain a different/better result in certain situations. By briefly increasing the intensity, you can greatly reduce the shutter speed of the camera, which gives you sharper images when the objects are moving. You can also overrule the influences of environments (natural and ambient light).
  2. The wavelength of the light determines which property of the object you want to see. Some objects reflect much more light of a particular wavelength than others. This allows objects with different properties to be distinguished from each other. For example, Near-Infrared light is sometimes used to detect bad spots in crops.
  3. Finally, there are also the polar properties of light. This is how the wave of light moves. With normal light, all possible directional angles are present, but by using polarized filters you can filter out the light with a certain angle. As a result, you can sometimes turn an overexposed image into a very sharp image or reduce the influence of external light sources. The most known example of this is polarized sunglasses.

What experience does Vintecc have in optical sorting?

Optical sorting of sprouts for Deman

Vintecc has experience in the optical sorting of crops such as sprouts (reference: use case Deman). The inspection was mainly based on the color properties & size of discolorations on sprouts. The training of the algorithm was done with machine learning techniques. The first ideas about Artificial Intelligence in optical sorting were launched in this project.

Highspeed camerabeeld van vallende spruitjes en indicatie van een slecht valtraject
High-speed camera image of falling sprouts and indication of a bad fall trajectory

Quality inspection of brownies at Poppies

For our use case Poppies, a rejection unit for brownies was made. This unit can reject a brownie in the production line based on size, cracks or deviations in shape. 3D scans were used to scan the brownies at a sub-millimeter level at a very high rate – the production was not interrupted or stopped during the scan.

We try to get more and more AI into our algorithms so that the detection becomes more robust. This means that we are less dependent on certain classical techniques and external influences that can distort the images or measurements. Currently, the core of our algorithms is always based on AI, but sometimes heuristics need to be built around it. In the future, we will increasingly move towards self-learning and self-labelling systems.

Poppies
Scan sensor at a speed of 1500 Hz and gives us a point cloud of each brownie with a precision of 0.8 mm

Tips for people thinking about optical sorting?

Glenn Aesaert, Project Manager at Vintecc: “I think a good tip is to strive for a good constant product flow, because that ultimately determines the capacity that your sorting installation can handle.”

If the product flow is in fits and starts, your capacity will be limited to the peaks in your product flow which will lead to a much lower flow rate. You will always need a lot more computing power for a short time and you are limited on that.

What does a process looks like in a collaboration with Vintecc?

Glenn: “Usually we start with a proof of concept where we test the technologies on a setup that has the best chance. We try to map the real product as well as possible. With these tests we try to capture the hardware and sensory information. If testing is not possible, we will simulate the test environment and in this way we will try to record the sensors and hardware as well as possible. After the proof of concept, a customer-tailored project follows.”

 

Watch the recorded webinar

Relevant Cases

Our solutions

Software programming / Sensor technologies / Automatic code generation

Process simulation / Virtual commissioning / Automated testing

Data collection & exploration / Reporting & alerting / IoT & cloud

Machine learning / Data driven insights / Vision systems