Image recognition for industrial purposes

Industrial processes are often highly automated, with conveyor belts, robots, packaging machines, etc. Sensors are also used to take measurements and make decisions.

People are often still needed to perform quality control in order to intervene if necessary. That is no longer necessary.

Image recognition and computer vision are already used in industry. They are often limited to recognizing fixed patterns within a certain bandwidth, such as circles or squares. This often doesn’t work sufficiently to recognize production errors in the process. The product itself must be identifiable first. Once that has been achieved, counting is possible and quality can be controlled throughout the entire process. Once this is done, insight into the amount of failure and immediate action can be taken if errors are found. Also automated off course.

Although computer vision has great applications in industry, it is unfortunately often insufficiently reliable or can only be used to a limited extent for an industrial application. What now ?

To recognize product quality, a more advanced form of image recognition is required than machine vision can offer. And that’s where Artificial Intelligence comes into play in the form of Machine Learning and computer vision.

We have developed software for this: Intra is our core product and is capable of recognizing people, animals and objects. Intra can be trained for the recognition of any product and with very high reliability.

It is also possible to train specifically on exceptions.

Because software never gets tired and has no problems with strange working hours, this offers interesting opportunities for efficiency improvement.

Operational excellence and the continuous pursuit of improving quality and increasing capacity at the lowest possible investment is characteristic of an industrial process. The requirements for product quality are high and inconsistencies must be removed before the product is sent to the customer. Intra gives a computer the intelligence to do meaningful things with it.

Because Intra can recognize people and you can define safe or dangerous zones. Image recognition is very useful to improve safety. We are not talking here about movement detection, which also works on mice and leaves for example. Intra can provide reliable recognition of people or suspicious behavior.

Property protection can be monitored in warehouses or production sites. We do this by establishing a relationship between people and objects. See the video below about theft detection with Intra at the KRO NCRV (dutch only)

Software can fail or needs improvement. Underlying hardware or cameras can break down or lose sight. We recommend industrial-quality computers that are more resistant to the difficult conditions in a production facility. The image quality of cameras is mostly not very important. Outdoor internet cameras are sufficient in most cases.

Most maintenance and software update can be done by the customer himself, but it is good to be able to call the developer. Additional training is important to prevent incorrect results from occurring along the way. That is why we offer help with this.

Want to know more about image recognition in the industry with Intra? View the presentation with examples and use cases