CN109862319A - A kind of video library processing system and method for supporting more AI algorithms - Google Patents
A kind of video library processing system and method for supporting more AI algorithms Download PDFInfo
- Publication number
- CN109862319A CN109862319A CN201910104118.1A CN201910104118A CN109862319A CN 109862319 A CN109862319 A CN 109862319A CN 201910104118 A CN201910104118 A CN 201910104118A CN 109862319 A CN109862319 A CN 109862319A
- Authority
- CN
- China
- Prior art keywords
- video
- transcoding
- algorithm
- layer
- algorithms
- 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
Abstract
The invention discloses a kind of video library processing systems and method for supporting more AI algorithms.The video library processing system includes: transcoding service layer, data resource layer, algorithm service layer, routine interface layer and dispatch service layer, by the transcoding service layer the original nonstandard video in the data resource layer is carried out transcoding, form transcoding normal video, then the dispatch service layer carries out United Dispatching to each algorithm service quotient in the algorithm service layer, to carry out analysis and comparison processing to the picture, primary standard video and/or transcoding normal video stored in the data resource layer.The video library processing system is that each algorithm service quotient saves the plenty of time, allows each algorithm service quotient to be absorbed in the research and development of AI algorithm, also reduces access threshold, so that industry competition is more abundant.
Description
Technical field
The present invention relates to video library processing technique more particularly to a kind of video library processing system for supporting more AI algorithms and sides
Method.
Background technique
With the maturation of video monitoring system construction and going deep into for informatization, security industry company and government's machine
It closes and needs to come by AI algorithm to carry out various videos the processing such as recognition of face, action recognition, trajectory track, behavior anticipation.Together
When, also occur numerous alternative algorithm service quotient in the market, each algorithm service quotient due to using different AI algorithms,
Therefore each have their own advantage, needs to realize that advantage is mutual using the AI algorithm of more algorithm service quotient simultaneously in some scenarios
It mends.
But since the monitoring camera of video monitoring system is by stages, is conducted batch-wise construction, buying it is very possible
It is the different product of different company, even the same product of same company can also issue different versions in different year
This causes there are the video of a variety of coded formats in existing video library, and especially some relatively early stages have just carried out monitoring
The area of camera construction, such as Shenzhen cause the coded format of video very more because historical time span is big.Each algorithm
Service provider needs first to carry out video decoding to video, decoding process is not before carrying out video analysis and comparison processing to video
Only heavy workload, and require each algorithm service quotient to have and the decoded ability of video is carried out to each coded format, still
And not all algorithm service commercial city is good at and carries out video decoding, and each algorithm service quotient will also be for different coding format
The API of video integration producer is developed and is debugged, and engineering construction progress is extended.
Summary of the invention
In order to solve above-mentioned the deficiencies in the prior art, the present invention provides a kind of video library processing system for supporting more AI algorithms
And method, a large amount of Capture Program Time is saved for each algorithm service quotient, each algorithm service quotient is absorbed in
The research and development of AI algorithm also reduce access threshold, so that industry competition is more abundant.
The technical problems to be solved by the invention are achieved by the following technical programs:
A kind of video library processing system for supporting more AI algorithms, comprising: transcoding service layer, data resource layer, algorithm service layer, journey
Sequence interface layer and dispatch service layer, wherein
The transcoding service layer is turned the original nonstandard video in the data resource layer for providing transcoding service
Code forms transcoding normal video, wherein the original nonstandard video is not specified standard coding format, the transcoding standard
Video is specified standard coding format；
The data resource layer, for storing data resource, including store for the algorithm service layer handled it is original
The video library of normal video and transcoding normal video, wherein the primary standard video is specified standard coding format；
The algorithm service layer, primary standard video and/or transcoding for being stored in the video library to the data resource layer
Normal video carries out analysis and comparison processing, including at least two algorithm service quotient；
Described program interface layer, each algorithm service quotient for accessing in the algorithm service layer, for the dispatch service
Layer carries out United Dispatching to each algorithm service quotient；
The dispatch service layer, for carrying out United Dispatching to each algorithm service quotient in the algorithm service layer.
Further, described program interface layer uses unified api interface standard, for each in the algorithm service layer
Family algorithm service quotient accesses.
Further, to the primary standard video and/or turned between each algorithm service quotient using different AI algorithms
Code normal video carries out analysis and comparison processing.
A kind of video library processing method for supporting more AI algorithms, step include:
Original nonstandard video in the video library is subjected to transcoding, is stored after forming transcoding normal video, wherein described
Original nonstandard video is not specified standard coding format, and the transcoding normal video is specified standard coding format；
Scheduling an at least algorithm service quotient in the video library primary standard video and/or transcoding normal video solve
Analysis and comparison processing, wherein the primary standard video is specified standard coding format.
Further, the original nonstandard video in the video library is subjected to transcoding, forms the transcoding normal video
Step includes:
The coded format of original video in the video library is detected, wherein the original video includes primary standard
Video and original nonstandard video；
If the original video is not specified standard coding format, it is determined as original nonstandard video；
Transcoding is carried out to the original nonstandard video, specified standard coding format is converted into, forms the transcoding standard
Video.
Further, to the primary standard video and/or turned between each algorithm service quotient using different AI algorithms
Code normal video carries out analysis and comparison processing.
Further, when dispatching the algorithm service quotient progress analysis and comparison processing, scheduling at least two algorithms clothes
The analysis and comparison that business quotient carries out Same Scene to same primary standard video and/or transcoding normal video are handled.
Further, when dispatching the algorithm service quotient progress analysis and comparison processing, according to each algorithm service quotient
Algorithm advantage, dispatch different algorithm service quotient and corresponding fields carried out to the primary standard video and/or transcoding normal video
The analysis and comparison of scape are handled.
Further, an at least algorithm service quotient is supported to regard the primary standard in the video library by FTP/HTTP
Frequency and/or transcoding normal video download in its local server.
Further, an at least algorithm service quotient is using multithreading in its local server and/or by altogether
Enjoy storage network connect with the video library reading the picture, primary standard video and/or transcoding normal video divided
File, segment processing.
Passing through the invention has the following beneficial effects: the video library processing system and method in advance will be in the video library
Original nonstandard Video Quality Metric is that transcoding normal video is stored, then again by picture, primary standard video and/or transcoding standard
Video is sent to each algorithm service quotient and individually carries out analysis and comparison processing, and each algorithm service quotient no longer needs to individually regard
Frequency decodes, and only each algorithm service quotient does not save a large amount of decoding time, and each algorithm service quotient is absorbed in
The research and development of AI algorithm, the industrial chain division of labor is more clear, more efficient, also reduces access threshold, can support some to be bad to regard
Frequently decoded new algorithm service provider quickly accesses, so that industry competition is more abundant.
Detailed description of the invention
Fig. 1 is the architecture principle figure of video library processing system provided by the invention；
Fig. 2 is the step schematic diagram of video library processing method provided by the invention.
Specific embodiment
The present invention will be described in detail with reference to the accompanying drawings and examples.
A kind of video library processing method for supporting more AI algorithms is applied in a video library processing system, and video library refers to
Video store comprising picture and one or more coded format videos, it is the same below.As shown in Figure 1, the video library processing system
System includes: transcoding service layer, data resource layer, algorithm service layer, routine interface layer and dispatch service layer, wherein
The transcoding service layer is turned the original nonstandard video in the data resource layer for providing transcoding service
Code forms transcoding normal video, wherein the original nonstandard video is not specified standard coding format, the transcoding standard
Video is specified standard coding format；
The data resource layer, for storing data resource, including store the picture handled for the algorithm service layer,
The video library of primary standard video, original nonstandard video and transcoding normal video, and store through at the algorithm service layer
The shared storage network of processing picture (big figure and small figure) after reason, wherein the primary standard video is that specified standard is compiled
Code format；
The algorithm service layer, picture, primary standard video for being stored in the video library to the data resource layer and/or
Transcoding normal video carries out analysis and comparison processing, including at least two algorithm service quotient, uses between each algorithm service quotient
Different AI algorithms；
Described program interface layer, each algorithm service quotient for accessing in the algorithm service layer, for the dispatch service
Layer carries out United Dispatching to each algorithm service quotient；
The dispatch service layer, for carrying out United Dispatching to each algorithm service quotient in the algorithm service layer.
As shown in Fig. 2, the video library processing method includes the following steps:
Original nonstandard video in the video library of the data resource layer is carried out transcoding by the transcoding service layer, forms transcoding mark
It is stored after quasi- video；
At least algorithm service quotient at the algorithm service is dispatched at the dispatch service to the figure in the video library
Piece, primary standard video and/or transcoding normal video carry out analysis and comparison processing.
It is transcoding by the original nonstandard Video Quality Metric in the video library that the video library processing system and method, which pass through in advance,
Normal video is stored, and picture, primary standard video and/or transcoding normal video are then sent to each algorithm service again
Quotient individually carries out analysis and comparison processing, and each algorithm service quotient no longer needs to individually carry out video decoding, not only each algorithm
Service provider saves a large amount of decoding time, and each algorithm service quotient is allowed to be absorbed in the research and development of AI algorithm, industrial chain point
Work is more clear, more efficient, also reduces access threshold, can support that some to be bad at the decoded new algorithm service provider of video fast
Quick access enters, so that industry competition is more abundant.
The primary standard video and original nonstandard video are the original video that monitoring camera shooting obtains, and have one kind
Or a variety of coded formats, wherein the original nonstandard video is first stored in the video library after being shot by the monitoring camera
In, after the transcoding service layer carries out transcoding, it can still be stored in the video library and give over to backup, it can also be from the view
Frequency is deleted in library, depending on specific requirements；The present embodiment is using MP4 format as specified standard coding format.
Wherein, the original nonstandard video in the video library is carried out transcoding by the transcoding service layer, forms the transcoding
The step of normal video includes:
The coded format of original video in the video library is detected, wherein the original video includes primary standard
Video and original nonstandard video；
If the original video is not specified standard coding format, it is determined as original nonstandard video；
Transcoding is carried out to the original nonstandard video, specified standard coding format is converted into, forms the transcoding standard
Video.
Wherein, if when judging the coded format of the original video, if the original video is specified standard code
Format is then determined as primary standard video, with no treatment.
Preferably, described program interface layer uses unified api interface standard, for each family in the algorithm service layer
Algorithm service quotient accesses.Each algorithm service quotient accesses the video library processing system by unified api interface standard, by
The dispatch service layer carries out United Dispatching, due to api interface standard be it is unified, system compatibility is bigger, in interface differential technique
Upper need to corresponding with the simple communication exchange of each algorithm service quotient progress and offer resource (such as computing resource and storage money
Source), each algorithm service quotient can first research and develop according to interface document, be connect with the program that exploitation meets interface standard
Mouthful, save the API exploitation and debug time of each algorithm service quotient.
In the present embodiment, described program interface layer include but is not limited to picture parsing, video parsing, vector depositary management reason, to
Buret reason, 1:N are compared, 1:1 is compared or scratch the api interface of at least one function programs such as figure.
When dispatching the algorithm service quotient progress analysis and comparison processing, schedulable at least two, the dispatch service layer
Algorithm service quotient carries out same picture, primary standard video and/or transcoding normal video at the analysis and comparison of Same Scene
Reason, such as recognition of face, the dispatch service layer can dispatch at least two algorithm service quotient to needing to carry out recognition of face
Same picture, primary standard video and/or transcoding normal video carry out the analysis and comparison processing of recognition of face, with lateral ratio
Processing capacity of more each algorithm service quotient to Same Scene；Alternatively, the dispatch service layer can also be according to each algorithm service
The algorithm advantage of quotient is dispatched different algorithm service quotient and is carried out to the picture, primary standard video and/or transcoding normal video
The analysis and comparison processing of corresponding scene, such as recognition of face and action recognition, the dispatch service layer, which can be dispatched, to be good at
An at least algorithm service quotient for recognition of face is to the picture, primary standard video and/or transcoding mark for needing to carry out recognition of face
Quasi- video carries out the analysis and comparison processing of recognition of face, and scheduling is good at an at least algorithm service quotient for action recognition to needs
The picture, primary standard video and/or transcoding normal video for carrying out action recognition carry out the analysis and comparison processing of action recognition,
To comprehensively utilize the algorithm advantage of each algorithm service quotient.
Each algorithm service quotient to picture, primary standard video and/or the transcoding normal video in the video library into
Corresponding processing picture (big figure and small figure) can be generated after row dissection process, and the processing picture is stored in described share and is deposited
It stores up in network.
Each algorithm service quotient is obtaining picture, primary standard video and/or transcoding normal video in the video library
When, support that at least an algorithm service quotient passes through FTP/HTTP for the picture, primary standard video and/or transcoding normal video
It downloads in its local server, then using multithreading in its local server and/or passing through the shared storage
Network connect the picture, primary standard video and/or transcoding normal video read with the video library and carries out single cent part, divides
Section processing, realizes high speed processing to efficiently use the processing capacity of local server.
The shared storage network can be, but not limited to store network for NAS.
Embodiments of the present invention above described embodiment only expresses, the description thereof is more specific and detailed, but can not
Therefore limitations on the scope of the patent of the present invention are interpreted as, as long as skill obtained in the form of equivalent substitutions or equivalent transformations
Art scheme should all be fallen within the scope and spirit of the invention.
Claims (10)
1. a kind of video library processing system for supporting more AI algorithms characterized by comprising transcoding service layer, data resource layer,
Algorithm service layer, routine interface layer and dispatch service layer, wherein
The transcoding service layer is turned the original nonstandard video in the data resource layer for providing transcoding service
Code forms transcoding normal video, wherein the original nonstandard video is not specified standard coding format, the transcoding standard
Video is specified standard coding format；
The data resource layer, for storing data resource, including store for the algorithm service layer handled it is original
The video library of normal video and transcoding normal video, wherein the primary standard video is specified standard coding format；
The algorithm service layer, primary standard video and/or transcoding for being stored in the video library to the data resource layer
Normal video carries out analysis and comparison processing, including at least two algorithm service quotient；
Described program interface layer, each algorithm service quotient for accessing in the algorithm service layer, for the dispatch service
Layer carries out United Dispatching to each algorithm service quotient；
The dispatch service layer, for carrying out United Dispatching to each algorithm service quotient in the algorithm service layer.
2. the video library processing system according to claim 1 for supporting more AI algorithms, which is characterized in that described program interface
Layer uses unified api interface standard, so that each algorithm service quotient in the algorithm service layer accesses.
3. the video library processing system according to claim 1 or 2 for supporting more AI algorithms, which is characterized in that each algorithm
The primary standard video and/or transcoding normal video are carried out at analysis and comparison using different AI algorithms between service provider
Reason.
4. a kind of video library processing method for supporting more AI algorithms, which is characterized in that step includes:
Original nonstandard video in the video library is subjected to transcoding, is stored after forming transcoding normal video, wherein described
Original nonstandard video is not specified standard coding format, and the transcoding normal video is specified standard coding format；
Scheduling an at least algorithm service quotient in the video library primary standard video and/or transcoding normal video solve
Analysis and comparison processing, wherein the primary standard video is specified standard coding format.
5. the video library processing method according to claim 4 for supporting more AI algorithms, which is characterized in that by the video library
In original nonstandard video the step of carrying out transcoding, forming the transcoding normal video include:
The coded format of original video in the video library is detected, wherein the original video includes primary standard
Video and original nonstandard video；
If the original video is not specified standard coding format, it is determined as original nonstandard video；
Transcoding is carried out to the original nonstandard video, specified standard coding format is converted into, forms the transcoding standard
Video.
6. the video library processing method according to claim 4 or 5 for supporting more AI algorithms, which is characterized in that each algorithm
The primary standard video and/or transcoding normal video are carried out at analysis and comparison using different AI algorithms between service provider
Reason.
7. the video library processing method according to claim 6 for supporting more AI algorithms, which is characterized in that dispatching the calculation
When method service provider carries out analysis and comparison processing, scheduling at least two algorithm service quotient are to same primary standard video and/or turn
Code normal video carries out the analysis and comparison processing of Same Scene.
8. the video library processing method according to claim 6 for supporting more AI algorithms, which is characterized in that dispatching the calculation
When method service provider carries out analysis and comparison processing, according to the algorithm advantage of each algorithm service quotient, different algorithm services is dispatched
The analysis and comparison that quotient carries out corresponding scene to the primary standard video and/or transcoding normal video are handled.
9. the video library processing method according to claim 4 or 5 for supporting more AI algorithms, it is characterised in that:
Support that at least an algorithm service quotient passes through FTP/HTTP for the primary standard video and/or transcoding mark in the video library
Quasi- video download is into its local server.
10. the video library processing method according to claim 9 for supporting more AI algorithms, which is characterized in that at least one calculation
Method service provider connect reading in its local server and/or by shared storage network with the video library using multithreading
The picture, primary standard video and/or the transcoding normal video taken carries out single cent part, segment processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910104118.1A CN109862319A (en) | 2019-02-01 | 2019-02-01 | A kind of video library processing system and method for supporting more AI algorithms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910104118.1A CN109862319A (en) | 2019-02-01 | 2019-02-01 | A kind of video library processing system and method for supporting more AI algorithms |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109862319A true CN109862319A (en) | 2019-06-07 |
Family
ID=66897430
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910104118.1A Pending CN109862319A (en) | 2019-02-01 | 2019-02-01 | A kind of video library processing system and method for supporting more AI algorithms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109862319A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114257701A (en) * | 2020-09-23 | 2022-03-29 | 北京字节跳动网络技术有限公司 | Access configuration method, device and storage medium of video processing algorithm |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101808233A (en) * | 2009-02-18 | 2010-08-18 | 杭州海康威视数字技术股份有限公司 | Universal system for realizing video monitoring of mobile phone and method thereof |
| CN105871872A (en) * | 2016-04-28 | 2016-08-17 | 珠海经济特区远宏科技有限公司 | Movable video investigation method |
| CN106375721A (en) * | 2016-09-14 | 2017-02-01 | 重庆邮电大学 | Smart video monitoring system based on cloud platform |
| US20180152715A1 (en) * | 2016-11-25 | 2018-05-31 | Canon Kabushiki Kaisha | Method and system for determining encoding parameters of video sources in large scale video surveillance systems |
| CN109246444A (en) * | 2018-11-28 | 2019-01-18 | 中电科（德阳广汉）特种飞机系统工程有限公司 | Operation monitoring system, method, apparatus and the readable storage medium storing program for executing of mission payload |
-
2019
- 2019-02-01 CN CN201910104118.1A patent/CN109862319A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101808233A (en) * | 2009-02-18 | 2010-08-18 | 杭州海康威视数字技术股份有限公司 | Universal system for realizing video monitoring of mobile phone and method thereof |
| CN105871872A (en) * | 2016-04-28 | 2016-08-17 | 珠海经济特区远宏科技有限公司 | Movable video investigation method |
| CN106375721A (en) * | 2016-09-14 | 2017-02-01 | 重庆邮电大学 | Smart video monitoring system based on cloud platform |
| US20180152715A1 (en) * | 2016-11-25 | 2018-05-31 | Canon Kabushiki Kaisha | Method and system for determining encoding parameters of video sources in large scale video surveillance systems |
| CN109246444A (en) * | 2018-11-28 | 2019-01-18 | 中电科（德阳广汉）特种飞机系统工程有限公司 | Operation monitoring system, method, apparatus and the readable storage medium storing program for executing of mission payload |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114257701A (en) * | 2020-09-23 | 2022-03-29 | 北京字节跳动网络技术有限公司 | Access configuration method, device and storage medium of video processing algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3735679B1 (en) | Distributing shaders between client machines for precaching | |
| CN110019240A (en) | A kind of service data interaction method, apparatus and system | |
| CN104331435B (en) | A kind of efficient mass data abstracting method of low influence based on Hadoop big data platforms | |
| CN102664967A (en) | Method and system for interacting cross-platform personal information and background server | |
| CN105144737A (en) | Media distribution and management platform | |
| CN108228709B (en) | Data storage method and system, electronic device, program, and medium | |
| CN113497955B (en) | Video processing system | |
| CN115462062A (en) | Information processing apparatus and method | |
| CN104778252B (en) | The storage method and device of index | |
| CN110430398B (en) | Video coding distributed method based on reinforcement learning | |
| CN109862319A (en) | A kind of video library processing system and method for supporting more AI algorithms | |
| CN117834833A (en) | Data processing method of point cloud media and related equipment | |
| US10412391B1 (en) | Minimize number of encoded video stream frames for content recognition | |
| CN104572964A (en) | Zip file unzipping method and device | |
| US10264273B2 (en) | Computed information for metadata extraction applied to transcoding | |
| JP2006020330A (en) | Process and device for compressing video documents | |
| Schmidt et al. | An approach for processing large and non-uniform media objects on mapreduce-based clusters | |
| CN105323593A (en) | Multimedia transcoding scheduling method and multimedia transcoding scheduling device | |
| US20230018087A1 (en) | Data coding method and apparatus, and computer-readable storage medium | |
| CN107194013B (en) | Processing method and processing system based on Distributed Message Queue middleware | |
| CN105338371A (en) | Multimedia transcoding scheduling method and apparatus | |
| CN110309138A (en) | Data merging method, the combiner based on FPGA and Database Systems | |
| CN110213583B (en) | Video encoding method, system, apparatus and computer program medium | |
| CN103577410A (en) | Massive video integrated processing system | |
| EP4115602A1 (en) | An encoding device and method for utility-driven video compression |
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 |
Application publication date: 20190607 |
|
| RJ01 | Rejection of invention patent application after publication |