The current status and application prospects of face recognition technology for commercial vehicles

This article briefly discusses the technical status, industrial structure and application prospects of face recognition, providing some reference for industry professionals.

Face recognition is a type of biometric recognition. It is a series of related technologies that use cameras to collect images or video streams containing faces, automatically detect and track faces during the collection process, and then perform facial recognition on the detected faces. As a rapidly developing human-computer interaction method in recent years, face recognition technology has brought people a new interactive experience and has become one of the new applications in the development of vehicle intelligence. Applying face recognition technology to the field of commercial vehicles is not only conducive to ensuring the basic safety of drivers and passengers, but also conducive to improving the regulatory efficiency of commercial vehicles, bringing convenience to administrative supervision, and ultimately promoting the development of smart cities.

1. Overview of Face Recognition Technology

(1) Technical indicators and identification effects

The face recognition technology process mainly includes the following four parts, namely, face image acquisition and preprocessing, face detection, face feature extraction, feature matching and recognition. The false recognition rate of face recognition is sometimes called the false recognition rate, or FAR (False Accept RATE), which refers to the probability of mistaking someone for someone in the biometric database and passing. The pass rate refers to the probability of passing through the correct recognition. Through the face recognition technology threshold setting in Table 1, it is found that:

First, the false recognition rate, pass rate and threshold are correlated. The higher the threshold, the lower the pass rate and false recognition rate, and the lower the threshold, the higher the pass rate and false recognition rate. It is necessary to comprehensively consider the two factors of customer experience and risk prevention ability to determine the threshold and the corresponding false recognition rate and pass rate. A more accurate statement should be that under the specified threshold, Company A's false recognition rate is lower than that of other companies, and its pass rate is higher than that of other companies, which means that Company A's face recognition algorithm is good.

Second, the false recognition rate, pass rate and threshold are nonlinear. As the threshold increases, the pass rate and false recognition rate drop rapidly. This means that the threshold cannot be set too high, and as long as the volume is large enough, false recognition will definitely occur. Therefore, face recognition can only be used as an auxiliary authentication method at present, and it cannot be equated with a strong authentication tool.

(2) Key technologies for face recognition

Face recognition requires face image acquisition, image detection, image preprocessing, image feature extraction, feature matching and recognition. Image acquisition is when the user is within the shooting range of the acquisition device, the acquisition device will automatically search and capture the user's face image. Image detection is to pick out some useful information in the image to provide reference materials for subsequent comparison. Due to various restrictions and random interference, the original image is often not used directly, and image preprocessing such as light compensation, grayscale correction, and image sharpening is required. Image feature extraction is the process of feature modeling of the facial features depicted in the image detection link. Feature matching and recognition is to form recognition parameters for various facial organs and feature parts through corresponding geometric relationships, and compare the recognition parameters with the information stored in the database to achieve identity confirmation and object recognition functions.

Among them, "image detection" and "feature matching and recognition" are the core of face recognition technology. The maturity of the technology greatly affects the accuracy, recognition speed and applicability of face recognition. The pattern features contained in the face image are very rich, such as color features, template features, structural features, etc. By depicting certain features in the face, the position and size of the face can be accurately calibrated, and then the useful information can be picked out and image detection can be achieved using these features. Feature matching and recognition mainly rely on software algorithms to compare and judge the facial features to be recognized with all the original parameters in the database, and judge the identity information based on the degree of similarity. Generally, the judgment time is required to be less than 1 second.

(3) Application fields of face recognition technology

The applications of face recognition can be divided into two categories: one is identity confirmation, which compares the face image with the image of the person already stored in the database to determine whether it is the same person. Through fast face recognition and comparison, identity authentication functions such as mobile payment authentication and security identity verification can be realized; the other is object recognition, which matches the face image with all the images already stored in the database to determine the identity of the detected object.

Facial recognition technology has gradually been applied to many fields such as finance, social and corporate management, education, campus security, public safety, mobile phones, judicial criminal investigation, transportation, and service industries.

2. Current Status of In-Vehicle Face Recognition Technology

(1) Application scenarios of in-vehicle facial recognition technology

With the improvement of automobile intelligence, the demand for automobile personalization and safety is becoming stronger and stronger, and the automobile industry is paying more and more attention to face recognition technology. As a new human-computer interaction method, face recognition technology has been applied in a small number of mass-produced passenger cars. As shown in Figure 1, at present, in-vehicle face recognition technology plays a vital role in automobile anti-theft, driving safety, and even automatic driving. It can currently realize two major functions: identity authentication and driver status monitoring. Among them, identity authentication functions mainly include: vehicle unlocking and starting, in-vehicle payment, personalized services, qualification certification, etc.; driver status monitoring functions mainly include: fatigue driving monitoring, distracted driving monitoring, health status monitoring, emotion recognition, etc.

Figure 1 Main functions of vehicle-mounted face recognition technology

During the driving of commercial vehicles, the main causes of accidents are the driver's fatigue, looking at the phone, distraction and other bad driving behaviors, which may cause the vehicle to drive out of the lane or collide with surrounding vehicles. Therefore, the application of face recognition technology in the commercial vehicle field to monitor the driver's status during driving has great application value.

Driver status monitoring based on face recognition technology will further enhance the intelligence level of commercial vehicle cockpits, mainly in two aspects: first, integration with the ADAS system to realize early warning functions, real-time monitoring and measurement of changes in driver's facial features, head movements and upper body movements through cameras, analysis of the driver's status through artificial intelligence algorithms, comparison with pre-set detection standards, judgment of the driver's attention level, determination of whether the driver has bad driving behavior, and timely issuance of early warning prompts. At the same time, trunk transport vehicles in L2 and L3 autonomous driving will strengthen real-time monitoring of the driver's ability to take over; second, it can realize personalized interaction in the cockpit and provide higher-level intelligent applications such as line of sight tracking and HUD linkage, in-vehicle perception during rescue, etc.

Driver status monitoring mainly detects the following characteristics of the driver: ① Changes in head and upper limb characteristics, such as facial expressions, facial contours, eye, nose and mouth positions, facial orientation, and hand movements; ② Eye signals, such as line of sight direction, eyelid opening and closing degree, blinking frequency, and pupil status; ③ Other biological indicators: such as analyzing facial structures such as skin, periorbital area, lips, and bones to identify body fat, BMI index, and blood pressure. For example, an early warning is issued when the driver's line of sight is more than 30% of the time or the line of sight is more than the threshold.

(2) Application mode

At present, there are two main modes for using facial recognition technology in automobiles: on-board application and management-side application.

The in-vehicle application is to install cameras inside and outside the car. The external camera is generally embedded on the A-pillar or B-pillar to identify user information, thereby realizing functions such as door sensor unlocking, engine start, and in-vehicle information system account login. In addition, various personalized services can also be enabled by "face scanning". After the user's identity verification is passed, the face recognition system will make personalized adjustments to the seat angle and infotainment content based on the user's unique ID usage record. This type of function is mainly used in passenger cars.

Commercial vehicle face recognition is to install the camera in the car on the center console, in front of the driver's head or in the rearview mirror. Its main function is to realize the driver status monitoring function. For example, dangerous goods transport vehicles, passenger vehicles, and slag transport vehicles can use face recognition technology to identify whether the driver is off duty or the driver's face is blocked. When the vehicle is in the forward gear, the camera does not detect the driver or the camera is blocked, and an immediate warning can be issued; during the driving process of commercial vehicles, when the driver is distracted, such as playing with mobile phones, smoking, making phone calls, looking for things, turning back to chat, etc., the system warning prompt will be triggered.