Advanced Object-Recognition technology developed by Mitsubishi Electric

Mitsubishi Electric Corporation (MEC) has developed what is believed to be the industry’s highest performing automotive camera technology that can detect various object types at distances of up to about 100 metres. This capability will enable drivers to receive advanced warning for enhanced driving safety in coming mirrorless cars.

The solution, which is based on MEC’s proprietary Maisart-brand artificial intelligence (AI) technology, is expected to help prevent accidents, especially when drivers change lanes. Mirrorless cars that replace rearview and side mirrors with camera-monitoring systems were approved for use by the United Nations in 2016, and the first commercial mirrorless cars are expected to be launched in Japan as early as next year.

Mirrorless systems will do away with the conventional rearview mirrors which use reflective surfaces to show the view behind or along the sided of the vehicle. Instead, small cameras are used to capture images in real-time and shown on a display panel. There are advantages and disadvantages to such a system and in the initial period of introduction, the displays will be in their traditional locations as drivers may need time to get used to looking at display panels in other areas.

Conventional mirrors have long used reflective surfaces to show the view behind, with mirrors on the doors being added from the 1970s to broaden coverage.
The use of small cameras to show the view directly behind the vehicle (left) has been available for many years already. It helps during parking as the driver can also see if there are any obstacles or even small children behind. The picture on the right shows Honda’s Lane Watch system which uses a small camera in the door mirror to show the view along the side of the car.

In fact, the use of cameras to provide real-time imagery has already been around some time – with rearview cameras. These have become more common as the cost of the hardware has fallen. In recent years, Honda has also introduced a useful feature called Lane Watch which employs a small camera mounted on the opposite side door mirror to show an image along the side of the car.

MEC’s technology employs a new computational visual-cognition model that mimics human visual behaviour to rapidly focus on appropriate regions within the field of view. Compared to conventional camera-based systems, the new technology can significantly extend the maximum distance of object detection from about 30 metres to as far as 100 metres. It is also said to improve object detection accuracy from 14% to 81%.

Many concept cars have proposed electronic camera-based mirrors. The Hy-Wire concept by GM in 2003 had small cameras on the doors which captured images shown on small screens in the car. However, drivers will have to get used to the idea of a rear view being shown in the middle of the steering wheel.

The computational visual-cognition model’s relatively simple algorithms free system resources for real-time performance, even in on-board systems. MEC’s Maisart AI technology, which makes possible compact AI for low-cost devices, can distinguish between object types such as pedestrians, cars and motorcycles. Maisart is an abbreviation for ‘Mitsubishi Electric’s AI creates the State-of-the-Art in technology’. MEC is continuing development of the technology to achieve better performance when driving in bad weather, at night and on winding roads.

[Chips Yap]