CN102929903A - Rapid video retrieval method based on layered structuralized description of video information - Google Patents

Abstract

The present invention provides a fast video retrieval method based on the hierarchical and structured description of video information, which is characterized in that: the comprehensive application framework and the research and judgment process of the hierarchical and structured description of video content based on three hierarchical databases and three engines, said The comprehensive application architecture is: In the video surveillance network system, according to the actual situation of each sub-network, multiple nodes are deployed hierarchically, each node takes the video layered structured description server as the core, and includes three complete databases and three Engine, the advantage of the present invention is that through the multi-level, multi-node three-database three-engine collaborative work, the surveillance video content is fully automatic hierarchically structured description, and logical reasoning and judgment are performed based on the topological relationship between nodes, realizing intelligent research and judgment Video surveillance information realizes automatic alarm and quick retrieval.

Description

A Fast Video Retrieval Method Based on Hierarchical and Structured Description of Video Information

technical field

The invention relates to a fast video retrieval method based on the layered and structured description of video information, which belongs to the field of security video monitoring.

Background technique

With the development of the economy and society, and the continuous improvement of people's demand for security, video surveillance systems have been rapidly developed and popularized in the past few decades. From the original single key security industries such as government agencies, road traffic, Financial institutions, etc., are gradually applied to more and more public facilities such as office buildings, shopping malls, residential quarters, etc.

Social security monitoring produces a large amount of video information. However, most of the video information has nothing to do with solving the case. Useful information is mixed with a large amount of useless information, which prolongs the search cycle and delays the time to solve the case. Manpower alone cannot distinguish between the actions of suspects or ordinary people in the mass video surveillance information, and give early warning of suspect behavior. Therefore, a dedicated video information analysis and query system is required to extract police attention information from these videos and perform video analysis. Layer-by-layer structured description, providing video screening services for public security departments to lock targets and provide early warnings in daily monitoring, or to retrieve targets, fix evidence, and discover clues during investigations.

The realization of such a video information analysis and query system usually needs to meet some basic conditions: first, the system must deploy multiple cameras to form a monitoring network; second, the system must automatically analyze the content of the monitoring video, extract and store useful information, and be able to Automatic alarm; Third, the system can provide the function of quick query and in-depth analysis of these useful information.

In the prior art, a monitoring network is deployed through an Ethernet or a wireless network. In the existing technology, in order to realize the automatic video description and analysis system, in order to extract the police attention information from these videos and realize the video monitoring fast retrieval and automatic alarm, the video monitoring information stored and obtained in real time can be analyzed at the camera terminal For processing, the data can also be transmitted to the monitoring center for centralized processing. Techniques used in video analysis include object detection, object tracking, scene analysis, etc. The results obtained from the analysis can be used for automatic alarm, and can also be stored for later query.

For example, patent announcement CN101783881A discloses a kind of intelligent network camera with video structure description function, which is aimed at the image visual features such as object color, texture, shape, motion, positioning and contour features in the camera to perform its attribute and content analysis. Reasoning and judgment, and writing description language through Extensible Markup Language (XML), Binary Extensible Markup Language (Binary XML). This invention mainly uses a video structured description device to reason and judge the characteristics and attributes of objects in a camera.

In the patent announcement CN101790064A, a hard disk video recording device with a video structured description function is disclosed. The invention uses a feature extraction unit to structurally describe multi-channel monitoring video image information, generate description information data about its attributes and contents, and realize intelligent monitoring and alarm functions.

In the patent announcement CN101827266A, a network video server is disclosed, which includes video acquisition, analysis and encoding output modules, which can use the attributes and contents of video features to judge and reason, and realize the encoding of small data volumes, so as to facilitate network transmission and video quick retrieval. This invention mainly uses a video analysis and description module to perform specific analysis on multi-channel video signals in the network to realize the structured description of the monitoring video.

However, in the above-mentioned prior art, even if part of the automatic analysis or automatic alarm function is realized, it is limited to separately analyzing the data collected by each camera terminal to form structured description information for query; The data of the camera terminal is comprehensively analyzed, and the quick query and cross-retrieval of video content in the case of multiple video surveillance sub-networks and massive video information are not considered. In the face of the needs of police ultra-large-scale networked video surveillance systems or inter-provincial and cross-regional networked video surveillance systems, the existing technology is unable to meet the performance requirements of such ultra-large-scale systems.

In addition, according to the video monitoring function module or network server of the prior art, it mainly faces a simple monitoring environment or a simple video monitoring network, and the analysis and retrieval of video information are limited to objects in a single video source/a few video sources For reasoning, judging, analyzing and storing, there is no mention of the method of realizing cross-network intelligent video monitoring through information interaction between sub-network nodes and logical reasoning through sub-network node topology between multiple video surveillance sub-networks. There is also no mention of a hierarchical and structured description of video content across multiple sub-networks and a distributed fast query method for video content. In the face of complex multiple sub-network integration systems, the above-mentioned prior art should not affect the respective video surveillance and alarm functions of the units, and at the same time satisfy the centralized video surveillance intelligent discrimination and alarm function, the permission setting will be very difficult, and the system The efficiency is not high, and there will be limitations on the requirement of being able to flexibly expand or disconnect a certain node. Therefore, the above-mentioned patents still cannot meet the needs of police large-scale cross-regional level video surveillance intelligent analysis system.

Contents of the invention

In order to solve the above technical problems, meet the needs of in-depth application of large-scale cross-regional video network monitoring system for police, quickly search for targets in massive video information, and intelligently detect suspects according to target characteristics in different regions or even across cities and regions The present invention provides a fast video retrieval method based on the layered and structured description of video information to meet the needs of networked tracking, intelligent arming and automatic alarm linkage in key monitoring areas.

A fast video retrieval method based on the layered structured description of video information of the present invention is based on the comprehensive application framework and the research and judgment process of the hierarchically structured description of video content based on three hierarchical databases and three engines. The comprehensive application framework is: In the video surveillance network system, according to the actual situation of each sub-network, multiple nodes are deployed hierarchically. Each node takes the video layered structured description server (Metadata server) as the core, and contains three complete databases and three engine, the three databases are video basic resource database, video layered structured description database (i.e. Metadata database) and monitoring blacklist database; the three engines are video layered structured description extraction engine (i.e. Metadata extraction engine ), an alarm research and judgment engine, and a video content retrieval engine; the research and judgment process is: through the multi-level, multi-node three databases and the three engines working together, a fully automatic hierarchical and structured description of the monitoring video content is carried out, and based on the node Logical reasoning and judgment of the topological relationship between them, realizing intelligent research and judgment of video surveillance information, and finally realizing automatic alarm and fast retrieval.

Metadata servers, three databases and three engines are set on each node of the video monitoring network system, so that it has functions such as independent video monitoring, automatic alarm, and user retrieval. The three databases may be centralized databases or distributed databases.

Each node of the video surveillance network system includes a dedicated video layered structured description server (i.e. Metadata server), which is used for inter-node interconnection and synchronous logic analysis results to realize simplified and complete police attention information Metadata data Subscription, push, storage, and query functions.

The fast video retrieval method of the present invention utilizes information such as the topological relationship of cameras in the basic video resource database to realize fast query by means of information correlation among different terminals when querying.

The fast video retrieval method of the present invention needs to follow the standardized format of the hierarchically structured description of the video content when performing hierarchically structured description of the video content.

The specific steps of the fast video retrieval method described in the present invention are as follows: establish a video basic resource database, collect and store in time the construction, update, changes and related basic information of various alarm and monitoring systems in various places; monitor video from different devices The video extraction work is carried out through the Metadata extraction engine, and the basic features of the video and the police attention information are extracted; these hierarchical content description information are stored in the storage medium in a standardized form, and are indexed uniformly; the basic video resources are stored in In the video basic resource database, the hierarchical description content is stored in the Metadata database; at the same time, the alarm research and judgment engine compares the generated video content hierarchical description information with the data in the key personnel, key vehicles, and suspicious object image databases. Yes, or compared with various suspect behavior interpretation principles and alarm limits based on the experience of public security systems at all levels, when a suspicious target is found, the alarm information is automatically output; when the user retrieves, the user enters the time and place through the client , Suspect features, suspicious vehicle features, suspicious item features and other retrieval conditions, the video content retrieval engine first extracts the features from the query conditions, and then combines the video basic resource information, through the Metadata server through distributed access to the sub-nodes at all levels The Metadata database performs parallel search, obtains the retrieval results and returns relevant videos to the user.

A fast video retrieval system based on the hierarchical and structured description of video information according to the present invention is characterized in that: a comprehensive application framework and a research and judgment process based on hierarchical and structured description of video content with three databases and three engines, the comprehensive application framework is : In the video surveillance network system, according to the actual situation of each sub-network, multiple nodes are deployed hierarchically, each node takes the video layered structured description server (ie Metadata server) as the core, and contains three complete databases and three three engines, the three databases are video basic resource database, video hierarchical structured description database (i.e. Metadata database) and monitoring blacklist database; the three engines are video hierarchical structured description extraction engine (i.e. Metadata extraction engine), an alarm research and judgment engine, and a video content retrieval engine; the research and judgment process is: through the multi-level, multi-node three databases and the three engines working together, a fully automatic hierarchical and structured description of the monitoring video content is carried out, and based on Perform logical reasoning and judgment on the topological relationship between nodes, realize intelligent research and judgment of video surveillance information, and finally realize automatic alarm and fast retrieval; each node in the video surveillance network system contains a dedicated network for inter-node interconnection and synchronous logic analysis results A video layered and structured description server (metadata server) to realize simplified and complete subscription, push, storage, and query functions of police attention information Metadata data.

A fast video retrieval system based on the layered and structured description of video information of the present invention, when querying, relies on the mutual correlation of information between different terminals, fully utilizes information such as the camera topology in the video basic resource database, and adopts a method based on The search method of logical reasoning realizes fast query.

The advantage of the present invention is that it designs and defines the layered and structured description content and format of the video content oriented to security prevention and criminal investigation and solving cases in combination with the needs of the front line of public security, and the video is analyzed from the two levels of basic video features and police attention features. Describe, and use the two levels to automatically retrieve and judge the video signal for police attention information, use its own three databases and three engine technologies to realize the functions of police attention information extraction, transmission, storage, retrieval, etc. In this way, a large-scale cross-provincial and cross-regional police video automatic retrieval and automatic alarm system can be realized, and theoretically it can be expanded indefinitely, and new terminals and new storage systems can be continuously connected. The new storage system continues to expand the logical reasoning ability and the scope of automatic alarms. Through the multi-level, multi-node three-database and three-engine collaborative work, it can fully automatically describe the surveillance video content in a hierarchical structure, and perform logical reasoning based on the topological relationship between nodes. Judgment, realizing intelligent research and judgment of video surveillance information, automatic alarm and quick retrieval.

Description of drawings

Fig. 1 shows the network structure diagram of three multi-level databases and three engines of the present invention;

Fig. 2 shows a schematic diagram of the network structure of three databases and three engines in a single node of the present invention.

Detailed ways

A kind of fast video retrieval method based on video information layered structured description of the present invention is described in detail below in conjunction with accompanying drawing; Multilevel three databases and three engine networks of the present invention, its principle is shown in Figure 1 shown.

In the multi-level three databases and three engines based video content hierarchically structured description comprehensive application architecture of the present invention, the multi-level three databases and three engines refer to the network including multiple nodes In , there are three database and three engine nodes deployed at a higher level relative to their child nodes. Through the three databases and three engines in the higher-level node, the resources in the three databases and three engines of all its child nodes can be scheduled. For example, a third-level three databases and three engine nodes can be deployed in each police station, these sub-node resources can be integrated in the sub-bureau to form a second-level three databases and three engine nodes, and the sub-node resources can be integrated in the city bureau to form A first-level three databases and three engine nodes. The sub-nodes at each level not only include independent Metadata servers, databases and engines for independent retrieval functions, but also call the resources of each sub-node and provide services to the upper level through the Metadata server. Sub-nodes at each level have all control systems that can work independently, and provide a unified interface to the upper and lower levels.

Metadata server and three databases and three engines are all set on each node of described video surveillance network system, make it have functions such as independent video surveillance, automatic alarm, user retrieval; Described three databases, both can be A centralized database can also be a distributed database.

In order to meet the performance requirements and flexible expansion requirements of a large-scale cross-regional networked video surveillance system, the present invention includes a dedicated video layered structured description server (Metadata server 1) in each node for inter-node interconnection and synchronous logic analysis As a result, streamlined and complete police attention information Metadata data subscription, push, storage, and query functions are realized.

Preferably, in the method provided by the present invention, by adopting a distributed database, the rapidly developing network processing capability can be utilized to transmit relatively small amounts of Metadata data for retrieval, and to locate relevant surveillance videos by means of logical reasoning, so that The time to retrieve information from a large number of surveillance video assets is reduced from days to hours. Moreover, due to the self-advantages of the distributed database, it is not necessary to adopt a specific video surveillance system for a certain design scheme in the prior art, but can be easily integrated with different video surveillance systems, or integrated with existing The integration of video surveillance systems only requires them to be supplemented with standardized and modularized Metadata servers, three databases and three engines. And because the grid in the distributed database is dynamic in nature, the resources in the grid can appear dynamically, allowing users to add resources to the grid or delete them from the grid as needed. After a node is deleted from the networked video surveillance system, the node has a Metadata server, three databases and three engines, so it still has independent video surveillance, automatic alarm, user retrieval and other functions, but it does not have the same functions as adjacent nodes. The associated research and judgment function and logical reasoning function, and the networked system after deleting this node still has all the related research and judgment functions and logical reasoning functions. In addition, the distributed database provides enhanced scalability. By transmitting streamlined and complete logically related video information, the system is not limited by physical proximity or network delay, avoiding the defects of centralized databases, and in this way , can flexibly realize easy interconnection between nodes, easy separation, convenient authority division, and fast cross-regional information transmission and video information query.

A schematic diagram of a dedicated Metadata server, three databases and three engines, three databases and three engines in a single node according to the present invention, and its principle is shown in FIG. 2 .

FIG. 2 also shows a comprehensive application architecture for police attention information retrieval based on the fast video retrieval method of the present invention. The comprehensive application architecture is that each node is based on the video layered structured description server (Metadata server 1) as the core, and contains three complete databases and three engines, and the three databases are basic video resources Database 4, video hierarchical structured description database (Metadata database 2) and monitoring blacklist database 3; the three engines are video hierarchical structured description extraction engine (Metadata extraction engine 5), alarm research and judgment engine 6 and video content search engine7.

The fast video retrieval method described in the present invention, when querying, relies on the mutual correlation of information between different terminals, makes full use of information such as the camera topological relationship in the video basic resource database 4, and adopts a logical reasoning-based search method to achieve Quick Search.

The hierarchical structured description of video information refers to the use of video intelligent analysis technology to generate textual structures and annotations from unstructured video data through different levels of abstraction and understanding of its content, that is, video hierarchical structure Descriptive data (i.e. metadata or Metadata data described below). The transmission and storage equipment resources required by these Metadata data are far less than the video data after compression and encoding, but it contains the main police attention information in the video. By indexing these Metadata data, advanced video retrieval and data mining applications can be realized. In the method provided by the present invention, in order to realize the data interaction between each video surveillance sub-network and the cross-subnet comprehensive reasoning application, it is first necessary to establish a standardized format for the hierarchical and structured description of the video content.

The search method based on logical reasoning refers to the use of hierarchically structured description data of video content to generate a logical reasoning evaluation result, i.e. logical data . Compile and index these logical data and compare them with various suspect behavior interpretation principles and alarm limits formulated based on the experience of public security systems at all levels to realize intelligent research and judgment of video surveillance information, and finally realize automatic alarm and fast retrieval .

The video layered structured description server, i.e. Metadata server 1, is the core of each sub-node, used for inter-node interconnection and synchronous logical analysis results, and realizes the subscription, push, and Storage and query functions.

The Metadata database 2 is used to store the hierarchically structured video description data corresponding to the video captured by each camera in the sub-node (that is, the Metadata data).

The monitoring blacklist database 3 stores a blacklist of personnel to be armed, facial features, clothing features, gait features, vehicle shape features, vehicle color features, license plate features, mass incidents, personnel wandering, vehicle retrograde, etc. Police attention event description, etc.

The video basic resource database 4 is used to store basic information such as the device model, installation location, working status, and camera topology diagram of each camera in the monitoring sub-node. The camera topology diagram describes the adjacent cameras of each camera and the necessary route and minimum time required for the suspect/vehicle to move from one camera to another camera's shooting range. The shortest time is determined by the camera installation location, installation angle, traffic mode, road conditions, etc.

The Metadata extraction engine 5 has a function of performing real-time data analysis on multi-channel surveillance video, extracting standard video layered structured description information, and storing it in the Metadata database 2 . The Metadata extraction engine is a general term for all devices or modules that can extract hierarchical and structured descriptions from videos: it can be one device or multiple devices; it can be a front-end embedded device, a central server or even a Cluster computing resources; it can be a stand-alone device or embedded in a camera or DVR; it can be connected to analog video or compressed digital video. The Metadata extraction engine can not only extract the captured video stream in real time, but also extract the stored video files.

The alarm research and judgment engine 6, by analyzing the Metadata data stream in real time, deploys defenses against the information in the monitoring blacklist database 3, and sends an alarm immediately when it detects that it meets the blacklist database and the suspect/car or event occurs .

The video content retrieval engine 7, according to the user's intelligent video retrieval or tracking request, generates Metadata retrieval computer instructions, searches the Metadata database 2, and returns to the user video clips containing objects matching the retrieval conditions.

The camera network refers to multiple cameras in which adjacent camera terminals are logically associated and communicate with each other.

In the method provided by the present invention, in order to realize video information retrieval based on logical reasoning between nodes, first, it is necessary to record the camera model, installation location, and working status in the monitoring network to the video basic resource database 4, and, through on-site On-the-spot investigation and experiment, record the neighboring cameras of each camera, as well as the necessary route and the shortest time required for the suspect/vehicle to move from one camera to another camera’s shooting range, and finally form a camera topology map and record it to the basic video resource database4.

Second, deploy various video intelligent analysis devices to form a Metadata extraction engine5. The Metadata extraction engine 5 can automatically extract conventional targets and their features (such as target shape, color, texture, outline, etc.) and police attention information from the video content captured by the associated node camera/video capture device 8 (such as suspects, suspicious vehicles, suspicious objects, abnormal events, etc.), and store them in the Metadata database 2 after being indexed. For videos stored in the video storage device 9 that have not been analyzed in real time and extracted Metadata data, the Metadata extraction engine 5 can be temporarily instructed to extract video recordings when retrieval is required, and the results are also stored in the Metadata database 2 .

When the user performs intelligent deployment and control, according to the blacklist and deployment strategy input by the user through the software interface, as well as the monitoring blacklist and deployment strategy assigned by the superior node, the monitoring blacklist database 3 is formed and dynamically updated. At the same time, in the alarm research and judgment engine 6, through the real-time analysis of the Metadata data stream, the information in the monitoring blacklist database 3 is deployed, or it is combined with various suspect behavior interpretation principles based on the experience of public security systems at all levels. When a suspect/vehicle that meets the blacklist, a suspicious suspect/vehicle or an emergency that exceeds the police preset alarm limit is detected, an alarm will be issued immediately. When there are lower-level nodes, the alarm research and judgment engine of this node can also assign the monitoring blacklist and arming strategy to the corresponding lower-level nodes for arming.

When the user performs intelligent video retrieval, the user specifies the retrieval time range, area range, target feature, behavior rules, event type, etc. in the video content retrieval engine 7, and the Metadata retrieval computer is generated by the video content retrieval engine 7. Instructions are used to search the Metadata database 2, and return to the user video clips containing objects matching the search conditions. When there are subordinate nodes, the video content retrieval engine 7 of this node can also assign the retrieval task to the subordinate nodes in the corresponding area.

When the user conducts intelligent network tracking of the suspect/car, with the help of the information correlation between different cameras, combined with the camera network node topology diagram, the search method based on logical reasoning is adopted to realize the comprehensive reasoning of the whereabouts of the suspect/car , get the most matching video-related information, or get the time and place where the suspect/car is most likely to be found again, and help the police automatically obtain relevant videos.

Through the geographically related information between each node and the adjacent system nodes, as well as the police field investigation and survey, a large amount of relationship information about the distance and time required for this node to the adjacent system nodes can be derived logically. The time required for a suspicious vehicle or a certain suspect to go to an adjacent system node through a certain mode of transportation, and from this it can be deduced that it may be necessary to call the priority of the adjacent node video information. For a suspicious vehicle or a remote system node that cannot be reached by a certain suspect, the priority of video information retrieval is reduced to zero, so there is no need for further retrieval or retrieval of video information on the network, thus saving police retrieval and The time to call the video ensures that the large-scale networked video surveillance system and automatic alarm system will not be restricted by system performance or network performance.

Preferably, in order to meet the requirements of large-scale cross-regional networked video surveillance system interconnection and comprehensive logical reasoning of adjacent nodes, the Metadata data stored in the Metadata database 2 in each node must be stored in the same format and the same indexing method , to ensure that the basic video data in adjacent nodes can be retrieved by the same logical retrieval command.

Preferably, the Metadata data described in the present invention consists of the following two parts of data:

1. Basic features of video

Video basic features are low-level information used to describe objects in a video. Include the following information:

1) Moving target detection information: including time, target number, target position, target size, speed;

2) Moving target tracking information (trajectory);

3) Object color histogram: The color histogram describes the distribution of different colors on the object. The dominant color feature describes the main colors that make up the object. Such as a green car, its main color is green;

4) Discrete cosine transform (DCT) coefficient: the coefficient obtained by performing frequency domain transformation on the target image, this feature effectively describes the texture characteristics of the target;

5) Edge direction feature: the vector obtained by performing histogram statistics on the object edge according to the direction, which is used to effectively distinguish objects with different shapes.

2. Police attention information

1) Target category information;

2) Face and license plate recognition information;

3) Event detection output information such as leftover detection, item removal detection, running detection, intrusion detection, retrograde detection, loitering detection, etc. The specific information includes: current time, event number, event type, event level, event status;

4) The output information of people flow statistics includes: the current time, the number of people passing through the section;

5) The output information of crowd density estimation includes: current time, current crowd density level;

6) Video quality inspection information.

The above two parts of information together constitute Metadata data, which is immediately extracted by the Metadata extraction engine and stored in the Metadata database.

In the hierarchical description process of video information in the present invention, preferably, the surveillance video source may be the video acquisition device 8 as shown in FIG. 2 , or a video file stored in the video storage device 9 .

In summary, according to the embodiment of the present invention, its specific process is: establish a video basic resource database 4, and collect and store in time the construction, update, changes and related basic information of various alarm and monitoring systems in various places; Surveillance videos from different devices use the Metadata extraction engine 5 to extract the basic features of the video and police attention information; these hierarchical content description information are stored in the storage medium in a standardized form and indexed uniformly; the basic video resources are Stored in the video basic resource database 4, the hierarchical description content is stored in the Metadata database 2; meanwhile, the alarm research engine 6 combines the above-mentioned generated video content hierarchical description information with key personnel, key vehicles, and suspicious object image libraries Compare the data in the database, or compare with various suspect behavior interpretation principles and alarm limits based on the experience of public security systems at all levels, and when suspicious targets are found, the alarm information will be automatically output; Input time, location, suspect features, suspicious vehicle features, suspicious item features and other retrieval conditions at the terminal, the video content retrieval engine 7 first extracts the features from the query conditions, and then combines the video basic resource information, through the Metadata server through distributed access 1 Parallel search is performed on the Metadata database 2 of sub-nodes at all levels, the retrieval results are obtained and relevant videos are returned to the user.

The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, more engines or more engines can be added on the basis of the present invention. More databases, thereby realizing more complete and fast video retrieval, these modifications and improvements are also considered as the protection scope of the present invention, and any changes that can be easily imagined by those skilled in the art within the scope of the disclosure of the present invention Or replacement, all should be covered in the scope of protection of the claims of the present invention.

Claims (6)

1. fast video search method of describing based on the video information hierarchical structured, it is characterized in that: describe the integrated application framework and study and judge flow process based on the video content hierarchical structured of three databases of classification and three engines, described integrated application framework is: in video monitoring network system, according to each sub-network actual conditions, a plurality of nodes are disposed in classification, each node is take video segmentation structural description server as core, and include complete three databases and three engines, described video segmentation structural description server is the Metadata server, described three databases are the video elementary resource database, video segmentation structural description database and monitoring blacklist database, described video segmentation structural description database is the Metadata database; Described three engines are that engine and video content search engine are studied and judged in video segmentation structural description extraction engine, warning, and described video segmentation structural description extraction engine is the Metadata extraction engine; The described flow process of studying and judging is: by three databases and three engine collaborative works multi-level, multinode, the monitor video content is carried out full-automatic hierarchical structured to be described, and carry out reasoning from logic based on topological relation between node and judge, video monitoring information is studied and judged in the realization intellectuality, finally realizes automatic alarm and quick-searching.

2. a kind of fast video search method of describing based on the video information hierarchical structured according to claim 1, it is characterized in that: Metadata server and three databases and three engines all are set on each node of described video monitoring network system, make it possess independently video monitoring, automatic alarm, user search function, described three databases are one of centralized data base or distributed data base.

3. a kind of fast video search method of describing based on the video information hierarchical structured according to claim 1, it is characterized in that: each node at described video monitoring network system all comprises special-purpose video segmentation structural description server, it is the Metadata server, be used for interconnected, synchronous logic analysis result between node, realize simplifying and subscription, propelling movement, storage, the query function of complete police concern information Metadata data.

4. a kind of fast video search method of describing based on the video information hierarchical structured according to claim 1, it is characterized in that: described fast video search method is when inquiring about, interrelated by means of the information between the different terminals, take full advantage of the information such as video camera topological relation in the video elementary resource database, adopt the searching method of logic-based reasoning to realize fast query.

5. a kind of fast video search method of describing based on the video information hierarchical structured according to claim 1, it is characterized in that: described fast video search method is being carried out hierarchical structured when describing to video content, need follow the standardized format that the video content hierarchical structured is described.

6. a kind of fast video search method of describing based on the video information hierarchical structured according to claim 1, it is characterized in that: the concrete steps of described fast video search method are: set up the video elementary resource database, construction, renewal, change conditions and the relevant rudimentary information of all kinds of warnings in various places and supervisory system is in time gathered warehouse-in; Monitor video from distinct device carries out video extraction work by the Metadata extraction engine, extracts video essential characteristic and police concern information; These stratification content description informations are kept in the storage medium with normalized form, and the unified index of setting up; Described video elementary resource is stored in the video elementary resource database, and described stratification is described content and is stored in the Metadata database; Simultaneously, warning is studied and judged engine the data in the video content stratification descriptor of above-mentioned generation and emphasis personnel, emphasis vehicle, the suspicious object image library is compared, or compare with various suspect's behavior interpretation principles and alarm threshold that the experience of solving a case based on public security systems at different levels is formulated, when finding suspicious object, automatic output alarm information; When user search, the user is by search conditions such as client input time, place, suspect's feature, suspect vehicle feature, suspicious object features, the video content search engine at first extracts feature from querying condition, then in conjunction with the video elementary resource information, mode by distributed access is carried out parallel search through the Metadata server to the Metadata database of child nodes at different levels, obtains result for retrieval and associated video is returned to the user.

Images