Tractors now have 20/20 vision, thanks to University of Illinois researchers.

For several years, U of I agricultural engineers have been mounting cameras on tractors and using them to observe the terrain and help steer the machine--without anyone behind the wheel. However, their first automatically guided tractors used monocular cameras (cameras with only one lens) to help steer, but now they have taken machine vision to a new level. Their automatically guided tractor is equipped with a stereo-vision camera mounted in front.

You might say that this camera has two eyes, or lenses.

Much like the human eye, an automatic guidance system for a tractor needs two eyes, or a stereo camera, to see depth, said Qin Zhang, U of I agricultural engineer. With a monocular camera, the field appears 2-D to the automatic guidance system, but a stereo-vision camera brings the field image into the three-dimensional world.

Also, a monocular camera is much more sensitive to light changes, said Zhang. When the light changes, contrast is lost, making it difficult for the tractor's camera to distinguish between the crop rows and the background soil. But the stereo camera is not as sensitive to light changes, making it significantly more accurate.

Zhang and his fellow engineers tested the new system in 2003 on a John Deere 7700 tractor with great success.

"With curved rows and straight rows, we successfully achieved a speed of 8 mph without any problem," said Zhang. The tractor stayed on track. But at higher speeds, he added, the tractor starts to bounce, making it difficult for the camera to study the terrain and guide the tractor.

However, Zhang believes that in the near future they will develop a stronger mounting frame, preventing camera bouncing and making the system accurate even up to 12 mph.

The camera is only one part of the tractor's automatic guidance system, Zhang noted. A Global Positioning System (GPS) sensor determines the tractor's precise position in the field while other sensors provide information about the tractor's motion and calculate its path. The "sensor fusion" system combines all of the information from the cameras and sensors and decides when to tell the electro-hydraulic steering valve to turn the wheel.

But exactly how safe is it for a tractor to be driving itself?

The stereo camera deals with this issue with a new safety feature, Zhang pointed out. When an object is far away, it appears white, while the background is black. But when the tractor gets close to something in its path, the object will turn black and the background will become white. This signals the tractor to automatically stop.

Some farmers may not be willing to completely give up the steering wheel to a machine, but Zhang said it is possible that in the future there will be a combination of traditional and nontraditional equipment. For example, one possibility is that the farmer's tractor could be accompanied by a fleet of smaller, cheaper, self-guided equipment, moving through the field together and getting the work done much faster.

Zhang is optimistic. As he put it, "We're hoping to make farming more enjoyable by reducing tedious or intense work levels."