A few days ago, auto blogger @ Changyan CY said that in the process of using xiaopeng automobile (09868.HK/XPEV.US) assisted driving function (NGP), due to small eyes, it is always judged by the on-board DMS (Driver Monitor System, driver monitoring system) as "frequent distraction when using assisted driving" and was warned.
The blogger said: I really just have small eyes, not driving to sleep! This problem you do not optimize and optimize, that small eyes are not worthy of NGP?
At 6 p.m. on July 26, He Xiaopeng, chairman and CEO of Xiaopeng Automobile, specially @ Liu Yilin, deputy general manager of Xiaopeng Automobile Internet Center, asked to deal with this issue. Liu Yilin immediately replied, in optimization. At the same time, the official Weibo of Xiaopeng Automobile also responded that the students of autonomous driving products received optimization demand overnight.
Pictured according to car blogger @ Changyan CY
Blogger narrator
Assist system warning triggered due to small eyes
In addition to Xiaopeng Motors, @Changyan CY posted a special post about the real-time warnings he has triggered many car brand driver assistance systems because of his small eyes. GM's Super Cruise would judge him to be overtired; LanTu FREE would turn on the cold wind in the winter in order to keep him from getting sleepy; NIO ET7 identified him as driving fatigue and wandering...
GM's project engineers said that in addition to the eyes being too small, the blogger's face undulations and height led to too high elevation angles and small targets for the detection equipment. Engineers at Xiaopeng Motors said that the blogger's eye occupancy pixel grid is too small. This case was recognized by the industry as a hellish corner case of DMS. NIO (09866.HK/NIO.US) set up a special research team overnight, and contacted auto blogger @ Changyan CY to prepare targeted optimization.
Pictured according to car blogger @ Changyan CY
Red Star Capital bureau understands that the on-board DMS is mainly to monitor driving fatigue, distraction and dangerous driving behavior. Real-time alerts and video uploads when abnormal driving behavior occurs.
The current mainstream perception scheme is based on visual perception, supplemented by other information. For example, cameras are installed on the steering wheel, A-pillar or inner mirrors to monitor the driver's condition. Infrared cameras and infrared fill lights effectively detect driver status at night or even when the driver is wearing sunglasses.
The current mainstream way is to grasp the driver's facial features through a single vision to make judgments. The degree of eye closure and mouth opening and closing are graded by the driver's facial features, so as to judge the driver's fatigue driving and distracted driving. This is complemented by a torque sensor and a capacitive HOD sensor that detects whether the driver's hand is on the steering wheel, as well as a fusion detection of vehicle status parameters.
In-vehicle DMS applications are still in their early stages
Some insiders pointed out to the Red Star Capital Bureau that "DMS's algorithm has a set of judgment rules, one of which is based on the proportion of time when the eyes are closed to a certain extent, and when the closure is more, it is judged to be sleepy." But this method is ineffective for small-eyed drivers, which is why the car blogger is always judged by DMS to be fatigued driving. At present, there is no unified quantitative indicators and judgment standards in the industry, but DMS suppliers and car companies themselves label and judge. ”
Automotive DMS applications are still in the early stages of the domestic passenger car market. According to the monitoring data of Gaogong Intelligent Automobile Research Institute, in 2021, the insurance volume of new passenger cars (excluding imports and exports) in China that is equipped with DMS will be 543,300 units, with a loading rate of 2.66%, but the year-on-year growth rate will reach 228.87%. With the widespread application of vision technologies such as 3D imaging, optical lenses, and infrared cameras, as well as AI algorithms, DMS functions have continued to grow with the help of the development of the assisted driving industry.
Yiou auto analyst Li Haocheng pointed out that with the in-depth development of vehicle intelligence, L2+ and above high-level intelligent driving functions have accelerated on the car. In the current stage of "man-machine co-driving", subjective factors such as fatigue driving, driving distraction, and aggressive driving have become the main causes of highway accidents and deaths, accounting for up to 37%. The main reason for this is that most of the driving operations in the current "man-machine co-driving stage" still need to be completed by the driver himself, so ensuring that the driver can take over the vehicle at any time has become an important solution to avoid traffic accidents, which is the role played by the DMS function.
At present, some AI algorithm companies are already trying to integrate in-vehicle DMS with intelligent cockpit domain, intelligent driving domain and intelligent interaction functions. Zhongke Chuangda (300496. Zhao Hongfei, founder of SZ), believes that "vehicle DMS should become the strategic hub of cockpit domain, self-driving domain, vehicle control domain and human-computer interaction." ”
Red Star News reporter Wu Danruo
Edited by Yang Cheng
(Download Red Star News, there are prizes for the newspaper!) ）