CN114037841A - Personnel area collision early warning model based on feature clustering - Google Patents
Personnel area collision early warning model based on feature clustering Download PDFInfo
- Publication number
- CN114037841A CN114037841A CN202111294962.9A CN202111294962A CN114037841A CN 114037841 A CN114037841 A CN 114037841A CN 202111294962 A CN202111294962 A CN 202111294962A CN 114037841 A CN114037841 A CN 114037841A
- Authority
- CN
- China
- Prior art keywords
- server
- clustering
- feature clustering
- snapshot
- interface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a personnel area collision early warning model based on feature clustering, which comprises a snapshot machine, a video analysis server, a feature clustering server, a column-type database server, an interface server and an application program, wherein the snapshot machine is in communication connection with the video analysis server, the video analysis server is in communication connection with the feature clustering server, the feature clustering server is in communication connection with an actual-name human face library, the feature clustering server is also in communication connection with the column-type database server, the column-type database server is in communication connection with the interface server, and the interface server is in communication connection with the application program; the clustering archive is stored by adopting the open-source column type database, and the clustering archive is more suitable for the field of data analysis than a general relational database and a full-text retrieval engine, and meets various complex operation requirements such as data aggregation and duplicate removal based on the archive.
Description
Technical Field
The invention relates to the technical field of video image pattern recognition and application, in particular to a personnel region collision early warning model based on feature clustering.
Background
Chinese patent CN109977863A discloses a campus periphery dangerous individual detection system based on monitoring, and relates to a method for detecting, tracking and identifying dangerous individuals in monitored images of multiple monitors, the system identifies dangerous behaviors around a campus which harm students, and considers that the detection rate of the dangerous behaviors subjected to disguise is low by simply carrying out video identification, so that the system adds movement path track identification on the basis of traditional video image mode identification, and improves the detection rate of the system on the disguised dangerous behaviors in a mode of combining video identification and movement path identification;
with the enhancement of security consciousness, a large number of face snapshot cameras are built in a place, according to statistics, more than 1500 million pictures are snapshot by a fertile snow project face snapshot camera every day, how to extract key information from the face pictures in time, quickly and efficiently and meet security protection requirements becomes a hotspot in snow project construction, a traditional processing mode is that a known face picture is used for snapshot library image search, or known face pictures are distributed and controlled through a blacklist for early warning, the situation can be further researched and judged after passively receiving the early warning or needing known clues, and the situation detection and screening effect cannot be achieved.
Disclosure of Invention
The invention aims to solve the problems of the background technology and provides a personnel region collision early warning model based on feature clustering.
The purpose of the invention can be realized by the following technical scheme:
the personnel area collision early warning model based on feature clustering comprises a snapshot machine, a video analysis server, a feature clustering server, a column type database server, an interface server and an application program, wherein the snapshot machine is in communication connection with the video analysis server, the video analysis server is in communication connection with the feature clustering server, the feature clustering server is in communication connection with a real-name human face library, the feature clustering server is also in communication connection with the column type database server, the column type database server is in communication connection with the interface server, and the interface server is in communication connection with the application program.
As a further scheme of the invention: the snapshot machine pushes the captured face picture to a video analysis server, and the video analysis server is responsible for extracting the features of the face picture.
As a further scheme of the invention: and the video analysis server pushes the extracted massive face feature information to a feature clustering server, and the feature clustering server carries out clustering archiving in batches according to the similarity.
As a further scheme of the invention: the clustering archiving comprises the following steps:
the face snapshot record takes the face object characteristic attribute in the view library as a data structure basis, and the video analysis server analyzes the face characteristic value and fills an extension field;
and (3) calculating vector distance through a density clustering algorithm, merging the similar points with the similar points, extracting and generating class center features and unique identification to form personnel file records, and writing back the file unique identification to the face snapshot record.
As a further scheme of the invention: the snapshot record comprises snapshot time, snapshot equipment, a snapshot size graph, an identity tag and a file identifier, and the snapshot record is stored in the column database.
As a further scheme of the invention: the interface server constructs a region collision analysis interface and provides a region collision data interface for an upper application program; the outgoing parameter is a person profile which appears under the corresponding camera within the corresponding time period and meets the repeated occurrence times.
As a further scheme of the invention: the interface server constructs a region collision analysis interface, and the incoming parameters are aggregated and deduplicated in the column database to obtain outgoing parameters.
As a further scheme of the invention: incoming parameters of the interface server include a plurality of sets of cameras, a plurality of sets of time periods, and a number of repetitions.
As a further scheme of the invention: the outgoing parameter of the interface server is the personnel file which appears under the corresponding camera in the corresponding time period and meets the repeated occurrence times.
The invention has the beneficial effects that: the density-based clustering algorithm is adopted, compared with the common mean clustering algorithm, the method is only suitable for the convex data set, and the density clustering algorithm can cluster the dense data set with any shape; the clustering archive is stored by adopting the open-source array type database, and the clustering archive is more suitable for the field of data analysis (OLAP) than a general relational database and a full-text retrieval engine, and meets various complex operation requirements such as data aggregation and duplicate removal based on the archive; a personnel space-time region collision model is constructed, repeated personnel information of a preset region and a preset time period can be rapidly pushed, and a research and judgment means is provided for case investigation of a public security organization; therefore, under the condition that no direct clue exists, the data collision of the time information and the space information can quickly reduce the mass face acquisition data in the time of the cases in a plurality of areas to hundreds or even several suspects, so that the screening target can be quickly reduced, the case detection time can be shortened, and the powerful guarantee is provided for the life and property safety of people.
Drawings
The invention will be further described with reference to the accompanying drawings.
FIG. 1 is a data flow diagram of the present invention;
FIG. 2 is a network topology of the present invention;
FIG. 3 is a flow chart of an operator of the region collision model according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1 and 2, the invention is a personnel area collision early warning model based on feature clustering, which includes a snapshot machine, a video analysis server, a feature clustering server, a column database server, an interface server and an application program;
the snapshot machine is in communication connection with the video analysis server, the face snapshot machine at the front end of the video network pushes a captured face picture to the video analysis server, and the video analysis server is responsible for extracting features of the face picture;
the video analysis server is in communication connection with the feature clustering server, the video analysis server pushes extracted massive face feature information to the feature clustering server, and the feature clustering server conducts clustering archiving in batches according to the similarity;
the method for clustering and archiving the massive face feature information according to the similarity comprises the following steps:
the face snapshot record is recorded according to the view library standard GAT 1400.3-2017 public security video image information application system part 3: the characteristic attribute of the face object in the database technical requirement is a data structure basis, and the video analysis server analyzes the face characteristic value and then fills an extension field;
calculating vector distance through a density clustering algorithm, combining the classes with close distance, extracting and generating class center features and unique identification to form personnel file records, and writing back the file unique identification to a face snapshot record;
the feature clustering server is in communication connection with the real-name face library, the person files which are filed are clustered by the feature clustering server to form class center features, the images of the permanent population with the real-name information are subjected to feature extraction to construct a real-name feature library, the class center features in the file records are compared with the real-name feature library, and after comparison, the real-name identity numbers are backfilled into the file records to form confidence files; the real-name face library comprises a standing population library and a real population library;
the characteristic clustering server is also in communication connection with the column-type database server, and the snapshot records after the characteristic clustering server clusters and archives are stored in the column-type database server; the snapshot record is transmitted to a columnar database in a storage field mode, and the storage field comprises snapshot time, snapshot equipment, a snapshot size graph, an identity mark and a file mark; the snapshot records are stored in a column-type database, so that aggregation and duplicate removal are facilitated;
the column-type database server is in communication connection with the interface server, the interface server is in communication connection with the application program, and the interface server constructs a region collision analysis interface and provides a region collision data interface for the upper application program; the incoming parameters comprise a plurality of camera sets, a plurality of time period sets and repeated occurrence times, and the outgoing parameters are personnel files which appear under the corresponding camera in the corresponding time period and meet the repeated occurrence times;
as shown in fig. 3, the incoming parameters of the region collision model include multiple camera sets, multiple time period sets and repeated occurrence times, the outgoing parameters are personnel files which appear under corresponding cameras in corresponding time periods and meet repeated occurrence times, the face snapshot records are stored in a column database, the key points are snapshot time and a snapshot camera, the snapshot records are aggregated through two fields of the snapshot time and the snapshot camera, it is considered that personnel may appear and snapshot in the same region for multiple times in the same time period, therefore, data in the same region of the same personnel in the same time period need to be deduplicated, the aggregation records of the personnel files meeting the incoming parameters are formed, filtering is performed through the repeated occurrence times, and finally, the region collision results are output.
Example 2
The working method of the personnel area collision early warning model based on the feature clustering comprises the following steps:
step 1: the front-end face snapshot machine of the video network pushes a captured face picture to a video analysis server, and the video analysis server is responsible for extracting the features of the face picture;
step 2: pushing the extracted massive face feature information to a feature clustering server, and clustering and archiving in batches according to the similarity;
and step 3: comparing the clustered personnel file with a picture library with real-name information;
and 4, step 4: the snapshot records are stored in a column database;
and 5: the interface server constructs an area collision interface and provides an area collision data interface for the upper application program.
The working principle of the invention is as follows: the method adopts a density-based clustering algorithm, is only suitable for a convex data set compared with a common mean clustering algorithm, and can cluster dense data sets in any shape by using the density clustering algorithm; the clustering archive is stored by adopting the open-source array type database, and the clustering archive is more suitable for the field of data analysis (OLAP) than a general relational database and a full-text retrieval engine, and meets various complex operation requirements such as data aggregation and duplicate removal based on the archive; a personnel space-time region collision model is established, repeated personnel information of a preset region and a preset time period can be rapidly pushed, and a research and judgment means is provided for case investigation of public security organs.
While one embodiment of the present invention has been described in detail, the description is only a preferred embodiment of the present invention and should not be taken as limiting the scope of the invention. All equivalent changes and modifications made within the scope of the present invention shall fall within the scope of the present invention.
Claims (7)
1. The personnel area collision early warning model based on feature clustering comprises a snapshot machine, a video analysis server, a feature clustering server, a column type database server, an interface server and an application program, and is characterized in that the snapshot machine, the video analysis server, the feature clustering server, the column type database server are in communication connection, and the interface server is in communication connection with the application program in sequence;
the video analysis server pushes the extracted massive face feature information to the feature clustering server, and the feature clustering server conducts clustering archiving in batches according to the similarity;
clustering is archived as: the face snapshot record takes the face object characteristic attribute in the view library as a data structure basis, and the video analysis server analyzes the face characteristic value and fills an extension field; and (3) calculating vector distance through a density clustering algorithm, merging the similar points with the similar points, extracting and generating class center features and unique identification to form personnel file records, and writing back the file unique identification to the face snapshot record.
2. The personnel area collision early warning model based on feature clustering as claimed in claim 1, wherein the snapshot machine pushes the captured face picture to a video resolution server, and the video resolution server is responsible for extracting features of the face picture.
3. The feature clustering-based personnel area collision early warning model as claimed in claim 1, wherein the snapshot records comprise snapshot time, snapshot equipment, a snapshot size graph, an identity tag and a file identifier, and are stored in a columnar database.
4. The personnel region collision early warning model based on feature clustering as claimed in claim 1, wherein the interface server constructs a region collision analysis interface to provide a region collision data interface for upper application programs; the outgoing parameter is a person profile which appears under the corresponding camera within the corresponding time period and meets the repeated occurrence times.
5. The personnel area collision early warning model based on feature clustering as claimed in claim 1, wherein the interface server constructs an area collision analysis interface, and the incoming parameters are aggregated and deduplicated in the column database to obtain outgoing parameters.
6. The feature clustering based personnel area collision warning model of claim 5, wherein incoming parameters of the interface server include multiple camera sets, multiple time period sets and number of repetitions.
7. The feature clustering-based personnel area collision warning model according to claim 6, wherein the outgoing parameters of the interface server are personnel profiles that appear under the corresponding camera within the corresponding time period and satisfy the number of repeated occurrences.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111294962.9A CN114037841A (en) | 2021-11-03 | 2021-11-03 | Personnel area collision early warning model based on feature clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111294962.9A CN114037841A (en) | 2021-11-03 | 2021-11-03 | Personnel area collision early warning model based on feature clustering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114037841A true CN114037841A (en) | 2022-02-11 |
Family
ID=80142747
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111294962.9A Pending CN114037841A (en) | 2021-11-03 | 2021-11-03 | Personnel area collision early warning model based on feature clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114037841A (en) |
-
2021
- 2021-11-03 CN CN202111294962.9A patent/CN114037841A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101778260B (en) | Method and system for monitoring and managing videos on basis of structured description | |
| CN109635146B (en) | Target query method and system based on image characteristics | |
| WO2020259099A1 (en) | Information processing method and device, and storage medium | |
| CN101692706B (en) | Intelligent storage equipment for security monitoring | |
| US20210357678A1 (en) | Information processing method and apparatus, and storage medium | |
| CN106355154B (en) | Method for detecting frequent passing of people in surveillance video | |
| EP2831842A2 (en) | An event triggered location based participatory surveillance | |
| SG174519A1 (en) | Systems and methods for detecting anomalies from data | |
| CN106295489B (en) | Information processing method, information processing device and video monitoring system | |
| Afra et al. | Early warning system: From face recognition by surveillance cameras to social media analysis to detecting suspicious people | |
| CN111078922A (en) | Information processing method and device and storage medium | |
| CN109492604A (en) | Faceform's characteristic statistics analysis system | |
| Khan et al. | Blockchain-enabled deep semantic video-to-video summarization for IoT devices | |
| CN110543584A (en) | method, device, processing server and storage medium for establishing face index | |
| CN111090777A (en) | Video data management method, management equipment and computer storage medium | |
| KR101170676B1 (en) | Face searching system and method based on face recognition | |
| CN114037841A (en) | Personnel area collision early warning model based on feature clustering | |
| CN110704660A (en) | Data processing method, device, equipment and computer storage medium | |
| Le et al. | Surveillance video retrieval: what we have already done? | |
| Priyanka et al. | Matrix decomposition based digital video forgery detection | |
| CN114817638A (en) | Backtracking system for fixed scene monitoring video | |
| Chaisorn et al. | Video analytics for surveillance camera networks | |
| CN112818863A (en) | Method and system for constructing personnel track based on face searching | |
| Cao et al. | Adaptive and robust feature selection for low bitrate mobile augmented reality applications | |
| Hannane et al. | An automatic video surveillance indexing based on facial feature descriptors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |