New electronic mirror recognizes distant vehicles with high accuracy

Mitsubishi Electric Corp has developed a safe driving support system that can recognize vehicles running up to 100m behind the user’s vehicle with a high accuracy by using automotive electronic mirrors.

For example, when the user’s vehicle is changing lanes on an expressway, the system can detect vehicles approaching from behind and give a warning. Mitsubishi Electric aims to apply the system to automatic lane changing systems used for “level 3” or more advanced automatic driving.

Electronic mirrors are used to replace door mirrors with cameras and monitors. In general, people point out that current electronic mirrors to be commercialized in 2018 cannot recognize distant vehicles. It is because a wide-angle camera is used to cover a wide area and the images of distant vehicles become small.

In view of this, Mitsubishi Electric applied the mechanism of the visual perception of human brain to the image processing of camera. Specifically, as in the case of human vision, which has a high sensitivity to “conspicuous objects,” the new system extracts only distant “conspicuous areas” (e.g. areas whose contrast is greatly changing) from images captured by the electronic mirrors.

The new system processes extracted images by using deep learning and judges, for example, whether the images are vehicles or trucks. It enables to reduce the amount of data processed because it extracts only conspicuous areas of images, and the data can be processed with a general-use CPU such as an existing automotive microcomputer, Mitsubishi Electric said.

For commercialization, Mitsubishi Electric will make improvements to the system so that it can recognize vehicles with a high accuracy even under severe conditions (e.g. nighttime, tunnel, rain, etc). Also, the company plans to make the system support pedestrians and two-wheeled vehicles so that it can be used on general roads.

Give it a share: