Analysis: Video image intelligent analysis and processing technology

Video surveillance is one of the most effective means in security systems. Since its widespread use in the 1990s, especially with the construction of large-scale projects such as safe cities, tens of thousands of video images have brought difficulties to the monitoring and management of the majority of staff. A monitoring staff often has to take into account hundreds or even thousands of monitoring video signals. After working for a long time, the monitoring staff will gradually enter a state of physiological fatigue, and often due to excessive fatigue, they will not be able to respond to potential security incidents in the video images in the first time.

How to automatically discover, analyze and process useful information from massive images, and make video surveillance systems an important means of preventing accidents before they happen, has become one of the important issues that security system construction and security management departments are increasingly concerned about. Faced with an increasing number of monitoring points, massive video surveillance systems will only be "visual" but not effectively "monitored" without video analysis. Therefore, people expect a product that can automatically capture suspicious targets through the recognition and analysis of video images, and promptly warn of possible security incidents, thereby reducing the workload of monitoring personnel.

The meaning of intelligent analysis and processing technology Domestic technical status Intelligent video analysis uses intelligent neural network technology to process video images in layers and separate people or objects that are useful to the system. Neural network is an information processing system based on simulating the intelligent characteristics and structure of the human brain. It explores the model that simulates the function of the human brain nervous system by modeling the basic units of the human brain and designing typical excitation functions, and develops a software simulation technology with parallel distributed processing and storage, high adaptability and self-learning functions, and the ability to analyze more complex nonlinear systems.

Video image intelligent analysis and processing technology is to automatically analyze and extract key information from video sources. If the camera is regarded as a person's eyes, the intelligent video system or equipment can be regarded as the human brain. Intelligent video technology uses the powerful data processing function of computers to analyze the massive data in the video screen at high speed, filter out redundant information, and provide useful key information for the monitor.

The "heart" of intelligent analysis: DSP+algorithm

The reason why intelligent video surveillance can play an important role in intelligent video analysis and change the traditional passive monitoring mode with active monitoring is that its fundamental means of realization lies in the high-speed real-time performance of high-speed digital signal processors (DSPs), which greatly improves the immediacy and effectiveness of monitoring. Zheng Xiaolong, manager of the general DSP business development department of Texas Instruments, believes that with the enhancement of DSP performance, video intelligent analysis algorithms can be transferred from CPU processing to DSP real-time processing, so video intelligent analysis has achieved the transition from background to front-end. Digital video surveillance products of more devices can use DSP to add video analysis functions to achieve comprehensive intelligence. After several years of promotion, TI's universal digital media processing platform (DaVinci) has been recognized by many digital video surveillance customers. When expanded to intelligent video analysis applications, it is beneficial for digital surveillance systems to reduce system costs, shorten development cycles, and support a variety of interface standards. Among them, the representative ones are the DM644x with AMR core and the DM643X series platform without AMR core. Both have C64x+ DSP cores, which are upgraded from C64x cores. They are compatible with C64x algorithm software and have higher performance and efficiency.

Scenarios are complex and promote the diversified development of intelligent analysis technology

In recent years, intelligent video analysis technology has developed rapidly. Intelligent video analysis technology and its market are slowly penetrating into various industries.

The combination of intelligent technology and existing security products can be used in many aspects. Judging from the current combination of technology and security products, the following needs can be met: detect and track moving targets, analyze their behaviors, and improve the performance of real-time warning; perform logical abstraction on the input image stream, generate video indexes, and improve the efficiency of video retrieval; perform lens control and video compression according to the scene content, which not only reduces the total amount of video data, but also ensures the quality of collected images when abnormal events occur; use facial images captured in the video during warning to conduct investigation and authentication.

Beijing Broadview Technology is a company founded in June 2007, dedicated to the research of intelligent video surveillance algorithms and the development of intelligent video surveillance systems. With strong intelligent video surveillance technology reserves and R&D capabilities, and a first-class R&D team with high practical capabilities, it is able to optimize the application of intelligent video surveillance technology in the development of intelligent video surveillance products, and has launched several series of intelligent video surveillance systems, including Broadview defense system, Broadview smoke detection system, Broadview flame detection system, Broadview crowd monitoring system, Broadview statistics system, Broadview tracking system, Broadview diagnosis system, which can meet the video analysis and processing requirements of multiple application sites.

In the current situation of increasingly complex video surveillance systems, intelligent video analysis must not only keep daily management operations simple and convenient, but also take into account the increasingly wide coverage of the system and the economy and effectiveness of the system. Finally, it must also integrate the monitoring system with the access control system and the alarm system into the same platform, allowing the integration of other related technologies while maintaining the flexibility and scalability of the system.

Shanghai Zhuoyang Technology Co., Ltd. has developed its own brand IVOSystem series of intelligent video analysis systems based on the world's leading ObjectVideoOnBoard? intelligent video analysis technology, which has attracted wide attention from all walks of life. The system perfectly combines cameras, intelligent video analysis, video encoding, and communication networks to automatically identify people, cars, and other objects, and identify, analyze, and judge their behaviors, and can actively warn of various security events. Rules can be set, including crossing single cordon, crossing double cordon, entering, leaving, appearing, disappearing, abandoning, taking away, scene change, and retention.

SmartCatch intelligent video analysis technology launched by Japan's NEC is a behavior recognition solution for CCTV monitoring. It can capture various pre-set illegal combinations, such as entering security-sensitive areas, illegal parking, leaving items behind, moving or destroying items, etc. The system can automatically generate alarm signals in various ways (ringing, short messages, emails). SmartCatch is suitable for existing monitoring systems, including use with CCTV video surveillance recording and access control systems. It can be expanded at any time as needed, and configured, managed, monitored and reported through software. SmartCatch is compatible with analog and digital cameras, and can also be combined with other sensors, such as access control, biometrics, and microwave equipment.

The role of video image intelligent analysis

Intelligent video analysis needs to automatically identify the monitored object, timely perceive the changes of the target object, and realize automatic alarm and other functions when abnormal situations occur. We know that "monitoring" and "control" are complementary. The former part of "monitoring" is not too difficult to do, but the latter part of "control" is crucial to truly do. In order to achieve the requirements of "control" better and faster, it is necessary to use intelligent video analysis technology to improve the effectiveness of "monitoring".

The industrial development of intelligent video analysis technology provides an excellent opportunity for intelligent video surveillance. The intelligent video analysis system extracts and screens abnormal behavior events in the video in real time and issues early warnings in a timely manner, changing the passive state of traditional video surveillance systems that can only "monitor" but not "control", solving the problem of difficult evidence collection after the event, and making monitoring more proactive. Compared with traditional video surveillance systems, its faster response time and more powerful data retrieval and analysis functions have greatly improved monitoring capabilities.

As technology continues to mature, some professional IVS research manufacturers have gradually emerged around the world, such as ObjectVideo and Vidient in the United States; NICE, Mate, IOImage in Israel; IOmniscient in Australia, etc. These manufacturers have entered the Chinese market one after another, which once caused foreign manufacturers to monopolize the domestic intelligent video analysis market. However, due to the differences between their technology and China's national conditions, they were unable to truly penetrate the Chinese market. This situation not only brings challenges to domestic intelligent video technology research manufacturers, but also provides a very good development opportunity. Professional intelligent video analysis research manufacturers such as Zhianbang, Feiruis, and Bellxin have gradually emerged in China, and related products have also achieved breakthrough applications in safe cities and various key industries.

The Ministry of Public Security clearly stated in its "Eleventh Five-Year Plan" that face recognition and intelligent target recognition and analysis are two of its seven key development directions, which shows its importance. Today, intelligent video surveillance has been applied to a certain extent in China: the Qinghai-Tibet Railway was completed and accepted in October 2007, and 1,300 channels on the entire line use intelligent video analysis to protect the entire railway from intrusion; Beijing Subway Line 5, 23 stations on the entire line, key areas have adopted bomb dropping and intrusion detection; Beijing Aviation Information Center computer room, using intrusion and anti-tailing detection; Beijing Olympic Games, Shanghai World Expo, Guangzhou Asian Games venue construction projects, have all built and budgeted to use intelligent video analysis technology...

Conclusion

Some experts point out that after analog, digital and IP surveillance, the next wave of emerging technologies will be intelligent surveillance technology that includes intelligent monitoring and intelligent analysis. Intelligent video surveillance is an advanced solution that performs intelligent video analysis and fully automatic video surveillance. It can automatically track and identify objects, analyze motion and extract video information from analog or digital video streams. Intelligent video surveillance allows users to easily monitor and ensure the safety of various areas with security cameras, while also reducing labor costs and improving production efficiency.

At present, some professional video analysis research manufacturers are also working to improve existing core technologies, further enhance and optimize existing algorithm structures, and strive to pay more attention to users' actual needs in terms of functions, applications and prices in order to achieve the best system product form.

In order to make intelligent video technology more widely used, TI launched a basic video analysis library, named VLIB, which supports the development of unique intelligent video application algorithms based on the C64+x core. The essence of VLIB is a collection of DSP function kernels of more than 40 software. By calling these function libraries, customers can implement background modeling and background extraction algorithms, target feature extraction, tracking and recognition, and low-level pixel processing. Zheng Xiaolong told reporters that if the development of video analysis products is regarded as the construction of an intelligent building, the VLIB function library is like building materials. Developers no longer need to start from scratch and only need to focus on specific application implementation.

According to Zhang Yongfu, general manager of Shenzhen Yongyan Technology Development Co., Ltd., in recent years, three methods have emerged in the video analysis market: single algorithm, object tracking algorithm, and multi-object behavior detection algorithm. Single algorithm: It can identify and capture moving objects that appear in the monitoring screen. Although this method has some effect, it cannot be used as a complete solution, and it cannot be used when there are many people in the picture or complex behaviors. Object tracking algorithm: By using a graphical map, the camera can capture moving objects, which are monitored completely secretly. Although this method has made some progress, it cannot identify some behaviors that can lead to violations, and this method of tracking objects requires overlapping camera images. Multi-object behavior detection: This is through a neural network algorithm that can identify multiple behaviors of multiple objects. It can be used as a complete security solution and can be effectively applied to the monitoring system to detect various violations in real time.