XUCHUAN CHEMICAL(SUZHOU) CO., LTD , https://www.xuchuanpu.com
Analysis of "Video Structure" in Public Security
I. Current status of public safety
The construction of the safe city originated from two major projects: the "Strengthening Science and Technology Strengthened Police" strategy and the construction of the city's alarm and monitoring system, the "3111" pilot project. Since 2004, after more than ten years of construction, the video storage scale in Pingan City has exceeded the EB level. These huge amounts of video data are increasingly playing an irreplaceable role in social public safety management and case detection. Under the background of the in-depth development of public safety information construction, the existing video systems have the outstanding problem of lack of in-depth application mode and low level of intelligence in video data. How to use new technologies to transform existing video systems so that they can better adapt to the Internet of Things in the era of video intelligence, intelligence application requirements is imminent. The current major problems are: the lack of standardized methods for generating video information intelligence, and the lack of new policing work models that use video information intelligence to guide investigations and solve crimes; and video informationization and policing applications lack uniform standards and norms.
The root of all these issues is the lack of understanding of video content, and there is no efficient, standardized method of video data exchange and video intelligence extraction. To meet the challenge of deep application of video data, the core and bottleneck is to solve the transformation from general video data to video informatization and video informatization by researching the video structured description technology, and to realize the innovation of the social public safety working model.
Second, the video structure description
Video structured description is a technology based on the extraction of video content information. It is based on the semantic relationship of video content, using spatial and temporal segmentation, feature extraction, object recognition and other processing means, organized into a structure of information for computers and human understanding technology . From the perspective of data processing, video structured description technology can transform unstructured video data into structured or semi-structured intelligence information that can be understood by humans and machines, and further transforms it into intelligence data used in public security operations to realize video. The data is transformed into the information and information-based direction, reaching the smart applications of the video-aware world. The video structured description is not only an effective technology for the informatization and information transformation of massive video, but also a directional solution to the video structuring in the current public safety field. In terms of the content of the structured description of video, public security focuses on video information mainly: people, vehicles, and behavior. In the video, the person is presented as a descriptive individual, which includes the precise positioning of the person's face, the extraction of facial features, the comparison of facial features, the gender, age range, approximate height of the person, hair accessories, clothing, articles carried, and walking A variety of configurable descriptive information such as form; description information for vehicles includes vehicle description information such as license plate, vehicle color, vehicle model, brand, sub-brand, car stickers, and car accessories information; description information for the behavior includes: cross-border , regional, defamation, legacy, aggregation and other behavior description information. After the video structured processing, the following objectives can be achieved: First, the video search speed is greatly improved. After the video is structured, find a suspected pedestrian target on a screenshot from a million-level target library (corresponding to hundreds of thousands of hours of high-definition video) and complete it in seconds; It can be completed in minutes (if cloudization is achieved, it will be faster). Search queries on a structured basis can solve the problem of fast target finding. Second, the storage capacity is greatly reduced. After the structured video, the structured search information and target data of the storage person are less than 2% of the video data volume; for the vehicle, less than 1%; and the behavior is reduced more. The storage capacity is greatly reduced, which can solve the problem of long-term video storage. Finally, video structuring can revitalize video data and can be used as the basis for data mining. After structured processing, the video is stored in a corresponding structured data warehouse. In-depth data mining can be performed on various types of data warehouses to make full use of big data. Enhance the application value of video data and improve the analysis and prediction functions of video data.
Third, the application of video structuring
"Public safety" is the name given to the safety of the people. It refers to the safety of the people. It does not refer to the safety of a certain person, nor the security of a particular group (such as a robbery gang), but refers to the safety of ordinary people around them. Public security "people-oriented" aims to protect national security and social stability. With the heightened development of human material civilization, the country, society, and individuals have reached an unprecedented height of reliance on safety and expectations. Public safety involves many fields: public safety includes information security, food safety, public health and safety, and public travel laws. Safety, safety of asylum behavior, site safety for evacuation, construction safety, urban lifeline safety, malicious and non-malicious personal safety and evacuation, etc. Public safety incidents include natural disasters, accidents and disasters, public health incidents, and social security incidents. As an important source of visual perception of the Internet of Things, video data plays an increasingly important role in the field of public security. The video structured description is an in-depth application of unstructured video data, making the video data a perceptible and descriptive intelligent data. Therefore, its application fields are extremely extensive. In terms of public safety, the video structured description almost penetrates into all aspects of public safety.
Fourth, video intelligence analysis
The video structured description is an intelligent structured analysis of video content, and the unstructured video data is intelligently analyzed to form descriptive structured data. Therefore, video intelligent analysis is the core technology of video structuring. The quality of intelligent video analysis technology has great influence on the video structured description. In order to better analyze the video structure, intelligent video analysis must be innovative in the following three directions: First, video pre-processing technology, including image protection Jitter and image enhancement. The main cause of video jitter is the high frequency of small jitter caused by overhead installation in road monitoring. Video stabilization can effectively suppress false alarms and false reports in intelligent analysis and improve the accuracy of intelligent analysis. Image enhancement is the source of video. Improve the visual quality, improve the image quality, improve the sharpness of the image, and make the original low-quality image clear and legible. Second, improve the accuracy of analysis technology. For example, the face recognition technology transitions from the initial feature face method to the neural network method, from visible light face recognition to multi-source light face recognition. Similarly, smart analysis of vehicles and behaviors has also led to more efficient analysis techniques. Third, video post-processing technology, including image restoration and image abstract retrieval. Image restoration is the comprehensive use of super-resolution, de-blur filtering, distortion correction, color adjustment, etc. to process the blurred video so that it is clearly identifiable.
V. Video structuring needs to break through the bottleneck
Massive video image data is an important data accumulated in the public security department's information construction. Through analysis and processing of video content, effective clues can be quickly and accurately discovered, and the role of video resources can be fully utilized. However, under the background of the in-depth development of information construction, existing video resources lack the mode of deep application. The bottleneck of its application is how to effectively extract video information, how to exchange standard data with other information systems, interconnection and interoperability. The core technology to solve this problem is the technology of video structured description. However, as a core technology of video processing, there are still some bottlenecks in its implementation that need to be broken. First, it is a breakthrough in the core technology of video structuring. As mentioned above, video structured description technology and video intelligence analysis technology are closely related, but current video intelligence analysis technology is subject to various application environment constraints. Take face recognition as an example, current face recognition is mostly cooperative and repeatable. In the application scenario, in this application scenario, the recognition rate of the human face can basically meet the practical requirements, and it is difficult to achieve a practical goal in a scene without coordination, multiple faces, and dynamic video.
In order to solve such problems, the face recognition algorithm also evolved from the initial pattern recognition to a deep neuron network learning pattern, which greatly improved the accuracy of face detection and recognition, but the negative effects were also considerable. Obviously, the first to go is the increase in computational complexity and the need to spend a lot of computing resources. Many industry companies have introduced various solutions to this bottleneck, such as the calculation of the front-end progress, and the use of GPUs to achieve clustered computing. All these The directions are to look forward to the practical application of the subsequent video structured description. The realization of efficient and accurate video structured description technology becomes the direction of efforts of various algorithm research institutions in the future. The second is the video structure description data storage, retrieval and application technology. With the rapid growth of data capacity, structured video descriptions increasingly have 4V characteristics of “big dataâ€. How to implement large-capacity, highly-efficient storage, efficient retrieval of video structured data, and rapid implementation of data applications to provide end-users with efficient and flexible services will all become major problems faced by major video integrated application manufacturers. Again the top-level design builds a standard system. Through the research on the characteristics and application modes of video structuring technology, a standard system model for video structured description is established, a standardized system for the implementation of coverage technologies and application systems is formulated, and related standards are formulated step by step to standardize technical research and device development. It guides all aspects of system construction, operation, and evaluation, and lays the foundation for the comprehensive development of the application of video information intelligence at the source.
In this regard, the national standard GB/T30147-2013 “Technical Requirements for Real-time Intelligent Analysis Equipment for Security Monitoring Video†and GB/T30148-2013 “Safety Alarms†are centralized and organized by the National Security Alarm System Standardization Technical Committee (SAC/TC100). The electromagnetic compatibility immunity requirements and test methods for equipment are published by the National Standards Management Committee and implemented on August 1, 2014. The formulation of these standards establishes a common observance of rules and codes of conduct for video structured descriptions, making it possible for large-scale video structured applications.
Finally, the application of large platforms for video structured data applications. With the maturity of video structuring technology, how to collect and manage these huge amounts of video structured description data, how to provide public security departments with fast, efficient, professional and personalized services It is also a difficult problem for service providers and owners. For example, many manufacturers use the "cloud" and "big data" framework models as the platform architecture of the video structured description technology to meet the large amount of data storage, life cycle management, and rapid data response.
Six, video structure description technology development prospects
The video structured description technology closely fits the analysis and extraction of video content and processes the unstructured video data into informatized data for quick retrieval and positioning. With the development and maturity of video structured description technology, it will inevitably provide strong support for the informatization, intelligence, and intelligence of video data. The passive defense of video changes will be actively identified, and the corresponding system will become intelligent and active. The control system becomes possible, so as to open up the transmission of system-to-system video intelligence and open up the application from front-end acquisition to back-end intelligence. With the construction of video systems in recent years, a huge amount of video data already exists in the society and still producing a large amount of video data in real time, all of which provide a broad market prospect for video structured description technology.