Abstract
In order to reduce the red light behavior of pedestrians at urban intersections and reduce the traffic accidents caused by pedestrians’ red lights, this paper proposes a pedestrian red light automatic warning solution based on face recognition technology. The pedestrian red light automatic warning system based on face recognition mainly includes information collection module, facial feature recognition module and display exposure module. The system realizes pedestrian crossing automatic reminding function, pedestrian red light recording function, pedestrian red light exposure function, information inquiry function. Wait. The actual results show that the system has good stability and practicability, can effectively prevent pedestrians from red light behavior and reduce traffic accidents.
Introduction
With the rapid development of China’s economy and society and the accelerating urbanization process, a large number of people have flooded into cities, and the number of motor vehicles has increased year by year. Urban intersections are the only way for motor vehicles, pedestrians and non-motor vehicles to travel. Traffic accidents continue appear [1]. As pedestrians in the urban road traffic process, traffic accidents involving pedestrians have been high. According to statistics, the number of people killed in traffic accidents in China exceeds 60,000 per year, 80% of which are caused by pedestrians’ violations. 2.6 people died from red light, and pedestrian red light has become one of the important causes of traffic accidents in China [2].
Due to the pedestrian’s own “constrained psychology” and the management department’s “law not to blame the public”, the pedestrian red light has become a common “Chinese style crossing the road”, which will not only threaten personal safety, but also induce traffic accidents and affect traffic order [3]. However, for the phenomenon of pedestrian crossings with many intersections, uninterrupted time and continuous large area, relying solely on the traffic police and the police to maintain order is far from satisfying the requirements. In this paper, based on the phenomenon of pedestrian red light, a pedestrian red light automatic warning system based on face recognition technology is proposed, which realizes automatic monitoring and alarming of pedestrian red light through video capture and face recognition, combined with voice prompts, close-up warnings, etc. [4]. Under the premise of manual intervention, the standardized management of pedestrian crossing is realized, the number of pedestrians red light is reduced, and the probability of traffic accidents is reduced, thereby constructing a safe, harmonious and intelligent urban transportation environment.
System overall description
Pedestrian red light automatic warning system includes front-end information acquisition/exposure, network transmission and central storage/management. The front-end information acquisition/exposure section includes a dedicated camera for pedestrian red light detection, a speaker, an outdoor display exposure large screen, and an intelligent terminal management host [5]. Realize pedestrian red light automatic detection, sound broadcast reminder, illegal behavior process in large screen display exposure, illegal evidence front-end storage and real-time upload center management platform.
The central storage/management part includes illegal data access platforms, illegal data sharing platforms, traffic information publishing platforms, facial feature recognition platforms, databases, and storage servers. Responsible for pedestrians, red light for evidence, reminders, exposure and other business logic control; data access, sharing, storage; and pedestrian facial feature recognition comparison, review and other functions.
System module design
The pedestrian red light illegal disposal process is divided into six steps: pedestrian red light detection, voice prompt, illegal process capture, large screen display exposure, facial feature recognition, and illegal disposal. As shown in Fig. 1.
Pedestrian red light disposal process.
Steps 1 through 5 are all done automatically by the system without manual intervention. After the identity information of the illegal red light pedestrian (name, age, unit, address, contact information, etc.) is obtained through facial feature recognition, the traffic control department can illegally dispose of it.
Through the analysis of the pedestrian red light process, the pedestrian red light automatic warning system needs to include the illegal forensic module, the illegal pedestrian display exposure module, and the face recognition module.
The illegal forensics module is the information collection module, which mainly monitors the zebra crossing by using the pedestrian red light special detection camera. The built-in intelligent algorithm automatically recognizes the status of the signal light and the pedestrian trajectory. When the signal light is in a red state and a pedestrian crosses the waiting area, the camera can The built-in capture program is automatically started, and the reminder and exposure are automatically triggered according to the result after the capture. The specific business process is shown in Fig. 2.
Flow chart of illegal forensics business.
Pedestrians red light will Capture 3 pictures and combine 1 picture.
The first one: When the signal light is red, the pedestrian crosses the waiting area, and the system captures the first picture. At this time, there will be a voice to remind “the current red light status, please return to the waiting area”.
The second one: Pedestrians ignore the sound reminder, continue to travel, and when they walk to the center of the camera’s field of view, capture the second picture.
Page 3: When the pedestrian walks under the camera’s field of view, capture the third picture;
The fourth picture: a close-up picture of a pedestrian with a red light, can clearly see and analyze the whole picture of the pedestrian. The close-up image can be selected from the first three images, and the close-up image of the face is captured and superimposed into the image.
Composite picture: The final four pictures will be combined into one picture, which will be uploaded to the central platform as a picture of the evidence reflecting the entire illegal process of the pedestrian red light.
The display exposure module is an illegal composite picture for real-time reception of a pedestrian detection camera to capture exposure. The display exposure unit includes an outdoor high-definition LED display, a supporting release program, and a central network publishing platform. It has illegal information release and network information publishing.
The central network information publishing platform can uniformly control the distribution procedures of the large screens at the intersections, and realize the flexible management of large-screen publishing content. For example, when there is no illegal behavior, weather information, traffic control information, safety and civilized propaganda slogans, real-time video of intersections, etc. can be released; when illegal activities occur, synthetic pictures can be released in violation of the law; It can also be released instantly by inserting it. The specific business logic is shown in Fig. 3.
Information display exposure business flow chart.
The detailed technical principle of face recognition comparison.
The face recognition module means that when a pedestrian blinks red light, the high-definition integrated capture unit captures and intercepts the face image according to the walking trajectory, and further analyzes the face image in the background server, and the face image and the face database are further analyzed. Compare and identify, and finally determine pedestrian identity information.
Illegal capture.
Face recognition.
The detailed technical principle of face recognition comparison is shown in Fig. 4. Firstly, the system will perform feature extraction modeling on all images in the imported resident population information database, and the modeled data will be saved to the system’s face feature database. When a pedestrian has a red light, the system will obtain a facial feature picture of the pedestrian, which is also modeled by feature extraction. The feature value is compared with each feature value in the face feature database by 1:N feature, and finally the comparison result including the pedestrian identity information and the similarity is returned.
The development tool for the pedestrian red light automatic warning system is Visual stdio, and the database is SQL 2008. At present, the system has been tested at certain intersections in Xi’an, Shaanxi Province. Figure 5 shows the illegal behavior of the pedestrian red light captured by the system, and Fig. 6 shows the face recognition comparison interface.
Conclusion
Pedestrians are a very common but difficult to manage traffic violation. Although the traffic management department has increased its enforcement, it has had little effect. The pedestrian red light automatic warning system based on face recognition technology proposed in this paper uses the high-definition camera to monitor the zebra crossing, automatically recognizes the status of the signal light and the pedestrian trajectory, automatically activates the illegal behavior of the built-in capture pedestrian red light, and automatically reminds according to the result after the capture. exposure. Through field testing, the system has good stability and practicability, which can effectively prevent pedestrians from appearing red light behavior.