VISION AND ARTIFICIAL INTELLIGENCE
Our own Artificial Vision Laboratory allows us to have the most suitable technology to carry out a wide variety of camera-based processes.
From classic 2D product localisation applications on a conveyor, to complex solutions that use artificial intelligence algorithms and deep learning that allow the classification of fish catches by species, to the reading of barcodes, QR codes or Datamatrix.
AUTOMATIC COUNTING OF FISH CATCHES
Thanks to the use of Artificial Intelligence algorithms, our fish counting booths are able to identify each fish on a conveyor and count them quickly and efficiently.
This is a very useful tool in the fish markets of fishermen's guilds, where they help to make an accurate calculation of the units per kilogramme that determine the cost of the catch.
You can find more information in this section.
Once the product is packaged, our machine vision solutions allow a large amount of information to be identified.
- Correct printing of the batch identifier.
- Correct printing of production and expiry dates.
- Correct sealing of the flow-pack packaging.
- Existence of cap.
- Partial or complete lack of the packaged product.
You can see an example of application in the cheese production sector in this link.
TRACEABILITY. READING OF MULTI-FORMAT CODES
Sometimes it is necessary to read several types of codes on the same package, we use artificial vision systems that allow greater flexibility than traditional readers.
Thus, we can identify one-dimensional codes such as traditional barcodes and two-dimensional codes such as QR or Datamatrix.
By combining this information with the ERP software and the WMS, we are able to guarantee complete production traceability.
LOCATION OF PRODUCTS FOR PICKING
In order to locate the products on a conveyor, we use 2D artificial vision cameras that measure and locate the centre of each product.
Industrial robots receive the information generated by the vision cameras and use it to accurately pick each product.
In this way we can fill boxes with the preset quantity of products.
You can see an example of boxing boiled potatoes here, and another of picking pre-cooked products here.
3D RECONSTRUCTION AND WEIGHT ESTIMATION
Using machine vision cameras with 3D Laser Triangulation technology we are able to perform an accurate 3D reconstruction of the object.
Knowing the volume of the object and the average density of the object, we are able to obtain an estimate of its weight.
Once we know the weights of various products, such as fish, we employ genetic algorithms that decide which fish we should select to fill trays that add up to a predetermined fixed weight.
This technique saves a lot of calculation time per tray, and improves the hygiene of the process, as the product is only touched once for banding.
Click on this link to access the application video.
The visible spectrum is only a small range of frequencies within the electromagnetic spectrum.
In these images, the different shades of the colour spectrum can be obtained as a combination of red, green and blue.
A hyperspectral image is a set of images of the same object, each represented with different wavelengths, over a wide range of frequencies of the electromagnetic spectrum.
In a hyperspectral image we have much more information about each material than just the information in the visible spectrum.
Each material has its own fingerprint in different bands of the spectrum: each material has its own spectral signature, and it is unique.
This machine vision technique has diverse applications within the quality inspection of primary products.
From the detection of parasites, or foreign bodies, to the classification of meat into different quality categories.
BIN PICKING. EXTRACTION OF PARTS FROM A CONTAINER
In order to pick up parts that are located inside a container and whose location is random, more than just 3D vision equipment and a robot are needed.
The vision system must know beforehand what the part to be captured looks like in order to be able to decide which one to pick up and its exact location in space.
To do this, it must take into account the relationship of each piece with the immediately closest ones.
We have developed a Bin Picking system for picking connectors from racking beams, but it can be adapted to almost any product.
DEEP LEARNING APPLIED TO ARTIFICIAL VISION
We use the most advanced artificial intelligence techniques to solve problems that would be impossible to tackle using classical tools.
We train our systems to "learn" to identify different elements of an image, just as we would train a person. To do this, we show hundreds of images of what we are looking for, so that the system is capable by itself of performing tasks such as:
- Classification of fish catches according to species.
- Quality inspection in search of possible surface defects.
- Identification of different materials in urban waste recycling plants.
- Segmentation - Segmentation. To observe or measure only a specific part of the animal or plant.
3D MACHINE VISION ASSISTED DEPALLETISING
At the beginning of the production line it may be necessary to depalletise the containers to be filled with the product.
In order to place the products, we use 3D machine vision cameras based on Fringe Projection technology.
This technology offers advantages such as flexibility when depalletising different products, accuracy and high reliability.
In addition to providing information about the location of the products, it is able to identify possible interferences that could ruin the process.
You can see an example of depalletising crates containing bottles of beer in this link.