Teaching machines to see what is in front of them

Last week we put a dummy in front of our test robot. Then a real person. First the dummy, so nothing can go wrong. Then the person, because in the end that is who the machine has to see.

What is SAVA?

SAVA is a research project where Conpleks works together with DTU, the Danish Technological Institute and a number of other companies. The goal is to teach machines working in fields and green areas to spot living beings in their path, and to know how far away they are.

Why it matters

A lot can hide in front of a machine. Young deer do not run when a machine approaches. They lie down in the grass and stay completely still. When it goes wrong, it is hard on the animal and hard on the farmer who has to deal with it afterwards. The field and the machine have to be cleaned, and nobody wants that job.

At the same time, more and more machines are starting to drive on their own. That means there is no longer a driver in the cab to keep watch.

Who it is for

SAVA is for farmers and operators who want to work safely. And for the machine manufacturers building the next generation of machines, both large and small.

How we do it

A camera recognises what is in front of the machine. A radar measures how far away it is. The two have to agree, in real time, that there is someone in front of the machine.

It sounds simple. But in a field full of grass, dust and changing light, it is not. The two sensors do not see the world in the same way, and on a moving machine everything has to line up at exactly the same moment.

Why we test in the real world

Test days like last week are where the theory meets the real world. It is where we see what works, and what still needs more work.

Want to know more about SAVA, or how the technology could be used on your machines? Get in touch.