Object detection in cars has gone to the dogs (and cats, cows and deer)
The technology makes driving less dangerous by detecting animals and other objects to avoid a collision, but it's inherently tough to program.
The technology uses sophisticated sensors, cameras and software to spot objects that the host car could collide with. Object-detection technology can be tied into a “driver assist” system so that the brakes are applied automatically, or integrated into night vision so that drivers better detect objects in the dark.
Whereas object detection and autonomous-braking collision prevention systems were initially available only on expensive luxury vehicles, they’ve recently begun to appear on lower-cost cars. The Subaru EyeSight system, for example, uses a stereo camera to detect slow-moving or stopped vehicles ahead. If it determines that a collision is imminent, the system will automatically apply the brakes.
And whereas object detection started out identifying only other vehicles, it’s being refined to recognize others who share the road, such as pedestrians, bicyclists and even animals.
In 2011, Volvo added pedestrian detection to its City Safety system, which was introduced in 2009 to prevent collisions with a car in front in low-speed, stop-and-go traffic. Recently, the automaker unveiled a bicycle-detection system to give pedal pushers a brake.
Now Autoliv, a company that supplies object-detection systems to Audi, Mercedes-Benz and Volvo is developing a night-vision system to detect animals, which it says will be available by next year. The system is designed to detect deer, moose, cows, horses and even wild boar, according to Stuart Klapper, managing director for Autoliv.
“In many cases, it will also detect dogs, cats, rabbits and other four-legged animals,” he told MSN Autos. Klapper added that in the U.S., there are 1.1 million deer-vehicle accidents each year, causing $3.5 billion in property damage.
To save Bambi as well as your bumper, object-detection technology is developing rapidly, said Anders Eugensson, director of government affairs for Volvo. “The reason is that once you have the basic algorithms in place, you can do more fine tuning and add more shapes into the library of objects,” he explained.
Object-detection system hardware is also becoming more refined. “We use a radar [sensor] and camera system that apply sensor fusion to merge the information,” Eugensson said. “Both are in constant development and becoming more sophisticated.” Klapper added that detection range has also improved and that this helps mitigate false alerts.
Even with better software and hardware, detecting additional objects requires extensive field testing. Eugensson said that Volvo logs “an immense number of miles” recording objects on the road, and that the fine tuning of the algorithms is compared with road data on an ongoing basis. Klapper added that it took a large team of engineers five years to develop the company’s animal detection algorithms, since animals come in all sizes and shapes and are more unpredictable than humans.
The Autoliv system is designed only for nighttime detection and doesn’t apply the brakes to prevent a collision with an animal, although it will alert the driver to an impending collision using visual and audible warnings. Eugensson said that while every type of object is challenging to identify, animals are very tough -- especially when the software must decide whether to brake the vehicle autonomously.
"For animals, there’s a decision point on what type of animals we should brake for and what type of animals we are not to brake for,” he said. "One of the biggest challenges is not to have activation when it is not desired, i.e. false positives."
“We spend extensive resources in trying to make sure we can avoid this. From a customer point of view, a false positive is the worst thing you can have. Braking in itself can cause a crash,” Eugensson added. “So we want to make sure not to cause a worse situation by avoiding one situation.”
Doug Newcomb has been covering car technology for more than 20 years for outlets ranging from Rolling Stone to Edmunds.com. In 2008, he published his first book, "Car Audio for Dummies" (Wiley). He lives and drives in Hood River, Ore., with his wife and two kids, who share his passion for cars and car technology, especially driving and listening to music.
EXPLORE NEW CARS
MORE ON MSN AUTOS
ABOUT EXHAUST NOTES
Cars are cool, and here at MSN Autos we love everything about them, but we also know they're more than simply speed and style: a car is an essential tool, a much-needed accessory to help you get through your day-to-day life. What you drive is also one of the most important investments you can make, so we'll help you navigate your way through the car buying and ownership experiences. We strive to be your daily destination for news, notes, tips and tricks from across the automotive world. So whether it's through original content from our world-class journalists or the latest buzz from the far corners of the Web, Exhaust Notes helps you make sense of your automotive world.
Have a story idea? Tip us off at firstname.lastname@example.org.