At present, subway video surveillance points are set up more and more densely. In multi-line transfer stations, the number of surveillance points can reach as many as 200. From the current subway security construction needs, more and more emphasis is placed on the video surveillance system without dead angle coverage, and the camera The scale is also getting bigger and bigger.

Since it is a special area of fully automated management. In addition to prevention after the event, prevention in advance is naturally essential. According to common sense, if we want to prevent trouble beforehand, we have to use one of the killer features of our security: intelligent analysis. From an extended point of view, intelligent analysis mainly uses software to analyze abnormal conditions in the image to alert the background to prevent problems before they occur. So, if you want to analyze the abnormality from the image, the premise is that the image must be clear! At this time, the demand for high-definition has become another important technology for subway security.

I have to mention: intelligent analysis

Due to the high cost and long cycle of subway construction, the subway lines of various cities in China are often distributed in busy areas in order to better solve the traffic congestion problem. This determines that the subway operation is basically at a high load, especially in the morning and evening peaks. They are all overcrowded. The greater the passenger flow, the greater the pressure on operations, and the more unsafe factors or hidden accidents. Therefore, the demand for operation and public security video surveillance systems has become inevitable, and the proportion in the subway system continues to increase.

Wang Wei, general manager of Beijing Besting Technology Co., Ltd. believes: “As the number of monitoring channels continues to increase, it becomes impossible for personnel to keep a close eye on the screen, and the method of post-mortem inspections cannot meet the requirements of monitoring management efficiency. Therefore, more attention is paid to accidents. The advance prevention of hidden dangers, and timely warning of abnormal events, etc. If this part of the investment actually plays a role, it will produce obvious benefits and high cost performance. Therefore, the application of intelligent video analysis technology in the subway system has become an irreversible trend, and there is more and more Many functions come into practical use.” Starting from the practical application, Wang Wei, Beijing Bothiting, summarized the following intelligent analysis functions.

The functions related to passenger flow detection have become the top priority, especially the counting and statistics of the cross-sectional flow of the crowd or the number of people in the area, mutation detection, density detection, etc., to detect abnormal crowding events in time, and strengthen personnel intervention in key locations in time. Passenger flow control or guidance, take precautions, take precautions, and prevent accidents such as trampling. But it may also be a major technical difficulty, because the camera viewing angles in halls, passages, platforms, etc. are all limited, and the passenger flow is more complicated and has different characteristics. At present, conventional passenger flow statistics algorithms (such as head detection at entrances and exits of shopping malls) generally cannot meet the statistical accuracy requirements in the subway environment. At present, some manufacturers have solved the problem by optimizing several sets of different passenger flow detection algorithms. The general time accuracy can reach more than 95%. The period can reach more than 85%;

The elevator accidents in Shenzhen and Beijing have made the subway also demand for video detection of abnormal events such as escalator retrograde and stop. If it works, it may reduce losses and reduce the degree of casualties. However, there are few suppliers of this algorithm at present. Some suppliers can only judge the elevator status based on the characteristics of the people on the elevator, and the effect of early warning is limited, and the technical performance needs to be further improved. At present, the video analysis algorithm developed by some manufacturers does not use the personnel on the elevator as a reference, and can alarm within 2 to 3 seconds, which has a high detection accuracy;

Face recognition function, generally by setting up high-definition cameras in the passageway or at the entrance and exit of the gate, collecting part of the front face image in the passenger flow for detection. After the face is detected, it is stored and compared with the target face database. When a suspicious person is found, the alarm is reported; face retrieval can also be performed. However, in this scenario, it is currently impossible to make requirements for the accuracy index of face recognition, so it can only be positioned for the purpose of intelligence collection;

Other functions include passenger chase and violent movement detection, real-time monitoring and alarming of passenger chasing, fighting, and violent activities; personnel entering the restricted area and trailing detection, timely detection of intrusion or tailing of people in key protection areas and alarm ;Video status detection/image abnormal alarm recognition, which can monitor each channel of video screen at all times and detect and diagnose the video status. 

When it is found that the video is abnormally working (such as video loss, video movement and deviation from the setting) Good scenes and videos are blocked so that they cannot be monitored normally, and due to various reasons such as focus deviation or water condensation, the video is blurred or the video quality is reduced) automatically alarm; alarm and display scenes are reorganized, and alarm events are automatically created after alarming. Capture pictures and videos (the length of time before and after the event can be adjusted and set by the user, up to 2 hours), and the whole process of the alarm event can be replayed at any time, realizing on-site reorganization, and easy to find the whole process of the event .

However, in subway video, due to camera viewing angle limitations and serious and frequent occlusion between targets, even some relatively familiar conventional video analysis functions (such as detection of personnel crossing the line, personnel stranded, personnel retrograde, etc., and detection of objects The accuracy of items left, items removed, etc.) has also increased considerable difficulty. In addition, the rail transit station should be one of the complex indoor environments, and many new issues have been raised for the algorithm, which require special efforts to deepen the analysis and solution. With its own years of experience, Bostin has significant effects in technology and algorithms. It can provide multiple sets of algorithms. Through embedded hardware or software, it can more accurately achieve passenger flow statistics under different camera angles on different occasions. Passenger flow data realizes local and remote statistical management in various forms.

Can’t use it: HD

As mentioned above, in order for intelligent analysis to exert a greater effect, it must first ensure the video surveillance in the subway station. First, it must achieve high-definition monitoring of the monitored object, so that every movement of the person can be clearly seen. The benefits of this are not only conducive to the back-end intelligent analysis, when the front-end security personnel cannot accurately perform security checks on the inspected personnel, the back-end monitoring personnel may be able to see the clues from different angles, so as to prevent problems before they occur.

Zhejiang Dahua Zhou Xiaojun introduced: "The early subway video surveillance system used analog front-end + analog matrix + digital coding storage equipment to build the system, which is also the main system used by most of the established subways. The current application status of the system is still It is relatively stable, but there are some problems. First of all, the image resolution collected by the analog front end is not high enough, and can only reach the resolution of D1 after encoding. The system with the analog matrix as the core has limitations in data sharing, remote transmission, etc. High-definition applications make the monitored images clearer. After networking, it makes data sharing and remote transmission easier and more flexible. "It is reported that network high-definition and intelligentization are also being used more and more in subway security. For example, Wuhan Metro Line 2 and Nanjing Metro Line 3 have adopted full HD solutions.

At present, the number of subway video surveillance points is becoming more and more dense. In multi-line transfer stations, the number of monitoring points can be as many as 200. A subway only operates about 1,000 monitoring points, and up to 2,000 monitoring points. From the perspective of current subway security construction needs, more and more emphasis is placed on the coverage of the video surveillance system without dead ends, and the scale of cameras is getting larger and larger. Since 2010, the demand for high-definition 1080P/720P video surveillance in subway security has gradually emerged. So what impact does HD surveillance have on traditional solutions, and what are the main advantages of the subway HD surveillance system?

First of all, there is no strong correlation between the scale of monitoring points and the number of servers. A single video management server can satisfy several sub-channels of SD and HD image monitoring. There is no major system bottleneck problem in the system;

Secondly, there is no station-level global failure point in the system. Due to the use of a soft switch architecture similar to that in the telecommunications field (separation of service flow and control flow), no matter the failure of the management server or the storage server, the monitoring real-time flow will not be interrupted. , The storage flow is not affected, which ensures that the monitoring data is still recorded normally, and greatly guarantees the high reliability of the system from the architecture;

Thirdly, when there are emergencies or accidents and multiple hotspots appear, large-scale burst traffic is replicated and distributed by switches through multicast technology, thus easily solving large-scale multi-user problems and satisfying public security and enterprises High-reliability scheduling requirements for management in emergency situations;

The direct storage of raw data greatly reduces the pressure on the performance of the disk array, so that it can easily handle large-scale high-definition image input while ensuring multi-channel high-definition playback. The bare data block method also enables data to be stored as you see it, what you store is what you use, and what you use is replayed with high performance, which guarantees the fundamental requirement of the subway industry to deal with emergencies quickly.

It is reported that more and more metro solution manufacturers are also beginning to take the all-digital road. For example, the IP all-digital surveillance solution launched by Zhejiang Yushi Technology is based on the NGN architecture and fully supports HD, SD, and virtualized storage management with high performance. Management platform, multiple backup technologies (management platform dual-system hot backup, N+1 backup, storage device N+M backup and network layer backup), and pay attention to the reliability of the system, the management of the business logic of different departments, and the realization of multiple lines Information sharing and collaboration, physical resource sharing, integrated network management. In addition,

Zhejiang Dahua launched a full HD solution, using the multicast function, that is, the front-end camera can send out two streams in real time, one is a unicast stream, the other is a multicast stream, and the multicast stream is mainly for previews. Used, the unicast stream is mainly used for storage. When the multicast mode is used for preview, the speed of real-time preview can be improved, and the immediacy of image display is guaranteed; the unicast mode is used for storage, which can improve the reliability of storage. And Bosch Ambow combines its rich experience in broadcasting systems and access control systems to develop subway security solutions that use high-definition and intelligent analysis to better realize humanized and high-security management.