CN113612952A - Method for adjusting dynamic allocation of AI server resources based on frequency conversion example - Google Patents
Method for adjusting dynamic allocation of AI server resources based on frequency conversion example Download PDFInfo
- Publication number
- CN113612952A CN113612952A CN202110916339.6A CN202110916339A CN113612952A CN 113612952 A CN113612952 A CN 113612952A CN 202110916339 A CN202110916339 A CN 202110916339A CN 113612952 A CN113612952 A CN 113612952A
- Authority
- CN
- China
- Prior art keywords
- analysis
- frequency
- data
- current
- video
- 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
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000004458 analytical method Methods 0.000 claims abstract description 389
- 238000007405 data analysis Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims description 19
- 230000001960 triggered effect Effects 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 6
- 238000009877 rendering Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 3
- 239000000779 smoke Substances 0.000 description 14
- 238000012544 monitoring process Methods 0.000 description 8
- 239000000284 extract Substances 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000002045 lasting effect Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000013468 resource allocation Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000036760 body temperature Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 239000012466 permeate Substances 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/76—Television signal recording
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/75—Media network packet handling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/80—Responding to QoS
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The invention relates to a method for dynamically allocating resources of an AI server based on a frequency conversion example, which comprises the following steps: data acquisition service step: collecting video data of different sources and converting the video data into video stream data capable of being analyzed in the data analysis service step; data analysis service step: and sequentially reading the latest analysis frequency configuration, analyzing the video data, extracting the video picture and recording the video according to the AI analysis frequency configuration, performing AI analysis service after generating AI analysis unit data, and performing frequency conversion analysis service according to result data returned by the AI analysis service. The invention has the advantages that: the AI server is dynamically called mainly according to the idea of supply as required, so that automatic frequency conversion work is realized, the utilization of AI service resources is maximized and the operation cost is reduced under the condition of not influencing the service effect.
Description
Technical Field
The invention relates to the technical field of server resource allocation, in particular to a method for adjusting AI server resource allocation based on a frequency conversion example.
Background
With the development of 5G, Internet of things and artificial intelligence technology, video monitoring systems are greatly constructed and popularized. The intelligent construction of the society is gradually perfected and enriched. At present, "wisdom" ization construction mainly uses in wisdom city, wisdom district, wisdom building site, wisdom tourism, wisdom trip, wisdom street lamp, wisdom parking area etc. and permeates each industry. And intelligent video analysis plays a great role in intelligent construction. The application with video AI analysis capability is formed by utilizing the existing video monitoring and AI visual analysis and recognition technology. The functions of face feature recognition, human body behavior analysis, body temperature monitoring, people stream density monitoring, illegal invasion of dangerous areas, smoke fire alarm, illegal lane occupation analysis and the like are realized, and irreplaceable effects are achieved in the security protection field and other video data acquisition fields. The realization of the video AI analysis function requires high-performance service equipment, a large amount of software and algorithm services for support, and frequent data recording, storage, analysis and processing. The resource consumption is high, the use cost is high, and the comprehensive popularization and application in the industry are difficult.
In a traditional video AI analysis mode, when a camera normally works, high-frequency AI analysis is required to be kept all the time; whether the returned analysis result has a reference value or not, high-frequency analysis needs to be maintained, which brings huge pressure to a system server and storage and also causes unnecessary consumption and high consumption of AI resource services.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a method for adjusting the dynamic allocation of AI server resources based on a frequency conversion example, and solves the defects of the traditional video AI analysis mode.
The purpose of the invention is realized by the following technical scheme: the method for adjusting the dynamic allocation of the resources of the AI server based on the frequency conversion example comprises the following steps:
data acquisition service step: collecting video data of different sources and converting the video data into video stream data capable of being analyzed in the data analysis service step;
data analysis service step: and sequentially reading the latest analysis frequency configuration, analyzing the video data, extracting the video picture and recording the video according to the AI analysis frequency configuration, performing AI analysis service after generating AI analysis unit data, and performing frequency conversion analysis service according to result data returned by the AI analysis service.
The frequency conversion analysis service comprises:
whether frequency conversion analysis is triggered is judged by analyzing and calculating an AI analysis service result, if the frequency conversion analysis is not triggered, the original AI analysis frequency is kept, and if the frequency conversion analysis is triggered, the AI analysis frequency is restored to the standard AI analysis frequency of the current AI analysis model or is reduced;
and calculating the optimal AI analysis frequency configuration of the current AI analysis model through an algorithm, and storing the latest AI analysis frequency configuration into a database/cache so as to facilitate the invocation of video data analysis service.
The most recent analysis frequency configuration reading comprises: and acquiring latest AI analysis frequency configuration information of the current analysis model in storage, wherein the configuration records the analysis modes of different AI analysis models, the current analysis frequency, the normal analysis frequency, the frequency reduction gradient, the frequency reduction period and the configuration information of the lowest frequency, and if the latest AI analysis frequency configuration is not inquired, the standard AI analysis frequency is used by default.
The analyzing of the latest analysis frequency configuration comprises: and analyzing the current latest analysis frequency configuration information, and analyzing the configuration information of the analysis mode, the current analysis frequency, the normal analysis frequency, the frequency reduction gradient, the frequency reduction period and the lowest frequency in the configuration.
The video data parsing comprises: and configuring and analyzing video stream data according to the latest AI analysis frequency of the current AI analysis model, and generating unit data required by the AI analysis service of the current AI analysis model.
The video picture frame extraction comprises the following steps: and performing picture frame extraction on video stream data according to the latest analysis frequency configuration of the current AI analysis module, and assembling the frame-extracted pictures into unit data required by the AI analysis service of the current AI model.
The video recording comprises: and performing video recording on video stream data according to the latest AI analysis frequency configuration of the current AI analysis model, and assembling the recorded video into unit data required by the AI analysis service of the current AI analysis model.
The dynamic analysis method further comprises an analysis result rendering step, wherein the analysis result rendering step is used for checking the AI analysis result data display value client.
The invention has the following advantages: a method for adjusting the dynamic allocation of resources of an AI server based on a frequency conversion example automatically reduces the frequency of AI analysis when a video monitoring area is in a low peak period, at night or without personnel and the like, and automatically improves the frequency of AI analysis to be normal when the video monitoring area is in a high peak period, within a working time and under the condition of people stream recovery. The intelligent AI analysis configuration can be carried out in real time according to the result of the AI analysis, and the aim of frequency conversion work is achieved by changing the frequency of calling the AI analysis service. The method for adjusting the dynamic allocation of the AI server resources based on the frequency conversion example maximizes the utilization of the AI server resources and reduces the operation cost under the condition of not influencing the service effect.
Drawings
FIG. 1 is a flow chart of the present invention for frequency conversion adjustment of resources of an AI server;
FIG. 2 is a flow chart of a data parsing service in a variable frequency adjustment AI server resource according to the present invention;
FIG. 3 is a flow chart of the variable frequency adjustment AI server resource variable frequency service of the present invention;
FIG. 4 is a detailed flowchart of the present invention for frequency conversion adjustment of resources of an AI server;
FIG. 5 is a business flow diagram of example 1 of frequency conversion regulating AI server resources in an example of the invention;
fig. 6 is a business flow diagram of example 2 of frequency conversion adjustment of AI server resources in an example of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the detailed description of the embodiments of the present application provided below in connection with the appended drawings is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application. The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, the present invention relates to a method for dynamically allocating resources of an AI server based on a frequency conversion example, which dynamically calls the AI server to implement automatic frequency conversion according to the concept of "supply as required". When the AI analysis is in demand, the frequency of calling the AI analysis server is reduced. (ii) a It mainly comprises the following contents:
the video data 1, 2 and 3 are used for accessing video data of different sources and different root modes, including a camera data stream and a video file;
the data acquisition service is used for acquiring video stream data of the video data and converting the video data of different sources into video stream data which can be analyzed by the data analysis service;
and the data analysis service is used for sequentially reading the latest analysis frequency configuration, analyzing video data, extracting frames of video pictures and recording videos according to the AI analysis frequency configuration, generating AI analysis unit data, then performing the AI analysis service, and performing frequency conversion analysis service according to result data returned by the AI analysis service.
Further, as shown in fig. 2, the latest analysis frequency configuration reading is used to obtain the latest AI analysis frequency configuration information of the current analysis model in the storage. The configuration records the analysis modes of different AI analysis models, the current analysis frequency, the normal analysis frequency, the frequency reduction gradient, the frequency reduction period, the lowest frequency and other configuration information. If the latest AI analysis frequency configuration is not inquired, the standard AI analysis frequency is used by default, and the standard AI analysis frequency is the AI analysis frequency used by the AI analysis model during initialization and frequency conversion recovery. While analyzing the frequency configuration, external condition parameters (campus announcements, working hours, climate, etc.) can also be added.
And analyzing the latest analysis frequency configuration, wherein the analysis frequency configuration is used for analyzing the current latest analysis frequency configuration information. And analyzing configuration information such as an analysis mode, current analysis frequency, normal analysis frequency, down-conversion gradient, down-conversion period, lowest frequency and the like in the configuration.
And video data analysis, configured to analyze video stream data according to the latest AI analysis frequency of the current AI analysis model, and generate unit data required by the AI analysis service of the current AI analysis model. Two analysis modes are mainly listed, namely video picture frame extraction and video recording.
And the video picture frame extraction is used for extracting the frame of the video stream data into picture data. And performing picture frame extraction on video stream data according to the latest AI analysis frequency configuration of the current AI analysis model, and assembling the frame-extracted pictures into unit data required by the AI analysis service of the current AI model.
And video recording, which is used for recording the video stream data. And performing video recording on video stream data according to the latest AI analysis frequency configuration of the current AI analysis model, and assembling the recorded video into unit data required by the AI analysis service of the current AI analysis model.
AI analysis cell data is generated for assembly into a cell data format required for an AI analysis service invoking the current AI analysis model.
And the frequency conversion analysis service is used for calculating the optimal AI analysis frequency configuration of the current AI analysis model through an algorithm according to result data returned by the AI analysis service, and updating the AI analysis frequency configuration of the current AI analysis model into an analysis frequency storage.
As shown in fig. 3, the specific flow of the frequency conversion service is as follows:
analyzing and calculating an AI analysis result, processing AI analysis result data, and calculating the optimal AI analysis frequency configuration of the current AI analysis model through an algorithm;
and whether frequency conversion analysis is triggered is used for judging whether frequency conversion analysis is triggered according to the AI analysis result data of the current AI analysis model. And if the frequency conversion is not triggered, keeping the original AI analysis frequency. If frequency conversion analysis is triggered, two frequency conversion conditions are adopted. One is to restore the AI analysis frequency to the standard AI analysis frequency of the current AI analysis model. One is to reduce the AI analysis frequency.
The method comprises the steps of maintaining original analysis frequency, and maintaining original AI analysis frequency of a current AI analysis model when frequency conversion processing is not needed; recovering standard frequency analysis, wherein the standard frequency analysis is used for frequency-converted analysis service and recovering the standard AI analysis frequency using the current AI analysis model, and the standard AI analysis frequency is used when the current AI analysis model is initialized and frequency conversion is recovered; and the frequency reduction analysis is used for modifying the AI analysis frequency configuration of the current AI analysis model according to the AI analysis result data of the current AI analysis model. The AI analysis frequency of the current AI analysis model is reduced.
And generating a latest analysis frequency configuration, calculating the optimal AI analysis frequency configuration of the current AI analysis model through an algorithm, and storing the latest AI analysis frequency configuration into a database/cache so as to facilitate the calling of video data analysis service.
And storing the latest AI analysis frequency configuration, and storing the latest AI analysis frequency configuration into a database/cache so as to be called by the video data analysis service. The database records the AI analysis frequency configuration after each modification, and the current latest AI analysis frequency configuration is recorded in the cache.
Rendering an analysis result, and displaying the AI analysis result data to the client for viewing.
Further, as shown in fig. 4, the data parsing service first reads the latest AI analysis frequency configuration information of the current AI analysis model in the cache, adjusts the parsing frequency of the video stream according to the AI analysis frequency configuration of the current AI analysis model, controls the frequency of frame extraction of the video stream, and controls the time interval at which the video stream records the video. Element data required for an AI analysis service of the current AI analysis model is generated. For example, people flow density information in video data needs to be analyzed, and people number information in a region needs to be counted. The AI analysis frequency configuration obtained from the people flow density analysis model is as follows: "{ analytical model: analyzing the density of the human stream in an analytic mode: frame extraction, current analysis frequency: 1/sec, standard analysis frequency: 1/sec, down gradient: current analysis frequency 2^ (number of downconversion), downconversion period: 5 minutes, lowest frequency: 1 times/10 minutes } ". When people stream density detection and analysis is normally started, the data analysis service extracts 1 picture per second and sends the picture to the AI analysis service of people stream density analysis, and people number information in the picture is counted. Since no people flow trace image exists in the video at the moment, the AI analysis service for people flow density analysis returns an analysis result of 0 people. And no sign of human flow lasts for 5 minutes, and the AI analysis service of the human flow density analysis always returns 0 people. Triggering the data analysis service to automatically reduce the frequency, gradually reducing the AI analysis frequency of the people flow density analysis, and extracting 1 picture every 2 seconds and sending the picture to the AI analysis service of the people flow density analysis. If the duration returns a result that is still 0 people. And continuing to reduce the AI analysis frequency of the people flow density analysis, extracting 1 picture every 4 seconds and sending the picture to the AI analysis service of the people flow density analysis. The duration returned results were 0 people. And then 1 picture is extracted every 10 minutes and sent to an AI analysis service for people stream density analysis. Until the number of persons who return the analysis result is not 0. The frequency of the AI analysis service that extracts 1 picture per second and sends it to the traffic density analysis is restored.
And transmitting the analyzed data to an AI analysis server of the current AI analysis model for AI analysis to obtain AI analysis result data.
And analyzing and calculating the result returned by the AI analysis to obtain the optimal AI analysis frequency of the current AI analysis model. And if the analysis result of the video monitoring area has no reference value for a long time, reducing the AI analysis frequency of the current AI analysis model. If the outdoor smoke fire analysis is reduced or stopped while the video surveillance area is in the rainy day. And the important security area continuously keeps reasonable analysis frequency. And when the video monitoring area subjected to the frequency reduction analysis has a result with reference value again, restoring the AI analysis frequency of the video data to the standard analysis frequency of the current AI analysis model.
The latest AI analysis frequency configuration is stored for data analysis service invocation.
And the data analysis service adjusts the analysis frequency of the video stream, controls the frequency of the frame extraction pictures of the video stream and controls the time interval of the video stream for recording the video according to the latest AI analysis frequency configuration. And finally, presenting the analyzed result data to a client.
As shown in fig. 5, the specific implementation steps of the embodiment 1 are as follows:
step 1, calling a data acquisition service to obtain video stream data, wherein the data acquisition service accesses video data of different sources and different formats, including a camera data stream and a video file.
And 2, reading and analyzing the latest AI analysis frequency configuration of the current AI analysis model.
And 3, analyzing the video stream data according to the latest AI analysis frequency to obtain unit data required by the AI analysis of the current AI analysis model. The data analysis service firstly reads the latest AI analysis frequency configuration information of the current AI analysis model in the cache, adjusts the analysis frequency of the video stream according to the AI analysis frequency configuration, controls the frequency of the frame extraction pictures of the video stream, and controls the time interval of the video stream recording video. Element data required for an AI analysis service of the current AI analysis model is generated. For example, people flow density information in video data needs to be analyzed, and people number information in a region needs to be counted. The AI analysis frequency configuration obtained from the people flow density analysis model is as follows: "{ analytical model: analyzing the density of the human stream in an analytic mode: frame extraction, current analysis frequency: 1/sec, standard analysis frequency: 1/sec, down gradient: current analysis frequency 2^ (number of downconversion), downconversion period: 5 minutes, lowest frequency: 1 times/10 minutes } ". The analysis mode of the people flow density analysis model is 'frame drawing picture', and the current analysis frequency is '1 time/second' obtained from the current configuration. And performing frame extraction on the video stream data according to the '1 time/second', and assembling the frame-extracted pictures into unit data required by an AI analysis service of people stream density analysis.
And 4, calling the AI analysis service of the current AI analysis model to obtain analysis result data.
And 5, calling a frequency conversion analysis service according to the AI analysis result data of the current AI analysis model, and calculating the configuration of the optimal AI analysis frequency of the current AI analysis model. For example, the AI analysis frequency configuration of the crowd density analysis model in step 3 is combined with the analysis result obtained in step 4. If the returned result indicates that there is human stream data, the current analysis frequency is kept at "1 time/second". The AI analysis frequency configuration of the people stream density analysis is not changed. If the AI analysis returns that the result indicates no-man-flow data, the current analysis frequency is kept for analysis. If the AI analyses lasting 5 minutes all return results indicating no-flow data. The AI analysis service triggering the people flow density analysis automatically reduces the frequency, gradually reduces the AI analysis frequency of the people flow density analysis, extracts 1 picture every 2 seconds after the first frequency reduction according to the rule that the frequency reduction gradient configured by the AI analysis of the people flow density analysis is 2 frequency reduction times, and sends the picture to the AI analysis service of the people flow density analysis. If the AI analysis lasting 5 minutes returns a result that is still 0 population. And continuing to reduce the AI analysis frequency of the people flow density analysis, extracting 1 picture every 4 seconds and sending the picture to the AI analysis service of the people flow density analysis. The duration returned results were 0 people. And then 1 picture is extracted every 10 minutes and sent to an AI analysis service for people stream density analysis. Until the number of persons who return the analysis result is not 0. The AI analysis frequency of the crowd density analysis is restored to the standard analysis frequency in the AI analysis frequency configuration of the crowd density analysis. The frequency of AI analysis services that extract 1 picture per second to send to traffic density analysis. And updating the frequency value of each frequency reduction to the AI analysis frequency configuration of the people flow density analysis model.
And 6, judging the latest AI analysis frequency configuration information.
And 7, storing and updating the latest AI analysis frequency configuration so as to facilitate the invocation of the data analysis service.
And 8, calling data analysis service, and automatically adjusting the frequency of video stream analysis according to the latest AI analysis frequency configuration of the current AI analysis model.
As shown in fig. 6, the specific implementation steps of the embodiment 2 are as follows:
step 1, calling a data acquisition service to obtain video stream data, wherein the data acquisition service accesses video data of different sources and different formats, including a camera data stream and a video file.
And 2, reading and analyzing the latest AI analysis frequency configuration of the current AI analysis model.
And 3, reading external condition parameters (park notification, working time, climate and the like) and integrating the external condition parameters into the latest AI analysis frequency configuration. The external parameters may be derived from the notification of the start-up and shutdown of the park, from the working time of the monitored area (or the time period required to be monitored), or from data such as the climate. The source of the final external parameters is set by the customer.
And 4, analyzing the video stream data according to the latest AI analysis frequency to obtain unit data required by the AI analysis of the current AI analysis model. The data analysis service firstly reads the latest AI analysis frequency configuration information of the current AI analysis model in the cache, adjusts the analysis frequency of the video stream according to the AI analysis frequency configuration, controls the frequency of the frame extraction pictures of the video stream, and controls the time interval of the video stream recording video. Element data required for an AI analysis service of the current AI analysis model is generated.
Step 5, for example, step 1, people stream density information in the video data needs to be analyzed, and people number information in the area needs to be counted. The AI analysis frequency configuration obtained from the people flow density analysis model is as follows: "{ analytical model: analyzing the density of the human stream in an analytic mode: frame extraction, current analysis frequency: 1/sec, standard analysis frequency: 1/sec, down gradient: current analysis frequency 2^ (number of downconversion), downconversion period: 5 minutes, lowest frequency: 1 times/10 minutes, other conditions: { area working time (or period of time for which statistical people flow information is required): 08: 30-17: 30}}". The analysis mode of the people flow density analysis model is 'frame drawing picture', and the current analysis frequency is '1 time/second' obtained from the current configuration. The video stream data should be decimated by "1 time/second". However, since the "zone operating time" in the "other conditions" is "08: 30-17: 30 ', it needs to be determined whether the current time is within the range of the ' regional work time '. And if the current time is in the range of the regional work time, performing frame extraction on the video stream data according to the rate of 1 time/second. Otherwise, frame extraction is carried out on the video stream data according to the lowest frequency of 1 time/10 minutes. And assembling the framed pictures into unit data required by AI analysis service of people flow density analysis.
Step 6, for example, step 2, the smoke fire condition in the video data needs to be analyzed, and whether the smoke fire occurs in the monitored area or not needs to be analyzed. The AI analysis frequency obtained into the smoke fire analysis model is configured as: "{ analytical model: smoke fire analysis, analytic mode: frame extraction, current analysis frequency: 1 times/5 seconds, standard analysis frequency: 1/5 sec, down gradient: ", the downconversion cycle: 5 minutes, lowest frequency: 1 times/5 min, other conditions: { climate: rain } } ". The analysis mode of the smoke fire analysis model is 'frame picture extraction', and the current analysis frequency is '1 time/5 seconds' according to the current configuration. The video stream data should be decimated by "1 time/5 seconds". However, since "weather" in "other conditions" is "rain", the probability of occurrence of a smoke fire in rainy days is extremely low. The video stream data is decimated at the lowest frequency of "1 time/5 minutes". And assembling the framed pictures into unit data required by AI analysis service of smoke fire analysis.
And 7, calling the AI analysis service of the current AI analysis model to obtain analysis result data.
And 8, calling a frequency conversion analysis service according to the AI analysis result data of the current AI analysis model, and calculating the configuration of the optimal AI analysis frequency of the current AI analysis model. For example, the frequency configuration is analyzed according to the analysis results obtained in step 6 and step 7 in combination with the AI analysis of the smoke fire analysis model in step 2 and step 3. If the returned result indicates a smoke-free fire, and the "climate" in the "other conditions" in the AI analysis frequency configuration of the smoke fire analysis model in steps 2 and 3 is "rain" anyway. The current analysis frequency is kept at "1 time/5 minutes". If the returned result indicates a smoke-free fire, "climate" in "other conditions" in the AI analysis frequency configuration of the smoke fire analysis model in steps 2 and 3 is not "rain". The current analysis frequency is restored to the standard analysis frequency of "1 time/5 seconds". The "down gradient" in the AI analysis frequency configuration due to the smoke fire analysis model in steps 2 and 3 is empty or no down gradient is configured. Then in the case where the "climate" in the "other conditions" is not "rain" in this configuration, the standard analysis frequency "1/5 seconds" is always maintained for analysis. In the case where "climate" in "other conditions" is "rain" in this configuration, the analysis is performed while keeping the standard analysis frequency "1 time/5 minutes" at all times. And if the returned result indicates that smoke fire exists, pushing the returned result to the monitoring center and sending out a danger warning. And updating the frequency value of each frequency reduction to the AI analysis frequency configuration of the people flow density analysis model.
And 9, judging the latest AI analysis frequency configuration information.
And step 10, storing and updating the latest AI analysis frequency configuration so as to facilitate the invocation of the data analysis service.
And step 11, calling a data analysis service, and automatically adjusting the frequency of video stream analysis according to the latest AI analysis frequency configuration of the current AI analysis model.
The foregoing is illustrative of the preferred embodiments of this invention, and it is to be understood that the invention is not limited to the precise form disclosed herein and that various other combinations, modifications, and environments may be resorted to, falling within the scope of the concept as disclosed herein, either as described above or as apparent to those skilled in the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (8)
1. The method for adjusting the dynamic allocation of the resources of the AI server based on the frequency conversion example is characterized in that: the dynamic allocation method comprises the following steps:
data acquisition service step: collecting video data of different sources and converting the video data into video stream data capable of being analyzed in the data analysis service step;
data analysis service step: and sequentially reading the latest analysis frequency configuration, analyzing the video data, extracting the video picture and recording the video according to the AI analysis frequency configuration, performing AI analysis service after generating AI analysis unit data, and performing frequency conversion analysis service according to result data returned by the AI analysis service.
2. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the frequency conversion analysis service comprises:
whether frequency conversion analysis is triggered is judged by analyzing and calculating an AI analysis service result, if the frequency conversion analysis is not triggered, the original AI analysis frequency is kept, and if the frequency conversion analysis is triggered, the AI analysis frequency is restored to the standard AI analysis frequency of the current AI analysis model or is reduced;
and calculating the optimal AI analysis frequency configuration of the current AI analysis model through an algorithm, and storing the latest AI analysis frequency configuration into a database/cache so as to facilitate the invocation of video data analysis service.
3. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the most recent analysis frequency configuration reading comprises: and acquiring latest AI analysis frequency configuration information of the current analysis model in storage, wherein the configuration records the analysis modes of different AI analysis models, the current analysis frequency, the normal analysis frequency, the frequency reduction gradient, the frequency reduction period and the configuration information of the lowest frequency, and if the latest AI analysis frequency configuration is not inquired, the standard AI analysis frequency is used by default.
4. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the analyzing of the latest analysis frequency configuration comprises: and analyzing the current latest analysis frequency configuration information, and analyzing the configuration information of the analysis mode, the current analysis frequency, the normal analysis frequency, the frequency reduction gradient, the frequency reduction period and the lowest frequency in the configuration.
5. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the video data parsing comprises: and configuring and analyzing video stream data according to the latest AI analysis frequency of the current AI analysis model, and generating unit data required by the AI analysis service of the current AI analysis model.
6. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the video picture frame extraction comprises the following steps: and performing picture frame extraction on video stream data according to the latest analysis frequency configuration of the current AI analysis module, and assembling the frame-extracted pictures into unit data required by the AI analysis service of the current AI model.
7. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the video recording comprises: and performing video recording on video stream data according to the latest AI analysis frequency configuration of the current AI analysis model, and assembling the recorded video into unit data required by the AI analysis service of the current AI analysis model.
8. The method according to claim 1, wherein the method for dynamically allocating resources of the AI server based on the frequency conversion algorithm comprises: the dynamic analysis method further comprises an analysis result rendering step, wherein the analysis result rendering step is used for checking the AI analysis result data display value client.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110916339.6A CN113612952A (en) | 2021-08-11 | 2021-08-11 | Method for adjusting dynamic allocation of AI server resources based on frequency conversion example |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110916339.6A CN113612952A (en) | 2021-08-11 | 2021-08-11 | Method for adjusting dynamic allocation of AI server resources based on frequency conversion example |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113612952A true CN113612952A (en) | 2021-11-05 |
Family
ID=78308062
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110916339.6A Pending CN113612952A (en) | 2021-08-11 | 2021-08-11 | Method for adjusting dynamic allocation of AI server resources based on frequency conversion example |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113612952A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104618679A (en) * | 2015-03-13 | 2015-05-13 | 南京知乎信息科技有限公司 | Method for extracting key information frame from monitoring video |
| CN104991830A (en) * | 2015-07-10 | 2015-10-21 | 山东大学 | YARN resource allocation and energy-saving scheduling method and system based on service level agreement |
| CN111260869A (en) * | 2020-01-19 | 2020-06-09 | 世纪龙信息网络有限责任公司 | Method and device for extracting video frames in monitoring video and computer equipment |
| CN113114508A (en) * | 2021-04-15 | 2021-07-13 | 上海理想信息产业（集团）有限公司 | Multistage variable-frequency network monitoring data acquisition method and device |
-
2021
- 2021-08-11 CN CN202110916339.6A patent/CN113612952A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104618679A (en) * | 2015-03-13 | 2015-05-13 | 南京知乎信息科技有限公司 | Method for extracting key information frame from monitoring video |
| CN104991830A (en) * | 2015-07-10 | 2015-10-21 | 山东大学 | YARN resource allocation and energy-saving scheduling method and system based on service level agreement |
| CN111260869A (en) * | 2020-01-19 | 2020-06-09 | 世纪龙信息网络有限责任公司 | Method and device for extracting video frames in monitoring video and computer equipment |
| CN113114508A (en) * | 2021-04-15 | 2021-07-13 | 上海理想信息产业（集团）有限公司 | Multistage variable-frequency network monitoring data acquisition method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108040221B (en) | Intelligent video analysis and monitoring system | |
| CN108965825B (en) | Video linkage scheduling method based on holographic position map | |
| CN201965714U (en) | Home intelligent early-warning security system based on human face recognition | |
| US20220092881A1 (en) | Method and apparatus for behavior analysis, electronic apparatus, storage medium, and computer program | |
| CN108197565A (en) | Target based on recognition of face seeks track method and system | |
| CN112969049B (en) | Intelligent detection system for ship violation behaviors | |
| CN106780250B (en) | Intelligent community security event processing method and system based on Internet of things technology | |
| CN110705482A (en) | Personnel behavior alarm prompt system based on video AI intelligent analysis | |
| CN101860731A (en) | Video information processing method, system and server | |
| JP2006514757A (en) | Method and system for efficiently performing event detection in multiple simultaneous video images | |
| CN106412522A (en) | Video analysis detection method and system of object in indoor and outdoor environment | |
| CN111405249A (en) | Monitoring method, monitoring device, server and computer-readable storage medium | |
| CN109905423B (en) | Intelligent management system | |
| CN114679607B (en) | Video frame rate control method and device, electronic equipment and storage medium | |
| CN111368741A (en) | High-altitude parabolic detection system | |
| CN109040703A (en) | A kind of monitoring security-protection management system | |
| CN103390347B (en) | A kind of taxi intelligent visual and pattern recognition device and method of work | |
| CN111178241A (en) | Intelligent monitoring system and method based on video analysis | |
| CN112699328A (en) | Network point service data processing method, device, system, equipment and storage medium | |
| CN101808229A (en) | Video stream rapid-playback system based on feature tag | |
| CN113612952A (en) | Method for adjusting dynamic allocation of AI server resources based on frequency conversion example | |
| CN116133031A (en) | Building network quality assessment method, device, electronic equipment and medium | |
| CN107424382A (en) | A kind of underground palace monitoring system | |
| CN111489276B (en) | Personnel management method and related device | |
| CN114528425A (en) | Three-dimensional visualization system and method for intelligent building |
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 |