Back

Eliko’s new industrial object recognition system makes welding more efficient

Machine vision is increasingly used in a range of industrial automation. Eliko delivered a novel industrial object recognition system that considerably speeds up the work of welding robots. Our client was a leading domestic industrial machinery company, Norcar BSB Eesti.

Norcar was looking to automate the process of feeding products to their welding robot. Previously the process took the operator’s time away from more important tasks. Norcar was looking for a system that would automatically recognize the prepared piece of machinery moving towards the welding robot and would start the right welding program corresponding to the identified product. One important condition was that the system had to operate without additional markers like barcodes, QR codes or similar.

Training the object recognition system needs optimal lighting

Before the image processing could start, the object recognition system needed training. For that, a set of pictures was taken of each product in different lighting. As the shape and size of the welded products varied significantly and they all needed optimal lighting conditions, we designed a custom lighting system. This consisted of a special luminaire with 16 sections each of which can be flexibly controlled, if necessary.

 

A system suitable for high-speed applications

The core of object recognition system is a set of sophisticated image processing and machine learning algorithms that are based on a technique called support vector machine. In order to identify the right object, the system needs to learn each object separately. Thanks to the automation and software optimization, the teaching process is very fast and takes less than a minute for each object. During normal operations, the recognition takes only a few hundreds of milliseconds, making our system suitable for high-speed applications.

 

The data acquisition and processing units are seamlessly integrated with the welding robot’s controller. This means that after recognizing the product in the feeding area, the robot can automatically start the right program. The object recognition system also has a web interface for remote monitoring the status of the system and the welding process.

 

Automation brings efficiency to welding

Our customer – Norcar BSB is a manufacturer of mini loaders, small farm tractors, and their accessories. Eliko’s object recognition system increases the welding robot’s operation time and reduces the number of personnel needed to work in the hazardous environment. The remote process monitoring option also helps the operator to conveniently follow the process status from a distance.