The idea of fitting cars with AI-powered cameras to replace vehicle mirrors may sound scary on the surface. However, from the overview of things, artificial intelligence technology seems to be the only way to curb road accidents, which so far claims over 1.25 million lives each year.
According to the recent research which looked to analyze the cause of road fatalities, it was seen that human error tops the list, that included: poor decision making on the part of the driver, and failure to detect blind spots among others. These two aspects contributed to more than 65 percent of all car crashes.
Improved AI-based Car Cameras Might Help
Now, even though we are yet to realize perfectly functional driverless cars, which is said will eliminate road carnage, Mitsubishi Electric is focusing on AI-powered car parts to be fitted on its long-awaited mirrorless vehicles. Hopefully, this will play a role in reducing the high rate of road fatalities as well.
During its annual R&D Open House meeting, Mitsubishi Electric stated that it has managed to develop what it trusts is the top version of mirrorless car technology -factored in its upgraded AI-based camera systems.
The company explained that what we have currently in the market as camera systems that feature motion detection technology cannot be relied upon. Because the far they can detect objects is up to 30 meters away. And sad of all is that they can only guarantee a 14 percent vision accuracy, which means they are not a safe option to be used as eyes on a car.
With the new Mitsubishi’s mirrorless technology, the recognition expanse extends to 100 meters, with 81% vision accuracy. That’s ideally a plus of 70 meters in recognition distance and 67% vision improvement to the technology.
The Difference Between Motion Detection Tech and Al-Based Object Recognition
Kazuo Sugimoto, Mitsubishi Electric’s, Senior Manager puts it clear saying, “Motion detection technology is limited because it can’t see objects which are of a certain distance away.” But when it comes to the new AI-based object recognition tech, the system is able to instantly sense objects up to about 100 meters away.”
How this AI-based Car Camera Works in Mirrorless Cars
To achieve that, the new system employs two complex processes, and the first process is a computational visual-cognition. This ideally mimics the way a human sees an area, to extracts object information after analyzing the background -in distant viewing.
In the second process, the collected data is received by Mitsubishi’s deep learning AI tech christened Maisart, where the object’s information is analyzed and placed into the relevant category. In other words, all pickup trucks, buses or small cars will appear in their respective categories, in the screen monitor that the driver will be using (which ideally would replace the mirrors we use.)
Interpreting the System
With the current state of the system, a yellow rectangle designates an approaching car, while a blue rectangle is meant to alert you of an oncoming truck. However, Sugimoto admitted that this can be presented better and so they are testing various ideas that would make it more natural to drivers.
“We also intend to improve the system so that it performs as expected even in bad weather or during night driving. Besides that, we will also be working to increase distance mastery and recognition accuracy,” he explained.
In summary, we expect this technology to be fully implementable before 2022 because that is the time which Japan says it would officially launch autonomous cars for public use. However, Mitsubishi expects to avail its mirrorless cars to the market before mid-2019.
To be honest, it’s hard to tell the number of people that will want to buy a mirrorless car right after the launch. Can you?