CN109033121A - A kind of video frequency search system and its method based on cloud storage - Google Patents
A kind of video frequency search system and its method based on cloud storage Download PDFInfo
- Publication number
- CN109033121A CN109033121A CN201810524297.XA CN201810524297A CN109033121A CN 109033121 A CN109033121 A CN 109033121A CN 201810524297 A CN201810524297 A CN 201810524297A CN 109033121 A CN109033121 A CN 109033121A
- Authority
- CN
- China
- Prior art keywords
- video
- feature vector
- model data
- image
- cloud storage
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of video retrieval method based on cloud storage, comprising the following steps: the video segment information that need to be retrieved is carried out interception and obtains several picture samples；Picture sample is converted to iconic model data；The iconic model data of conversion are carried out with the foundation of feature vector；The feature vector of the corresponding original image model data of several video segment informations stored in the feature vector of foundation and cloud storage is subjected to registration detection, obtains registration；Filter out the video-frequency band that registration is greater than registration threshold values；The video segment information of screening is sequentially output corresponding video-frequency band according to the size of Superposition degree modulus.The present invention is by intercepting the video-frequency band that need to be retrieved, converting and match index, the video segment information that registration is greater than registration threshold values can be filtered out, substantially increase the video frequency searching efficiency in massive video data, the accuracy for improving retrieval simultaneously has far-reaching influence in field of video retrieval.
Description
Technical field
The invention belongs to technical field of video monitoring, it is related to a kind of video frequency search system based on cloud storage and its side
Method.
Background technique
Video cloud storage is the important component in video monitoring, accounts for the Large videos such as city video monitoring monitoring network
In core status.
Video cloud storage mainly manages cluster by cloud storage and video storage cluster forms, and cloud storage manages cluster and video
Storage cluster passes through network connection.Wherein, video storage cluster is made of multiple video memory nodes, each video memory node
The video data from video camera is stored, video storage cluster is in video cloud storage system by the equipment of storage video data
Composed cluster；Cloud storage management cluster is made of multiple management nodes, and each management node is used for video memory node
It is managed and externally (such as user platform) provides service, cloud storage management cluster is external in video cloud storage system
Cluster composed by the equipment of service and management video memory node is provided.
It is all that video record is carried out by the camera on thousands of roads in typical video monitoring system, it is every to take the photograph all the way
As head is all stored with more days video datas, the video data of these magnanimity is stored in video cloud storage system, and video cloud is deposited
While massive video data in storage provides information resources abundant for user, also brought very to the inquiry of video data
Big challenge.In public safety field, being investigated into a case by video investigation technology has been the technology being in daily use, in face of the video of magnanimity
How data, even trans-regional data, user efficiently get desired information, become an apparent problem.
Currently, the mode for searching information from the multitude of video data stored in cloud storage substantially manually plays back
Video is completed, inefficiency, and manpower waste is serious, and it is existing efficiency checked to improve video by video retrieval technology,
But it is generally only semantic retrieval, some features are difficult to be described with semanteme, there is a problem of that recall precision is low, simultaneously
The low problem of the accuracy wasted time during retrieval, retrieved.
Summary of the invention
The purpose of the present invention is to provide a kind of video frequency search system and its method based on cloud storage, by needing to retrieve
Video-frequency band carry out interception and obtain picture sample, and iconic model data are converted to image pattern, by iconic model number
According to feature vector is established, the video segment information in cloud storage can be fast and accurately filtered out, and then solves existing video inspection
There are problems that video frequency searching low efficiency and accuracy rate during rope.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of video retrieval method based on cloud storage, comprising the following steps:
The video segment information that S1, reception need to be retrieved, video segment information is intercepted, and obtains several picture samples, institute
State several picture samples constitute set A (B1, B2 ..., Bi ..., Bm), m is expressed as picture sample in the video segment information
Quantity, Bi indicate i-th of picture sample；
S2, the picture sample of acquisition is modeled, is converted to iconic model data；
S3, the foundation that the iconic model data of conversion are carried out with feature vector, the feature vector of described image model data
Including essential characteristic vector sum specific characteristic vector, the specific characteristic vector is that every image is divided into n*n histogram
Picture, be successively set as the 1st feature vector, the 2nd feature vector ..., the n-th * n feature vector, several specific characteristic vectors constitute set
Bi (b i 1, b i2 ..., b i j ..., b i t*t), b i j is expressed as corresponding j-th of the feature of the Bi image pattern
Vector；
S4, by the corresponding original image model data of several video segment informations stored in the feature vector of foundation and cloud storage
Feature vector carry out registration detection, obtain registration；
The registration threshold value of S5, the registration that will test and setting compare, and filter out registration and are greater than coincidence bottom valve
The video-frequency band of value；
S6, to the video segment information screened in step S5 according to the size of Superposition degree modulus, be sequentially output corresponding video
Section.
Further, different weight parameters is set in the step S1 to different image patterns, constitutes weight set
The corresponding weighted value of i-th of image pattern, and g1+g2+...+gi+...+ are indicated for G (g1, g2 ..., gi ..., gm), gi
Gm=1.
Further, the video information in the step S4 in cloud storage includes several video-frequency bands, constitutes video-frequency band set W
(w1, w2 ..., wh), each video-frequency band include several image informations, and different image informations corresponds to different iconic model numbers
According to image collection is C (C1, C2 ..., Ck ..., Cn) in the Cloud Server, and n indicates the quantity (n > m) of image, and every
A image corresponds to different feature vectors, feature vector in each image constitute collection be combined into Dk (dk1, dk2 ...,
Dkj ..., dkt*t), dkj is expressed as corresponding j-th of the feature vector of the Dk image；
The 1st feature vector in image pattern is compared with f-th of feature vector in original image model data, structure
At being overlapped vector set Bi ' (bi ' 1, bi ' 2 ..., bi ' f ..., bi ' n*n), wherein f-th of feature in image pattern to
Amount is 1 with the value that f-th of feature vector in original image model data coincides, and is 0, bi ' f indicates the Bi image otherwise
The value that the f feature vector in f feature vector and original image model data in sample coincides；
According to registration calculation formula, calculates the video segment information that need to be retrieved and several video-frequency bands stored in cloud storage are believed
Breath carries out registration,Wherein, Q indicates Superposition degree modulus,
A kind of video frequency search system based on cloud storage, including video-frequency band interception module, modeling conversion module and matching rope
Draw module, the video-frequency band interception module is connect with modeling conversion module, and modeling conversion module is connect with match index module；
The video-frequency band interception module cuts received video segment information for receiving the video segment information that need to be retrieved
It takes, obtains several image patterns, and image pattern is sent to modeling conversion module；
The modeling conversion module receives the picture sample that video-frequency band interception module is sent, and models to image pattern,
Iconic model data are converted to, and the iconic model data of conversion are carried out with the foundation of feature vector, the spy of iconic model data
Levying vector includes essential characteristic vector sum specific characteristic vector；
The match index module receives the feature vector for the iconic model data that modeling conversion module is sent, and by feature
The feature vector of the corresponding original image model data of several video segment informations stored in vector and cloud storage carries out registration inspection
The registration of several video-frequency bands for surveying, and then obtaining the video-frequency band that need to be retrieved and storing in cloud storage, filters out registration and is greater than
The video-frequency band of the registration threshold value of setting, and corresponding video-frequency band is sequentially output according to Superposition degree modulus size order.
Further, the data stored in the cloud storage are the iconic model data of modeled conversion module conversion,
And the iconic model data stored in cloud storage are original image model data.
Beneficial effects of the present invention:
The present invention is converted to iconic model data by carrying out interception to received video-frequency band, and extracts iconic model data
In feature vector, the corresponding feature vector of the video segment information stored in the feature vector of extraction and cloud storage is overlapped
Degree detection, and then the video segment information that registration is met the requirements successively is filtered out, it substantially increases in massive video data
Video frequency searching efficiency, while the accuracy of retrieval is improved, avoid time-consuming existing for manual retrieval and semantic retrieval more, accurately
The low problem of rate has far-reaching influence in field of video retrieval.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, will be described below to embodiment required
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ability
For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is a kind of schematic diagram of the video frequency search system based on cloud storage of the present invention；
Fig. 2 is a kind of flow chart of the video retrieval method based on cloud storage of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other
Embodiment shall fall within the protection scope of the present invention.
Refering to Figure 1, the present invention is a kind of video frequency search system based on cloud storage, including video-frequency band intercepts mould
Block, modeling conversion module and match index module, the video-frequency band interception module are connect with modeling conversion module, model modulus of conversion
Block is connect with match index module；
The video-frequency band interception module cuts received video segment information for receiving the video segment information that need to be retrieved
It takes, obtains several image patterns, and image pattern is sent to modeling conversion module；
The modeling conversion module receives the picture sample that video-frequency band interception module is sent, and models to image pattern,
Iconic model data are converted to, and feature vector, the feature of described image model data are established to the iconic model data of conversion
Vector includes essential characteristic vector sum specific characteristic vector；
The match index module receives the feature vector for the iconic model data that modeling conversion module is sent, and by feature
The feature vector of the corresponding original image model data of several video segment informations stored in vector and cloud storage carries out registration inspection
The registration of several video-frequency bands for surveying, and then obtaining the video-frequency band that need to be retrieved and storing in cloud storage, filters out registration and is greater than
The video-frequency band of the registration threshold value of setting, and corresponding video-frequency band is sequentially output according to Superposition degree modulus size order.
Wherein, the data stored in cloud storage are the iconic model data of modeled conversion module conversion, and cloud storage
The iconic model data of middle storage are original image model data.
As shown in Fig. 2, a kind of video frequency search system method based on cloud storage, comprising the following steps:
The video segment information that S1, reception need to be retrieved, video segment information is intercepted, and obtains several picture samples, institute
State several picture samples constitute set A (B1, B2 ..., Bi ..., Bm), m is expressed as picture sample in the video segment information
Quantity, Bi indicates i-th of picture sample, and different weight parameters is set to different image patterns, constitutes weight sets and is combined into G
(g1, g2 ..., gi ..., gm), m are expressed as the quantity of picture sample in the video segment information, and gi indicates i-th of image pattern
Corresponding weighted value, the g1+g2+...+gi+...+gm=1；
S2, the picture sample of acquisition is modeled, is converted to iconic model data；
S3, the foundation that the iconic model data of conversion are carried out with feature vector, wherein the feature of described image model data
Vector includes essential characteristic vector sum specific characteristic vector, and the essential characteristic vector includes time parameter, image volume, image
Color and color category etc., the specific characteristic vector is that every image is divided into n*n rectangular image, and according to from upper
To lower and from left to right sequence, be successively set as the 1st feature vector, the 2nd feature vector ..., the n-th * n feature vector, Ruo Gante
It levies vector and constitutes set Bi (bi 1, bi2 ..., bi j ..., bit*t), it is corresponding that bi j is expressed as the Bi image pattern
J-th of feature vector, described eigenvector are color, the distributing position etc. of image；
S4, by the corresponding original image model data of several video segment informations stored in the feature vector of foundation and cloud storage
Feature vector carry out registration detection, wherein the video information in cloud storage includes several video-frequency bands, constitute video-frequency band set
W (w1, w2 ..., wh), each video-frequency band include several image informations, and different image informations corresponds to different iconic model numbers
According to image collection is C (C1, C2 ..., Ck ..., Cn) in the Cloud Server, and n indicates the quantity (n > m) of image, and every
A image corresponds to different feature vectors, feature vector in each image constitute collection be combined into Dk (dk1, dk2 ...,
Dkj ..., dkt*t), dkj is expressed as corresponding j-th of the feature vector of the Dk image；
The 1st feature vector in image pattern is compared with f-th of feature vector in original image model data, structure
At being overlapped vector set Bi ' (bi ' 1, bi ' 2 ..., bi ' f ..., bi ' n*n), wherein f-th of feature in image pattern to
Amount is 1 with the value that f-th of feature vector in original image model data coincides, and is 0, bi ' f indicates the Bi image otherwise
The value that the f feature vector in f feature vector and original image model data in sample coincides；
According to registration calculation formula, calculates the video segment information that need to be retrieved and several video-frequency bands stored in cloud storage are believed
Breath carries out registration,Wherein, Q indicates Superposition degree modulus,
The registration threshold value of S5, the registration that will test and setting compare, and filter out registration and are greater than coincidence bottom valve
The video-frequency band of value；
S6, to the video segment information screened in step S5 according to the size of Superposition degree modulus, be sequentially output corresponding video
Section.
The present invention is converted to iconic model data by carrying out interception to received video-frequency band, and extracts iconic model data
In feature vector, the corresponding feature vector of the video segment information stored in the feature vector of extraction and cloud storage is overlapped
Degree detection, and then the video segment information that registration is met the requirements successively is filtered out, it substantially increases in massive video data
Video frequency searching efficiency, while the accuracy of retrieval is improved, avoid time-consuming existing for manual retrieval and semantic retrieval more, accurately
The low problem of rate has far-reaching influence in field of video retrieval.
The above content is just an example and description of the concept of the present invention, affiliated those skilled in the art
It makes various modifications or additions to the described embodiments or is substituted in a similar manner, without departing from invention
Design or beyond the scope defined by this claim, be within the scope of protection of the invention.
Claims (5)
1. a kind of video retrieval method based on cloud storage, which comprises the following steps:
The video segment information that S1, reception need to be retrieved, video segment information is intercepted, and obtains several picture samples, if described
Dry picture sample constitutes set A (B1, B2 ..., Bi ..., Bm), and m is expressed as the quantity of picture sample in the video segment information,
Bi indicates i-th of picture sample；
S2, the picture sample of acquisition is modeled, is converted to iconic model data；
S3, the foundation that the iconic model data of conversion are carried out with feature vector, the feature vector of described image model data include
Essential characteristic vector sum specific characteristic vector, the specific characteristic vector are that every image is divided into n*n rectangular image, according to
It is secondary be set as the 1st feature vector, the 2nd feature vector ..., the n-th * n feature vector, several specific characteristic vectors constitute set Bi
(bi1, bi2 ..., bij ..., bit*t), bij are expressed as corresponding j-th of the feature vector of the Bi image pattern；
S4, by the spy of the corresponding original image model data of several video segment informations stored in the feature vector of foundation and cloud storage
It levies vector and carries out registration detection, obtain registration；
The registration threshold value of S5, the registration that will test and setting compare, and filter out registration greater than registration threshold values
Video-frequency band；
S6, to the video segment information screened in step S5 according to the size of Superposition degree modulus, be sequentially output corresponding video-frequency band.
2. a kind of video retrieval method based on cloud storage according to claim 1, it is characterised in that: in the step S1
Set different weight parameters to different image patterns, constitute weight sets be combined into G (g1, g2 ..., gi ..., gm), gi table
Show the corresponding weighted value of i-th of image pattern, and g1+g2+...+gi+...+gm=1.
3. a kind of video retrieval method based on cloud storage according to claim 1, it is characterised in that: in the step S4
Video information in cloud storage includes several video-frequency bands, is constituted video-frequency band set W (w1, w2 ..., wh), and each video-frequency band includes
Several image informations, different image informations correspond to different iconic model data, and image collection is C in the Cloud Server
(C1, C2 ..., Ck ..., Cn), n indicates the quantity (n > m) of image, and each image corresponds to different feature vectors, each
Feature vector in image constitutes collection and is combined into Dk (dk1, dk2 ..., dkj ..., dkt*t), and dkj is expressed as the Dk image pair
J-th of the feature vector answered；
The 1st feature vector in image pattern is compared with f-th of feature vector in original image model data, constitutes weight
Resultant vector set Bi ' (bi ' 1, bi ' 2 ..., bi ' f ..., bi ' n*n), wherein f-th feature vector in image pattern with
The value that f-th of feature vector in original image model data coincides is 1, is 0, bi ' f indicates the Bi image pattern otherwise
In f feature vector and original image model data in the value that coincides of f feature vector；
According to registration calculation formula, calculate the video segment information that need to be retrieved and several video segment informations stored in cloud storage into
Row registration,Wherein, Q indicates Superposition degree modulus,
4. a kind of video frequency search system based on cloud storage according to claim 1, it is characterised in that: cut including video-frequency band
Modulus block, modeling conversion module and match index module, the video-frequency band interception module are connect with modeling conversion module, and modeling turns
Mold changing block is connect with match index module；
The video-frequency band interception module is used to receive the video segment information that need to be retrieved, and intercepts to received video segment information,
Several image patterns are obtained, and image pattern is sent to modeling conversion module；
The modeling conversion module receives the picture sample that video-frequency band interception module is sent, and models, converts to image pattern
For iconic model data, and the iconic model data of conversion are carried out with the foundation of feature vector, the features of iconic model data to
Amount includes essential characteristic vector sum specific characteristic vector；
The match index module receives the feature vector for the iconic model data that modeling conversion module is sent, and by feature vector
The feature vector of original image model data corresponding with several video segment informations stored in cloud storage carries out registration detection, into
And the registration of several video-frequency bands for obtaining the video-frequency band that need to be retrieved and being stored in cloud storage, it filters out registration and is greater than setting
The video-frequency band of registration threshold value, and corresponding video-frequency band is sequentially output according to Superposition degree modulus size order.
5. a kind of video frequency search system based on cloud storage according to claim 4, it is characterised in that: in the cloud storage
The data of storage are the iconic model data of modeled conversion module conversion, and the iconic model data stored in cloud storage are
Original image model data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810524297.XA CN109033121A (en) | 2018-05-28 | 2018-05-28 | A kind of video frequency search system and its method based on cloud storage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810524297.XA CN109033121A (en) | 2018-05-28 | 2018-05-28 | A kind of video frequency search system and its method based on cloud storage |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109033121A true CN109033121A (en) | 2018-12-18 |
Family
ID=64611597
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810524297.XA Pending CN109033121A (en) | 2018-05-28 | 2018-05-28 | A kind of video frequency search system and its method based on cloud storage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109033121A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108965906A (en) * | 2018-05-28 | 2018-12-07 | 安徽维德工业自动化有限公司 | A kind of video storage method based on cloud storage service device |
| CN113886629A (en) * | 2021-12-09 | 2022-01-04 | 深圳行动派成长科技有限公司 | Course picture retrieval model establishing method |
-
2018
- 2018-05-28 CN CN201810524297.XA patent/CN109033121A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108965906A (en) * | 2018-05-28 | 2018-12-07 | 安徽维德工业自动化有限公司 | A kind of video storage method based on cloud storage service device |
| CN108965906B (en) * | 2018-05-28 | 2020-12-29 | 安徽维德工业自动化有限公司 | Video storage method based on cloud storage server |
| CN113886629A (en) * | 2021-12-09 | 2022-01-04 | 深圳行动派成长科技有限公司 | Course picture retrieval model establishing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104392338B (en) | Building Fire Protection examination ＆ verification Acceptance Test System, method, audit terminal and mobile examination terminal | |
| CN108446293A (en) | A method of based on urban multi-source isomeric data structure city portrait | |
| CN105389375B (en) | A kind of image index setting method, system and search method based on visible range | |
| CN114092769B (en) | Transformer substation multi-scene inspection analysis method based on federal learning | |
| CN110598532B (en) | Tree pest and disease damage monitoring system and method | |
| CN107896160A (en) | A kind of data center network flowmeter factor method based on distributed system | |
| CN109033121A (en) | A kind of video frequency search system and its method based on cloud storage | |
| CN106503223A (en) | A kind of binding site and the online source of houses searching method and device of key word information | |
| CN113792667A (en) | Method and device for automatically classifying properties of buildings in villages and towns based on three-dimensional remote sensing image | |
| CN109284760A (en) | A kind of furniture detection method and device based on depth convolutional neural networks | |
| CN112749663A (en) | Agricultural fruit maturity detection system based on Internet of things and CCNN model | |
| CN115033707A (en) | Power transmission equipment portrait knowledge map construction method based on big data analysis technology | |
| CN113626437B (en) | Method and system for rapidly inquiring mass vector data | |
| CN107276807B (en) | Hierarchical network community tree pruning method based on community dynamic compactness | |
| CN106372262A (en) | System and method for urban outdoor public space urban home furnishing management | |
| CN104703059B (en) | A kind of method and device for planning of broadband access network | |
| CN112528833A (en) | Plant sample prescription investigation method, device and system | |
| CN111563193B (en) | Medical insurance fixed-point management system based on information technology | |
| CN107492085A (en) | Stereo image quality evaluation method based on dual-tree complex wavelet transform | |
| CN115858713A (en) | Data management method and data management system for agricultural condition geographic information | |
| CN110442749A (en) | Video frame processing method and processing device | |
| CN115713603A (en) | Multi-type block building group form intelligent generation method based on building space map | |
| CN108629521A (en) | A kind of mapping Planning preparation system for building system suitable for rustic home | |
| CN115147835A (en) | Pineapple maturity detection method in natural orchard scene based on improved RetinaNet | |
| CN106503199A (en) | A kind of network machine information retrieval system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |