China News Network, March 3 (Zhongxin Finance Ge Cheng) At the moment when smart cars are gradually popularizing, car radar sometimes brings "shock" to car owners.

Recently, "car radar shows all human figures in the unmanned cemetery" has attracted attention. Online video shows an ideal car parked in an unmanned cemetery, and the car's radar shows that it is surrounded by people, not only walking, but also cyclists.

What's going on? A relevant person of Ideal Auto told Zhongxin Finance, "Because the ideal Max series models use lidar and visual camera fusion perception, limited by the limitations of sensor recognition capabilities in the current market, vehicles will display abnormalities in some scenarios, and we will optimize the algorithm in the subsequent OTA (remote upgrade) to strengthen the recognition ability." ”

A screenshot of the video transmitted on the Internet shows that an ideal car is parked in an unmanned cemetery, and the car's radar shows that there are all people around.

Some industry insiders analyzed that the automotive radar sensor is through electromagnetic waves to detect the distance and position of surrounding objects, if it encounters the interference or reflection of other electromagnetic waves, it may lead to misjudgment or ghosting. But this is generally rare and doesn't last long. Therefore, this matter needs further investigation and analysis.

A technician engaged in the research and development of autonomous driving told Zhongxin Finance that the situation of "car radar showing all human figures in the unmanned cemetery" may be caused by objects that "cannot be determined what it is, it may also be human" being identified as a person, because the lidar is not fused with the visual camera, the tombstone in the video may be recognized by the system as a bush, tree or person.

As for the lidar and visual camera mentioned in the response of the ideal car, according to him, the most basic of the two is that the sensors are different. Further, lidar sees a bunch of point clouds, identified by point clouds, visual cameras are more like the human eye, and the data of different sensors needs to be fused.

How to solve the problem of identifying deviations? The technician said that if it is at the software level, it can be solved through OTA (remote upgrade), but if it is a hardware problem, such as the number of lidar and vision cameras, it is difficult to solve from the cost point of view.

Screenshot of the video related to "ideal car owner driving at night and the central control shows that someone is chasing the car".

In fact, "car radar shows all human figures in the unmanned cemetery" is not the first time that the radar "phantom" has appeared in ideal cars. Previously, the video of "the ideal owner driving at night shows that someone is chasing the car" has also attracted attention.

In the video at that time, the owner drove on a rainy night, driving at a speed of about 60 km / h, and the central control screen suddenly showed that someone was chasing the car behind, one would be directly behind the car, and the other would run to the left rear of the car. After the owner stopped and checked, it was found that there was no one behind the car.

At that time, Li Auto quickly responded, "The above phenomenon is a bug in the visual perception algorithm of Li L8 Pro (AD Pro platform version 4.2). When water droplets are present at the rear windshield camera, a halo will form on the image if illuminated by the lights of the vehicle behind or street lights. The current production version of AD Pro's algorithm has a certain probability of misidentifying halos as human features, so it causes a bug in the display of the perception environment. ”

So, are the reasons for the above two different problems of ideal car radar the same?

People related to Ideal Auto told Zhongxin Finance that the two are not the same. "The uninhabited cemetery is full of human figures" is caused by the recognition ability of lidar and visual cameras; The "night driving is chased by people" is caused by the misidentification of light perception and caused by abnormal display. (End)