In-vehicle machine vision technology ensures your road safety

2.2.2 Vehicle Detection

Vehicle detection is the use of various sensors to detect information about vehicles in front, on the side, and behind, including the speed and position of the front and rear vehicles, as well as the size and position of obstacles. Related automobile driving safety auxiliary support systems include adaptive cruise control system (ACC), forward collision warning system (FCW), lateral collision warning system (LDW), and parking assistance system (Parking Assistance System). In ACC and FCW, 77GHZ microwave radar or camera is used to collect information in front of the road, and the road geometry and electronic map data are integrated as input signals for automobile cruise control or displayed to the driver. In LDW, cameras, front detection radars, and side detection radars are used to collect the front and side information of the vehicle, and the road width and other data are integrated as input data for the LDW system. In the parking assistance system, ultrasonic sensors or radars are used to detect obstacle information behind and on the side of the vehicle and display it to the driver. ACC, FCW, LDW, parking assistance systems, etc. have all been studied in Japan's ASV (Advanced Safety Vehicle), the United States' IVI (Intelligent Vehicle Initiative), and Europe's e-Safety projects.

2.2.3 Traffic Sign Detection

Road traffic signs are important ancillary facilities for road traffic safety, which can provide drivers with various guidance and restraint information.

Drivers can obtain traffic sign information correctly in real time to ensure safer driving. In the automobile safety assisted driving system, the detection of traffic signs is achieved through the image recognition system. DaimlerChrysler is currently conducting research on a new generation of image recognition systems. The system first judges the shape of the road sign method, and then reads the text and graphic information in the above shape to make a final judgment. When it is difficult to judge the sign, the driver can also use the pre-recorded electronic map data related to the road sign for identification. BMW also used image recognition technology to study traffic signs in the ADAS (Advanced Driver Assistance Systems) project research. In addition, Toyota of Japan is also actively developing an automatic traffic sign recognition system. Abroad, many researchers have conducted multi-faceted explorations in the study of traffic sign image recognition algorithms. Traffic sign image recognition includes several processes such as traffic sign positioning (i.e. determining the area of ​​interest) and classifier design. The color of the traffic sign and the background and the shape of the traffic sign are clearly specified in the traffic engineering standards, so positioning research can be carried out based on the color and shape of the traffic sign. Due to the large number of traffic signs and the many environmental factors affecting the shooting of traffic sign images, most of the traffic sign pattern classifier design studies are nonlinear classifiers. Traffic sign morphological skeleton and use matching algorithm to recognize traffic signs.

2.2.4 Pedestrian Detection Technology

Pedestrian detection based on computer vision in vehicle assisted driving systems refers to using a camera installed on a moving vehicle to obtain video information in front of the vehicle, and then detecting the position of pedestrians from the video sequence. Pedestrian detection systems based on computer vision generally include two modules: ROIs segmentation and target recognition. The purpose of ROIs segmentation is to quickly determine the area where pedestrians may appear and narrow the search space. The commonly used method is to use a distance-based method using a stereo camera or radar, which has the advantage of being relatively fast. The purpose of target recognition is to accurately detect the position of pedestrians in ROIs. The commonly used method is a shape recognition method based on statistical classification, which has the advantage of being relatively robust. Due to its great application prospects in pedestrian safety, the European Union continuously funded the PROTECTOR and SAVE-U projects from 2000 to 2005, and developed two pedestrian detection systems with computer vision as the core; the ARGO smart car developed by the University of Parma in Italy also includes a pedestrian detection module; Israel's MobilEye has developed a chip-level pedestrian detection system; Japan's Honda Motor Company has developed a pedestrian detection system based on infrared cameras; in China, Xi'an Jiaotong University, Tsinghua University, and Jilin University have also done a lot of research in this field.

3. Machine vision-assisted driving technology for vehicle interior information

Machine vision-assisted driving technology for vehicle internal information uses the on-board video camera to determine the driver's status, position and other information, and implement necessary safety measures, including driver's line of sight adjustment and driving fatigue detection.

3.1 Sight Adjustment

The driver's sight adjustment is to make each driver's eyes at the same relative height, to ensure an unobstructed view of the road and surrounding lanes and the best visibility, thus ensuring driving safety. This technology includes:

(1) The eye position sensor can measure the position of the driver's eyes and then determine and adjust the seat position accordingly;

(2) The motor automatically raises the seat to the optimal height, providing the driver with the best view of the road conditions;

(3) The motor automatically adjusts the steering wheel, pedals, central console and even the floor height to provide the most comfortable driving position possible. The sight line adjustment system has been applied in some high-end cars, such as Volvo's sight line adjustment system, which uses a video camera located in the windshield trim to scan the driver's seat area to find a pattern representing the driver's face, and then scans the driver's face to determine the position of his eyes, and then finds the center of each eye. It takes less than 1 second to complete these three steps.

3.2 Fatigue and distraction detection

Since fatigue driving is the main cause of major traffic accidents, research institutions at home and abroad have carried out research in this field. Compared with awake driving, the more specific indicators of fatigue driving are: fine adjustment of the steering wheel, forward tilt of the head, blinking of the eyelids, and even closing. In the current research on driving fatigue detection and monitoring systems, vehicle-mounted machine vision systems are often used to monitor human posture and operating behavior information to determine fatigue status. The AWAKE driving diagnosis system was developed in the European e-Safety project. The system uses visual sensors and steering wheel control force sensors to obtain driver information in real time, and uses artificial intelligence algorithms to determine the driver's status (awake, possible drowsiness, drowsiness). When the driver is in a fatigued state, the driver is stimulated by sound, light, vibration, etc. to restore the driver to a sober state. Reference [34] uses self-developed special cameras, electroencephalographs and other instruments to accurately measure head movements, pupil diameter changes and blinking frequency to study driving fatigue problems. The research results show that:

Normally, people's eyes are closed for 0.12 to 0.13 seconds. If the eyes are closed for 0.15 seconds while driving, it is very likely to cause a traffic accident.

4 Conclusion

More than 80% of the information of the driver is obtained through vision. In view of the insufficiency of the driver's vision, the development of automobile safety assisted driving system based on vehicle-mounted machine vision has always been one of the research hotspots of intelligent transportation. This paper reviews the current status of the technology in this field and the conclusions are as follows:

1) Analyze the driving operation process and describe the three stages of driving operation;

2) Based on the scope of information acquisition, automobile safety assisted driving is divided into: machine vision for external information and machine vision technology for internal information. Machine vision technology for external information is divided into: visual enhancement, field of view expansion, and road environment understanding, while machine vision technology for internal information is divided into: line of sight tracking and driver fatigue monitoring. The research status of machine vision technology in automobile safety assisted driving systems is reviewed;

3) The article analyzes the current research deficiencies in machine vision technology in automotive safety assisted driving systems, and points out that technologies such as low-visibility driver vision enhancement methods, road environment understanding information fusion, and driver fatigue detection need further research.