CN116610834A - Monitoring video storage and quick query method based on AI analysis - Google Patents

Monitoring video storage and quick query method based on AI analysis Download PDF

Info

Publication number
CN116610834A
CN116610834A CN202310542271.9A CN202310542271A CN116610834A CN 116610834 A CN116610834 A CN 116610834A CN 202310542271 A CN202310542271 A CN 202310542271A CN 116610834 A CN116610834 A CN 116610834A
Authority
CN
China
Prior art keywords
target
image
video
monitoring
preset
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.)
Granted
Application number
CN202310542271.9A
Other languages
Chinese (zh)
Other versions
CN116610834B (en
Inventor
陈庆锋
张晨
邱生顺
余雅滢
郑黎明
李晓波
陈昌浩
王青林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Three Gorges Changdian Big Data Technology Yichang Co ltd
Three Gorges High Technology Information Technology Co ltd
Three Gorges Technology Co ltd
Original Assignee
Three Gorges Changdian Big Data Technology Yichang Co ltd
Three Gorges High Technology Information Technology Co ltd
Three Gorges Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Three Gorges Changdian Big Data Technology Yichang Co ltd, Three Gorges High Technology Information Technology Co ltd, Three Gorges Technology Co ltd filed Critical Three Gorges Changdian Big Data Technology Yichang Co ltd
Priority to CN202310542271.9A priority Critical patent/CN116610834B/en
Publication of CN116610834A publication Critical patent/CN116610834A/en
Application granted granted Critical
Publication of CN116610834B publication Critical patent/CN116610834B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00—Television systems
    • H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
    • H04N7/181—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast for receiving images from a plurality of remote sources
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/71—Indexing; Data structures therefor; Storage structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/73—Querying
    • G06F16/732—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7837—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using objects detected or recognised in the video content
    • G06F16/784—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using objects detected or recognised in the video content the detected or recognised objects being people
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7847—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using low-level visual features of the video content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/48—Matching video sequences
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/49—Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
    • H04N21/231—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion
    • H04N21/23106—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion involving caching operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Library & Information Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Medical Informatics (AREA)
  • Mathematical Physics (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Human Computer Interaction (AREA)
  • Computational Linguistics (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

The invention provides a monitoring video storage and quick query method based on AI analysis, which comprises the following steps: video monitoring is carried out on a target monitoring area based on a monitoring camera to obtain a video stream, the video stream is split based on a preset unit video cache frame, and different preset channels in a streaming media server are subjected to queue cache on average; image interception is carried out on the obtained video stream through a streaming media server based on a target time interval, and the intercepted image is transmitted to an AI analysis model for analysis to determine an abnormal image; marking the abnormal image and transmitting the abnormal image back to a corresponding target preset channel in the streaming media server; and storing the video cached in the target preset channel based on the streaming media server, and calling and feeding back the stored video to the management terminal according to the query requirement of the management terminal based on the storage result. The efficiency of locating the abnormal video segment is improved, the video storage workload is reduced, and the efficiency of calling the abnormal video segment is improved.

Description

Monitoring video storage and quick query method based on AI analysis
Technical Field
The invention relates to the technical field of image and data processing, in particular to a monitoring video storage and quick query method based on AI analysis.
Background
At present, with the development of the Internet and the general improvement of people's safety consciousness, more and more people choose to monitor the area to be monitored by adopting a monitoring camera, and the real-time state of the current area can be effectively known by monitoring the area to be monitored by the monitoring camera, so that the security effect is improved;
however, monitoring the area to be monitored by adopting the monitoring cameras has a plurality of defects, in the present stage, a plurality of monitoring cameras can realize real-time monitoring, and the non-compliance phenomenon can be timely found through AI analysis, notification is found, but when the monitoring time of the cameras is long, a large amount of monitoring videos can be generated, in the prior art, all videos are uniformly stored, the storage workload of the videos is increased, and when the area to be monitored is abnormal, the positioning videos are difficult to be adjusted and positioned for the non-compliance phenomenon, so that the monitoring effect is greatly reduced;
therefore, the invention provides a monitoring video storage and quick query method based on AI analysis.
Disclosure of Invention
The invention provides a monitoring video storage and quick query method based on AI analysis, which is used for intercepting an image of a collected video stream through a streaming media server, inputting the intercepted image into an AI analysis model for high-efficiency and reliable analysis, and realizing quick locking of an image with an abnormality, thereby facilitating locking and storing of a video segment corresponding to the image with the abnormality, improving efficiency of locating the abnormal video segment, reducing workload of storing a large amount of videos, simultaneously facilitating quick retrieval of the abnormal video segment according to a user query requirement, guaranteeing accuracy and improving working efficiency.
The invention provides a monitoring video storage and quick query method based on AI analysis, which comprises the following steps:
step 1: video monitoring is carried out on a target monitoring area based on a monitoring camera to obtain a video stream, the video stream is split based on a preset unit video cache frame, and different preset channels in a streaming media server are subjected to queue cache on average;
step 2: image interception is carried out on the obtained video stream through a streaming media server based on a target time interval, and the intercepted image is transmitted to an AI analysis model for analysis, so that an abnormal image is determined;
Step 3: marking the abnormal image, and transmitting the marked abnormal image back to a corresponding target preset channel in the streaming media server;
step 4: and storing the video cached in the target preset channel based on the streaming media server, and calling and feeding back the stored video to the management terminal according to the query requirement of the management terminal based on the storage result.
Preferably, in step 1, video monitoring is performed on a target monitoring area based on a monitoring camera to obtain a video stream, which includes:
acquiring regional characteristics of a target monitoring region, determining effective monitoring cameras in the target monitoring region based on the regional characteristics, and determining terminal identity information of each effective monitoring camera based on a communication protocol of each effective monitoring camera in a local area network;
transmitting a video acquisition instruction to each effective monitoring camera through a local area network based on the streaming media server, and controlling each effective monitoring camera to acquire video in each monitoring range based on the video acquisition instruction;
and correlating and binding videos acquired by each effective monitoring camera with terminal identity information of each effective monitoring camera, transmitting the videos acquired by each effective monitoring camera to a streaming media server based on a local area network based on a binding result, and summarizing the received videos based on the streaming media server to obtain a final video stream.
Preferably, in step 1, a video stream is split based on a preset unit video buffer frame, and different preset channels in a streaming media server are buffered in a queue on average, which includes:
acquiring configuration information of a streaming media server, determining the number of preset channels contained in the streaming media server based on the configuration information, determining preset unit video cache frames of the video in unit time of each preset channel based on the configuration information of the streaming media server, and determining splitting position points of the video stream based on the preset unit video cache frames and the preset channel number;
carrying out average splitting on the video stream based on the splitting position point to obtain a video segment, and packaging the video segment obtained after splitting to obtain a target video segment;
determining a starting identifier and a terminating identifier of a cache queue in a preset channel, and caching the target video segment in the cache queue between the starting identifier and the terminating identifier based on the starting identifier and the terminating identifier;
and configuring a data temporary storage thread for a cache queue in each preset channel based on a cache result, and locking a target video segment currently cached in the cache queue based on the data temporary storage thread after the target video segment in the cache queue is cached.
Preferably, in step 2, image capturing is performed on an obtained video stream by a streaming media server based on a preset time interval, which includes:
the method comprises the steps of obtaining an obtained video stream, and playing the obtained video stream based on a preset video playing interface;
acquiring a target requirement for image interception of the video stream, determining a target time interval for image interception of the video stream based on the target requirement, and generating a periodic image interception instruction based on the target time interval;
triggering a streaming media server to execute image interception operation based on the periodical image interception instruction, and determining a triggering time point corresponding to the image interception operation in a video stream in play;
and determining an image to be captured corresponding to the video stream in playing based on the trigger time point, and capturing the image to be captured based on the streaming media server.
Preferably, in step 2, the captured image is transmitted to an AI analysis model for analysis, and an abnormal image is determined, which includes:
acquiring monitoring types of a target monitoring area, and calling standard monitoring images corresponding to each monitoring type and image evaluation indexes corresponding to each monitoring type from a preset database based on the monitoring types;
Performing image scanning on the standard monitoring images of each monitoring type based on a column scanning mode to obtain image characteristics of each standard monitoring image, and determining a mapping relation between the image characteristics and image evaluation indexes;
determining target weight of image characteristics of each standard monitoring image based on the monitoring requirements, and determining index values corresponding to each image evaluation index based on the target weight and the mapping relation;
constructing a monitoring behavior analysis information base based on image evaluation indexes corresponding to each monitoring type, index values corresponding to each image evaluation index and standard monitoring images corresponding to each monitoring type, matching a basic model frame from a preset model base based on monitoring requirements, performing iterative training on the basic model frame based on the monitoring behavior analysis information base, configuring image analysis strategies corresponding to each monitoring type in the basic model frame, and adapting each image analysis strategy to basic configuration parameters of the basic model frame to obtain an AI analysis model;
inputting the intercepted image into an AI analysis model, extracting image characteristics of the intercepted image based on the AI analysis model, and determining a target monitoring type based on the image characteristics;
Invoking a target image evaluation index based on a target monitoring type, and analyzing the intercepted image based on the target image evaluation index to obtain a target difference value of the intercepted image and a standard monitoring image;
when the target difference value is larger than a preset threshold value, judging that the intercepted image is an abnormal image, and meanwhile, adding a type identifier to the abnormal image based on the target monitoring type to obtain a final abnormal image.
Preferably, in step 3, marking an abnormal image includes:
reading the abnormal image, determining scene characteristics of the abnormal image, and determining an image identifier of the abnormal image according to the scene characteristics of the abnormal image;
inputting the image identification of the abnormal image into a preset image library for matching, and obtaining a reference image consistent with the image identification;
comparing the abnormal image with a reference image, determining a sensitive area of the abnormal image based on a comparison result, identifying the sensitive area, and determining pixel point distribution of the sensitive area and pixel point edges of the sensitive area;
and marking the abnormal image based on the pixel point distribution of the sensitive area and the pixel point edge of the sensitive area.
Preferably, in step 3, the marked abnormal image is returned to a corresponding target preset channel in the streaming media server, which comprises the following steps:
obtaining a cache address of an abnormal image, and reading channel addresses of different preset channels;
respectively matching the cache address of the abnormal image with the channel addresses of different preset channels, and determining a target preset channel corresponding to the abnormal image based on a matching result;
reading a target address corresponding to a target preset channel, generating a data return request based on the cache address and the target address, sending the data return request to a streaming media server for reading, simultaneously, calling cache logs of the video stream in different preset channels of the streaming media server for queue cache, verifying the data return request based on the cache logs, and judging whether the abnormal image is consistent with the target preset channel;
when the data return request does not have a corresponding matching relation in the cache log, judging that the abnormal image is inconsistent with the target preset channel, outputting a verification failure report, and simultaneously re-matching the target preset channel based on the verification failure report;
When the data feedback request has a corresponding matching relation in the cache log, the abnormal image is judged to be consistent with the target preset channel, meanwhile, the running state of the current target preset channel is read based on the streaming media server, when the current target preset channel outputs an idle signal, a verification success report is generated, and meanwhile, the marked abnormal image is transmitted back to the corresponding target preset channel in the streaming media server based on the verification success report.
Preferably, in step 4, the video cached in the target preset channel is saved and stored based on the streaming media server, which includes:
determining the video quantity in a target preset channel based on a preset unit video cache frame, and simultaneously determining the video information cached in the target preset channel;
generating a first monitoring factor in the streaming media server based on the video amount in the target preset channel, and generating a second monitoring factor in the streaming media server based on the video information cached in the target preset channel;
when the first monitoring factor and the second monitoring factor are generated, receiving and storing the video cached in the target preset channel based on the streaming media server, and simultaneously, performing first monitoring on the video storage cached in the target preset channel based on the first monitoring factor and performing second monitoring on the video storage cached in the target preset channel based on the second monitoring factor;
When the first monitoring result reaches a first preset standard and the second monitoring result reaches a second preset standard, judging that the streaming media server finishes storing the cache video in the target preset channel;
and after the storage of the video cached in the target preset channel is completed, reading the channel information of the target preset channel, generating a data storage identifier based on the channel information of the target preset channel, constructing a target storage data block based on the data storage identifier, and storing the video cached in the target preset channel into the corresponding target storage data block.
Preferably, a method for storing and rapidly querying a surveillance video based on AI analysis, which retrieves and feeds back the stored video to a management terminal according to a query requirement of the management terminal based on a storage result, includes:
generating a query requirement based on a management terminal, performing first reading on the query requirement, determining contents to be queried, and simultaneously reading user information of a query user;
inputting the content to be queried to a management terminal to call a user with the queriable authority, matching the user information in the user with the queriable authority, and determining whether the user information is in the user with the queriable authority;
When the user information is in the inquireable right, inputting the inquiry requirement to the streaming media server, simultaneously, carrying out second reading on the inquiry requirement based on the streaming media server, determining inquiry keywords of the inquiry requirement, and simultaneously, generating a video inquiry index based on the inquiry keywords;
and matching the video query indexes in the stored videos, calling the video segment of interest based on the matching result, and feeding the video segment of interest back to the management terminal.
Preferably, a method for storing and rapidly querying surveillance videos based on AI analysis, step 2 further includes:
when the intercepted image is transmitted to an AI analysis model for analysis, the method comprises the steps of monitoring personnel climbing in a target monitoring area, wherein the specific process is as follows:
acquiring adjacent frame numbers in the intercepted images, acquiring corresponding first target images and second target images based on the adjacent frame numbers of the screenshot images, and respectively determining a human body detection frame of the first target images and a human body detection frame of the second target images;
calculating a target aspect ratio change rate between the human body detection frame of the first target image and the human body detection frame of the second target image and a centroid change rate of the human body detection frame of the first target image and the second target image based on the change conditions between the human body detection frame of the first target image and the human body detection frame of the second target image;
Acquiring a preset length-width ratio change rate threshold and a preset centroid change rate threshold;
comparing the target length-width ratio change rate with a preset length-width ratio change rate threshold, and simultaneously comparing the centroid change rate with a preset centroid change rate threshold to judge whether personnel climbing behaviors exist in a target detection area;
when the target length-width ratio change rate is equal to or greater than a preset length-width ratio change rate threshold value and the centroid change rate is equal to or greater than a preset centroid change rate threshold value, determining that a person climbing behavior exists in the target detection area;
otherwise, judging that no personnel climbing behavior exists in the target detection area.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a flowchart of a monitoring video storage and quick query method based on AI analysis in an embodiment of the invention;
FIG. 2 is a flowchart of step 1 in a monitoring video storage and quick query method based on AI analysis in an embodiment of the invention;
fig. 3 is a flowchart of step 2 in a monitoring video storage and quick query method based on AI analysis in an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
Example 1:
the embodiment provides a hot monitoring video storage and quick query method based on AI analysis, as shown in FIG. 1, comprising the following steps:
step 1: video monitoring is carried out on a target monitoring area based on a monitoring camera to obtain a video stream, the video stream is split based on a preset unit video cache frame, and different preset channels in a streaming media server are subjected to queue cache on average;
step 2: image interception is carried out on the obtained video stream through a streaming media server based on a target time interval, and the intercepted image is transmitted to an AI analysis model for analysis, so that an abnormal image is determined;
Step 3: marking the abnormal image, and transmitting the marked abnormal image back to a corresponding target preset channel in the streaming media server;
step 4: and storing the video cached in the target preset channel based on the streaming media server, and calling and feeding back the stored video to the management terminal according to the query requirement of the management terminal based on the storage result.
In this embodiment, the target monitoring area is set in advance, that is, a place where monitoring by the monitoring camera is required.
In this embodiment, the video monitoring may be a video image obtained by performing video acquisition on the target monitoring area through the monitoring camera, so as to facilitate effective analysis on the abnormal situation of the current video.
In this embodiment, the video stream may be a plurality of continuous frame images obtained after video monitoring of the target monitoring area by the monitoring camera.
In this embodiment, the preset unit video buffer frame is used to represent the number of frames of images that can be buffered in each preset channel in the streaming media server, for example, 200 frames of images can be stored in each preset channel.
In this embodiment, the streaming media server is set in advance, and is used for capturing images and caching the acquired video stream.
In this embodiment, the preset channel is a video buffer channel in the streaming media server, and is not unique.
In this embodiment, the step of buffering the video stream in the queues of different preset channels in the streaming media server based on the preset unit video buffering frame may be performed by splitting the collected video stream according to the number of the preset unit video buffering frames and the preset channels, and storing the video obtained by splitting into the corresponding preset channels, so as to facilitate buffering the video segments corresponding to the abnormal images, and avoid storing a large number of videos.
In this embodiment, the target time interval is a frequency used to characterize the image capture of the video stream received in the streaming media server, thereby facilitating the determination of whether an anomaly exists in the current video stream.
In this embodiment, the image capturing may be to obtain a certain static image in the video stream, where the static image is used to characterize a storage state of each object in the target monitoring area at a certain moment, a current security condition of the target monitoring area, and so on.
In this embodiment, the AI analysis model is obtained after training a preset convolutional network through preset various evaluation indexes, and is used for analyzing the intercepted image, so as to facilitate determining whether the intercepted image is normal under the current various evaluation indexes.
In this embodiment, the abnormal image refers to an image in which the captured image does not satisfy the analysis requirements of the AI analysis model, that is, an image that does not satisfy the evaluation index requirements.
In this embodiment, the marking of the abnormal image may be marking the obtained abnormal image by a preset marking symbol, so that the streaming media server is convenient to lock the video segment corresponding to the marked abnormal image.
In this embodiment, the target preset channel may be a preset channel in the streaming media server where the video corresponding to the intercepted abnormal image is located, so that the video cached in the current preset channel is convenient to store.
In this embodiment, the storage and the preservation of the video cached in the target preset channel based on the streaming media server may be to store a preset number of frame images before and after the corresponding abnormal image, so as to reduce the amount of video to be saved.
In this embodiment, the query requirement may be a video type characterizing that the management terminal needs to query for the presence of an abnormal situation, for example, may be a stranger running into a video, etc.
The beneficial effects of the technical scheme are as follows: the acquired video stream is intercepted through the streaming media server, the intercepted image is input into the AI analysis model for high-efficiency and reliable analysis, and the abnormal image is locked quickly, so that video segments corresponding to the abnormal image are locked conveniently and stored, the efficiency of locating the abnormal video segments is improved, the workload of storing a large amount of videos is reduced, the abnormal video segments are quickly called according to the query requirement of a user, and the work efficiency is improved while the accuracy is ensured.
Example 2:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying a surveillance video based on AI analysis, as shown in fig. 2, in step 1, video surveillance is performed on a target surveillance area based on a surveillance camera, and a video stream is obtained, including:
step 101: acquiring regional characteristics of a target monitoring region, determining effective monitoring cameras in the target monitoring region based on the regional characteristics, and determining terminal identity information of each effective monitoring camera based on a communication protocol of each effective monitoring camera in a local area network;
step 102: transmitting a video acquisition instruction to each effective monitoring camera through a local area network based on the streaming media server, and controlling each effective monitoring camera to acquire video in each monitoring range based on the video acquisition instruction;
step 103: and correlating and binding videos acquired by each effective monitoring camera with terminal identity information of each effective monitoring camera, transmitting the videos acquired by each effective monitoring camera to a streaming media server based on a local area network based on a binding result, and summarizing the received videos based on the streaming media server to obtain a final video stream.
In this embodiment, the region feature may be a region shape of the target monitoring region, thereby facilitating determination of a specific camera that is capable of effectively monitoring the target monitoring region.
In this embodiment, the effective monitoring camera may be a camera that can monitor an effective monitoring picture in an area that can be monitored in the target monitoring area, and is not unique.
In this embodiment, the communication protocol is used to characterize rules to be followed by each effective monitoring camera when communicating in the lan, and communication addresses when each effective monitoring camera communicates with the streaming server, etc.
In this embodiment, the terminal identity information may be reference information capable of characterizing that different effective monitoring cameras are different from other effective monitoring cameras, for example, may be communication address information of each effective monitoring camera, and the like.
In the embodiment, the video acquired by each effective monitoring camera is associated with the terminal identity information of each effective monitoring camera and the binding purpose is to determine the video source of each acquired video, so that the streaming media server analyzes and processes the video.
The beneficial effects of the technical scheme are as follows: the effective monitoring cameras in the target monitoring area are accurately and effectively confirmed according to the area characteristics of the target monitoring area, and are in butt joint with the streaming media server, so that the video collected by each effective monitoring camera is summarized through the streaming media server, the video stream which is finally needed is accurately and effectively obtained, and convenience and guarantee are provided for effectively storing and rapidly inquiring the monitoring video.
Example 3:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying surveillance video based on AI analysis, in step 1, splitting a video stream based on a preset unit video buffer frame, and buffering in queues on different preset channels in a streaming media server on average, including:
acquiring configuration information of a streaming media server, determining the number of preset channels contained in the streaming media server based on the configuration information, determining preset unit video cache frames of the video in unit time of each preset channel based on the configuration information of the streaming media server, and determining splitting position points of the video stream based on the preset unit video cache frames and the preset channel number;
carrying out average splitting on the video stream based on the splitting position point to obtain a video segment, and packaging the video segment obtained after splitting to obtain a target video segment;
determining a starting identifier and a terminating identifier of a cache queue in a preset channel, and caching the target video segment in the cache queue between the starting identifier and the terminating identifier based on the starting identifier and the terminating identifier;
and configuring a data temporary storage thread for a cache queue in each preset channel based on a cache result, and locking a target video segment currently cached in the cache queue based on the data temporary storage thread after the target video segment in the cache queue is cached.
In this embodiment, the configuration information may be a processing rule, a processing manner, and the like for characterizing the video when the streaming media server collects and forwards the video.
In this embodiment, the predetermined number of channels included in the streaming server determined based on the configuration information may be, but is not limited to, a space for buffering video in the streaming server.
In this embodiment, the splitting position point may be a splitting position for characterizing splitting of the video stream, and the obtained video stream may be split into a plurality of video segments by the splitting position point, so as to implement buffering in a preset channel.
In this embodiment, the video segment may be a plurality of short videos obtained by splitting a video stream, which is a part of an original video stream.
In this embodiment, the target video segment may be a video segment that is obtained by encapsulating an obtained video segment and is capable of being cached in a preset channel.
In this embodiment, the start identifier may be start position information characterizing that the target video can be stored in a buffer queue in the preset channel, and is a specific position corresponding to the start position.
In this embodiment, the termination identifier may be end position information characterizing that the target video can be stored in a buffer queue in the preset channel, and is a specific position corresponding to the start position of the mark.
In this embodiment, the temporary data storage thread is set in advance, and is used for temporarily storing the data cached in the cache queue, so as to ensure the reliability of the data and ensure that the data cannot be lost.
The beneficial effects of the technical scheme are as follows: by analyzing the configuration information of the streaming media server, the number of preset channels contained in the streaming media server and the number of videos which can be cached in each preset channel are accurately and effectively determined, so that the obtained video stream is accurately and effectively split, and video segments obtained after the splitting are accurately and reliably cached in the preset channels, the targeted storage of the current video segment is facilitated when different video segments are abnormal, the workload of storing videos is reduced, and the corresponding video segments are conveniently and timely called according to the query requirement of a user.
Example 4:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying a surveillance video based on AI analysis, as shown in fig. 3, in step 2, image capturing is performed on an obtained video stream by a streaming media server based on a preset time interval, including:
Step 201: the method comprises the steps of obtaining an obtained video stream, and playing the obtained video stream based on a preset video playing interface;
step 202: acquiring a target requirement for image interception of the video stream, determining a target time interval for image interception of the video stream based on the target requirement, and generating a periodic image interception instruction based on the target time interval;
step 203: triggering a streaming media server to execute image interception operation based on the periodical image interception instruction, and determining a triggering time point corresponding to the image interception operation in a video stream in play;
step 204: and determining an image to be captured corresponding to the video stream in playing based on the trigger time point, and capturing the image to be captured based on the streaming media server.
In this embodiment, the preset video playing interface is set in advance, so as to play the obtained video stream, thereby facilitating capturing the required image from the video stream.
In this embodiment, the target requirement may be to characterize the frequency with which the video stream is truncated, the number of truncations, etc.
In this embodiment, the periodic image capturing instruction may be generated according to the target time intervals, and is used to generate a corresponding instruction in each target time interval, so as to control the streaming media server to capture images of the video stream in the target time intervals.
In this embodiment, the trigger time point is a specific video playing time point used for representing the video stream and corresponding to the operation of triggering the streaming media server to execute image capturing, so as to achieve capturing of the current image according to the specific video playing time point.
In this embodiment, the image to be intercepted may be an image that needs to be intercepted by the streaming server within the target time interval.
The beneficial effects of the technical scheme are as follows: the method comprises the steps of playing an obtained video stream, analyzing the target requirement of image interception of the video stream, accurately and effectively determining the target time interval of image interception of the video stream, and generating a periodic image interception instruction according to the target time interval, so that the streaming media server is controlled to intercept the video stream through the periodic image interception instruction, the anomaly analysis of the currently intercepted image is facilitated through an AI model, whether the current video segment is abnormal or not is conveniently determined according to an analysis result, and the efficiency of monitoring video storage is improved.
Example 5:
on the basis of embodiment 1, the present embodiment provides a monitoring video storage and quick query method based on AI analysis, in step 2, the intercepted image is transmitted to an AI analysis model for analysis, and abnormal images are determined, including:
Acquiring monitoring types of a target monitoring area, and calling standard monitoring images corresponding to each monitoring type and image evaluation indexes corresponding to each monitoring type from a preset database based on the monitoring types;
performing image scanning on the standard monitoring images of each monitoring type based on a column scanning mode to obtain image characteristics of each standard monitoring image, and determining a mapping relation between the image characteristics and image evaluation indexes;
determining target weight of image characteristics of each standard monitoring image based on the monitoring requirements, and determining index values corresponding to each image evaluation index based on the target weight and the mapping relation;
constructing a monitoring behavior analysis information base based on image evaluation indexes corresponding to each monitoring type, index values corresponding to each image evaluation index and standard monitoring images corresponding to each monitoring type, matching a basic model frame from a preset model base based on monitoring requirements, performing iterative training on the basic model frame based on the monitoring behavior analysis information base, configuring image analysis strategies corresponding to each monitoring type in the basic model frame, and adapting each image analysis strategy to basic configuration parameters of the basic model frame to obtain an AI analysis model;
Inputting the intercepted image into an AI analysis model, extracting image characteristics of the intercepted image based on the AI analysis model, and determining a target monitoring type based on the image characteristics;
invoking a target image evaluation index based on a target monitoring type, and analyzing the intercepted image based on the target image evaluation index to obtain a target difference value of the intercepted image and a standard monitoring image;
when the target difference value is larger than a preset threshold value, judging that the intercepted image is an abnormal image, and meanwhile, adding a type identifier to the abnormal image based on the target monitoring type to obtain a final abnormal image.
In this embodiment, the monitoring type may be a type that characterizes monitoring of the target monitoring area, such as behavior monitoring, article storage status monitoring, and the like.
In this embodiment, the preset database is set in advance, and is used for storing standard monitoring images corresponding to different monitoring types and corresponding image evaluation indexes.
In this embodiment, the standard monitoring image may be an image corresponding to each monitoring type in satisfying the monitoring requirement (i.e., normal state).
In this embodiment, the image evaluation index is a reference basis and a criterion for evaluating whether the intercepted image meets the monitoring requirement of the current monitoring type, and each monitoring type corresponds to a different image evaluation index.
In this embodiment, the column scan mode may be a scan mode for characterizing an image, in order to determine image features corresponding to the image.
In this embodiment, the image feature may be an object included in the standard monitoring image corresponding to each monitoring type and a state corresponding to each object under normal conditions.
In this embodiment, the mapping relationship is used to map the correspondence between different image evaluation indexes and image features, so as to facilitate the construction of an AI analysis model according to the image features and the evaluation indexes.
In this embodiment, the target weight is used to characterize the degree of importance of each image feature in performing anomaly analysis.
In this embodiment, the index value corresponding to the image evaluation index is a specific execution standard for characterizing each image evaluation index in performing anomaly monitoring on the corresponding image feature, that is, when the image feature of the image does not meet the index value requirement of the current image evaluation index, it is determined that the image is an anomaly image.
In this embodiment, the monitoring behavior analysis information base may be configured to aggregate the image evaluation index corresponding to each monitoring type, the index value corresponding to each image evaluation index, and the standard monitoring image corresponding to each monitoring type, so as to facilitate training to obtain a corresponding AI analysis model, and provide data support when in purpose.
In this embodiment, the preset model library is set in advance, and is used to store different basic model frames.
In this embodiment, the basic model framework may be a basic model suitable for constructing an AI analysis model, and training the basic model may result in an AI analysis model capable of analyzing an image.
In this embodiment, the image analysis policy may be to construct an image analysis manner corresponding to each monitoring type in the basic model framework, so as to facilitate integration of image analysis policies corresponding to different monitoring types in the same model.
In this embodiment, the basic configuration parameters may be basic structural parameters of the basic model framework, including the operand of the model, the compatibility of the model with the image analysis strategy, and the like.
In this embodiment, the target monitoring type may be a monitoring type corresponding to the captured image, which is one of preset monitoring types.
In this embodiment, the target image evaluation index may be a specific index suitable for evaluating the currently captured image.
In this embodiment, the target difference value may be a difference degree between an image requirement corresponding to the standard monitoring image and an analysis result obtained by characterizing the intercepted image after the analysis of the AI analysis model.
In this embodiment, the preset threshold is set in advance, and is used to characterize the maximum extent to which the distinction is allowed, and can be adjusted.
In this embodiment, the type identifier is a tag label for marking the monitoring type corresponding to the currently intercepted image.
The beneficial effects of the technical scheme are as follows: the monitoring type of the target monitoring area is determined, the standard monitoring image and the image evaluation index corresponding to each monitoring type are determined according to the monitoring type, the obtained basic model frame is trained according to the standard monitoring image and the image evaluation index, a reliable AI analysis model is obtained, finally, the cut image is subjected to anomaly detection through the AI analysis model, the analysis accuracy and the analysis efficiency of the cut image are improved, the video segment with anomalies is conveniently locked in time according to the analysis result, the efficiency of positioning the anomaly video is improved, and the storage workload of the video is reduced.
Example 6:
on the basis of embodiment 1, the present embodiment provides a monitoring video storage and quick query method based on AI analysis, and in step 3, marking an abnormal image includes:
Reading the abnormal image, determining scene characteristics of the abnormal image, and determining an image identifier of the abnormal image according to the scene characteristics of the abnormal image;
inputting the image identification of the abnormal image into a preset image library for matching, and obtaining a reference image consistent with the image identification;
comparing the abnormal image with a reference image, determining a sensitive area of the abnormal image based on a comparison result, identifying the sensitive area, and determining pixel point distribution of the sensitive area and pixel point edges of the sensitive area;
and marking the abnormal image based on the pixel point distribution of the sensitive area and the pixel point edge of the sensitive area.
In this embodiment, the scene feature may be a background distribution condition of an abnormal image, such as a position distribution of flowers, plants, trees, buildings, and the like in the background.
In this embodiment, the image identification may be an identification of background features characterizing the anomaly image, which is used to distinguish between the anomaly images of different scenes.
In this embodiment, the preset image library may be set in advance, and is used to store reference images corresponding to different image identifiers.
In this embodiment, the reference image may be an image without an abnormality in conformity with the background feature of the abnormality image.
In this embodiment, the sensitive area may be a portion in which the abnormal image and the reference image do not coincide as the sensitive area in the abnormal image.
The beneficial effects of the technical scheme are as follows: through the image identification of the abnormal image, the reference image corresponding to the abnormal image is accurately matched in the preset image library, the identification of the sensitive area in the abnormal image is facilitated, the labeling of the abnormal image is further realized, and the comprehensiveness and the accuracy of the labeling of the abnormal image are improved.
Example 7:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying a surveillance video based on AI analysis, in step 3, the method includes that a marked abnormal image is returned to a corresponding target preset channel in a streaming media server, including:
obtaining a cache address of an abnormal image, and reading channel addresses of different preset channels;
respectively matching the cache address of the abnormal image with the channel addresses of different preset channels, and determining a target preset channel corresponding to the abnormal image based on a matching result;
reading a target address corresponding to a target preset channel, generating a data return request based on the cache address and the target address, sending the data return request to a streaming media server for reading, simultaneously, calling cache logs of the video stream in different preset channels of the streaming media server for queue cache, verifying the data return request based on the cache logs, and judging whether the abnormal image is consistent with the target preset channel;
When the data return request does not have a corresponding matching relation in the cache log, judging that the abnormal image is inconsistent with the target preset channel, outputting a verification failure report, and simultaneously re-matching the target preset channel based on the verification failure report;
when the data feedback request has a corresponding matching relation in the cache log, the abnormal image is judged to be consistent with the target preset channel, meanwhile, the running state of the current target preset channel is read based on the streaming media server, when the current target preset channel outputs an idle signal, a verification success report is generated, and meanwhile, the marked abnormal image is transmitted back to the corresponding target preset channel in the streaming media server based on the verification success report.
In this embodiment, the buffer address may be an address where the monitoring camera captures an image.
In this embodiment, the buffer log may be video information for recording video buffered in each preset channel during the queue buffering process, and the like.
In this embodiment, verification is performed on the data backhaul request based on the cache log, for example, the target address of the target preset channel is determined based on the data backhaul request, the target address is correspondingly matched in the cache log, the video cached in the target preset channel is determined, whether the video corresponding to the abnormal image exists in the target preset channel is read, when the video corresponding to the abnormal image exists in the target preset channel, it is determined that the data backhaul request has a corresponding matching relationship in the cache log, and the verification is passed, or else, it is determined that the data backhaul request does not have a corresponding matching relationship in the cache log, and the verification is not passed.
In this embodiment, the target preset channel output idle signal may be a signal output when the target preset channel does not perform the buffering operation.
The beneficial effects of the technical scheme are as follows: and determining a target preset channel by determining the cache address of the abnormal image and matching, and verifying the data return request according to the streaming media server, so that the accuracy of the abnormal image return is ensured, and the efficiency of monitoring video query is ensured.
Example 8:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying surveillance videos based on AI analysis, in step 4, video cached in a target preset channel is stored based on a streaming media server, including:
determining the video quantity in a target preset channel based on a preset unit video cache frame, and simultaneously determining the video information cached in the target preset channel;
generating a first monitoring factor in the streaming media server based on the video amount in the target preset channel, and generating a second monitoring factor in the streaming media server based on the video information cached in the target preset channel;
when the first monitoring factor and the second monitoring factor are generated, receiving and storing the video cached in the target preset channel based on the streaming media server, and simultaneously, performing first monitoring on the video storage cached in the target preset channel based on the first monitoring factor and performing second monitoring on the video storage cached in the target preset channel based on the second monitoring factor;
When the first monitoring result reaches a first preset standard and the second monitoring result reaches a second preset standard, judging that the streaming media server finishes storing the cache video in the target preset channel;
and after the storage of the video cached in the target preset channel is completed, reading the channel information of the target preset channel, generating a data storage identifier based on the channel information of the target preset channel, constructing a target storage data block based on the data storage identifier, and storing the video cached in the target preset channel into the corresponding target storage data block.
In this embodiment, the first monitoring may be a means for monitoring the number of video stores cached in the target preset channel, that is, the first monitoring factor is a means for monitoring the total amount of the cached video data.
In this embodiment, the second monitoring may be a means for performing content integrity monitoring on the video stored in the target preset channel, that is, the second monitoring factor is a means for monitoring the video content stored in the preset channel.
In this embodiment, the first preset criterion may be that the streaming server receives the entire video amount in the target preset channel.
In this embodiment, the second preset criterion may be that the streaming server presets the video content received in its entirety in the channel.
In this embodiment, the channel information of the target preset channel may be address information of the channel, operation parameters of the channel, and the like.
In this embodiment, the data storage identifier may be a tag token used to distinguish between the transmission of each target preset channel back to the streaming server.
The beneficial effects of the technical scheme are as follows: the streaming media server monitors the video cached in the target preset channel for the first and second monitoring, so that the integrity and accuracy of video receiving are guaranteed, and video data of different preset channels are effectively distinguished by setting the data storage identifier, so that reasonable planning of storage is realized, and quick inquiry is facilitated.
Example 9:
on the basis of embodiment 1, the present embodiment provides a method for storing and rapidly querying a surveillance video based on AI analysis, wherein the method includes that the stored video is retrieved according to a query requirement of a management terminal based on a storage result and fed back to the management terminal, and the method includes:
generating a query requirement based on a management terminal, performing first reading on the query requirement, determining contents to be queried, and simultaneously reading user information of a query user;
inputting the content to be queried to a management terminal to call a user with the queriable authority, matching the user information in the user with the queriable authority, and determining whether the user information is in the user with the queriable authority;
When the user information is in the inquireable right, inputting the inquiry requirement to the streaming media server, simultaneously, carrying out second reading on the inquiry requirement based on the streaming media server, determining inquiry keywords of the inquiry requirement, and simultaneously, generating a video inquiry index based on the inquiry keywords;
and matching the video query indexes in the stored videos, calling the video segment of interest based on the matching result, and feeding the video segment of interest back to the management terminal.
In this embodiment, the content to be queried may be a query target of the querying user.
In this embodiment, the first reading may be to analyze the query requirement to thereby implement determining what the reading requirement needs to read.
In this embodiment, the content to be queried may be a query target of the querying user, that is, a video segment that the user needs to query.
In this embodiment, the user information may be information such as a user name of a user who needs to make a video query.
In this embodiment, the queriable authority user may be a different user having an inquiry authority for the content to be inquired in the management terminal.
In this embodiment, the second reading may be that, after the user corresponding to the query requirement satisfies the query authority, the query requirement is read through the streaming media server, so as to achieve the purpose of calling the video to be queried.
In this embodiment, the query terms may be pieces of data that can characterize the subject matter of the query requirement.
In this embodiment, the video query index may be guide information capable of querying the video.
In this embodiment, the video segment of interest may be a video segment that the user needs to retrieve.
The beneficial effects of the technical scheme are as follows: the query requirement generated by the management terminal is analyzed, so that the content to be queried is locked, the query authority of the querying user is verified, and the video to be queried is queried and retrieved after the authority verification is passed, so that the query and retrieval efficiency and accuracy of the video to be queried are improved.
Example 10:
on the basis of embodiment 1, in step 2, further includes:
when the intercepted image is transmitted to an AI analysis model for analysis, the method comprises the steps of monitoring personnel climbing in a target monitoring area, wherein the specific process is as follows:
acquiring adjacent frame numbers in the intercepted images, acquiring corresponding first target images and second target images based on the adjacent frame numbers of the screenshot images, and respectively determining a human body detection frame of the first target images and a human body detection frame of the second target images;
Calculating a target aspect ratio change rate between the human body detection frame of the first target image and the human body detection frame of the second target image and a centroid change rate of the human body detection frame of the first target image and the second target image based on the change conditions between the human body detection frame of the first target image and the human body detection frame of the second target image;
calculating a target aspect ratio change rate between a human body detection frame of the first target image and a human body detection frame of the second target image according to the following formula;
wherein K represents a target aspect ratio change rate between the human body detection frame of the first target image and the human body detection frame of the second target image; t1 represents a video frame number of the first target image; t2 represents the video frame number of the second target image; l (L) t1 Representing the length of a human body detection frame in the first target image; h t1 Representing the height of a human body detection frame in the second target image; l (L) t2 Representing the length of a human body detection frame in the second target image; h t2 Representing the height of a human body detection frame in the second target image;
calculating the mass center change rate of the human body detection frame of the first target image and the mass center change rate of the second target image according to the following formula;
wherein,,human body detection frame representing first target image and quality of second target image Heart rate of change; y is t1 The centroid coordinates representing the first target image are straight; y is t2 A centroid coordinate value representing a second target image; e represents a natural constant; ln (·) represents a logarithmic function based on e; delta represents an error factor, and the value range is (0.01,0.03);
acquiring a preset length-width ratio change rate threshold and a preset centroid change rate threshold;
comparing the target length-width ratio change rate with a preset length-width ratio change rate threshold, and simultaneously comparing the centroid change rate with a preset centroid change rate threshold to judge whether personnel climbing behaviors exist in a target detection area;
when the target length-width ratio change rate is equal to or greater than a preset length-width ratio change rate threshold value and the centroid change rate is equal to or greater than a preset centroid change rate threshold value, determining that a person climbing behavior exists in the target detection area;
otherwise, judging that no personnel climbing behavior exists in the target detection area.
In this embodiment, the preset aspect ratio change rate threshold and the preset centroid change rate threshold are set in advance, so as to measure whether a person climbing behavior exists in the target detection area, and when the target aspect ratio change rate is equal to or greater than the preset aspect ratio change rate threshold, and simultaneously, the centroid change rate is equal to or greater than the preset centroid change rate threshold, the person climbing behavior can be determined to exist in the target detection area.
The beneficial effects of the technical scheme are as follows: by accurately calculating the change rate of the aspect ratio and the change rate of the mass center of the target, whether personnel climb in the target detection area is effectively measured, and the accuracy and the effectiveness of judgment are improved.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (10)

1. A monitoring video storage and quick query method based on AI analysis is characterized by comprising the following steps:
step 1: video monitoring is carried out on a target monitoring area based on a monitoring camera to obtain a video stream, the video stream is split based on a preset unit video cache frame, and different preset channels in a streaming media server are subjected to queue cache on average;
step 2: image interception is carried out on the obtained video stream through a streaming media server based on a target time interval, and the intercepted image is transmitted to an AI analysis model for analysis, so that an abnormal image is determined;
step 3: marking the abnormal image, and transmitting the marked abnormal image back to a corresponding target preset channel in the streaming media server;
Step 4: and storing the video cached in the target preset channel based on the streaming media server, and calling and feeding back the stored video to the management terminal according to the query requirement of the management terminal based on the storage result.
2. The method for storing and rapidly searching surveillance videos based on AI analysis according to claim 1, wherein in step 1, video surveillance is performed on a target surveillance area based on a surveillance camera to obtain a video stream, comprising:
acquiring regional characteristics of a target monitoring region, determining effective monitoring cameras in the target monitoring region based on the regional characteristics, and determining terminal identity information of each effective monitoring camera based on a communication protocol of each effective monitoring camera in a local area network;
transmitting a video acquisition instruction to each effective monitoring camera through a local area network based on the streaming media server, and controlling each effective monitoring camera to acquire video in each monitoring range based on the video acquisition instruction;
and correlating and binding videos acquired by each effective monitoring camera with terminal identity information of each effective monitoring camera, transmitting the videos acquired by each effective monitoring camera to a streaming media server based on a local area network based on a binding result, and summarizing the received videos based on the streaming media server to obtain a final video stream.
3. The method for storing and rapidly querying surveillance video based on AI analysis of claim 1, wherein in step 1, video streams are split based on a preset unit video buffer frame, and different preset channels in a streaming media server are buffered in a queue on average, comprising:
acquiring configuration information of a streaming media server, determining the number of preset channels contained in the streaming media server based on the configuration information, determining preset unit video cache frames of the video in unit time of each preset channel based on the configuration information of the streaming media server, and determining splitting position points of the video stream based on the preset unit video cache frames and the preset channel number;
carrying out average splitting on the video stream based on the splitting position point to obtain a video segment, and packaging the video segment obtained after splitting to obtain a target video segment;
determining a starting identifier and a terminating identifier of a cache queue in a preset channel, and caching the target video segment in the cache queue between the starting identifier and the terminating identifier based on the starting identifier and the terminating identifier;
and configuring a data temporary storage thread for a cache queue in each preset channel based on a cache result, and locking a target video segment currently cached in the cache queue based on the data temporary storage thread after the target video segment in the cache queue is cached.
4. The method for storing and rapidly querying surveillance video based on AI analysis of claim 1, wherein in step 2, image capturing is performed on the obtained video stream by a streaming media server based on a preset time interval, comprising:
the method comprises the steps of obtaining an obtained video stream, and playing the obtained video stream based on a preset video playing interface;
acquiring a target requirement for image interception of the video stream, determining a target time interval for image interception of the video stream based on the target requirement, and generating a periodic image interception instruction based on the target time interval;
triggering a streaming media server to execute image interception operation based on the periodical image interception instruction, and determining a triggering time point corresponding to the image interception operation in a video stream in play;
and determining an image to be captured corresponding to the video stream in playing based on the trigger time point, and capturing the image to be captured based on the streaming media server.
5. The method for storing and rapidly searching surveillance videos based on AI analysis according to claim 1, wherein in step 2, the captured image is transmitted to an AI analysis model for analysis, and the determining of the abnormal image includes:
Acquiring monitoring types of a target monitoring area, and calling standard monitoring images corresponding to each monitoring type and image evaluation indexes corresponding to each monitoring type from a preset database based on the monitoring types;
performing image scanning on the standard monitoring images of each monitoring type based on a column scanning mode to obtain image characteristics of each standard monitoring image, and determining a mapping relation between the image characteristics and image evaluation indexes;
determining target weight of image characteristics of each standard monitoring image based on the monitoring requirements, and determining index values corresponding to each image evaluation index based on the target weight and the mapping relation;
constructing a monitoring behavior analysis information base based on image evaluation indexes corresponding to each monitoring type, index values corresponding to each image evaluation index and standard monitoring images corresponding to each monitoring type, matching a basic model frame from a preset model base based on monitoring requirements, performing iterative training on the basic model frame based on the monitoring behavior analysis information base, configuring image analysis strategies corresponding to each monitoring type in the basic model frame, and adapting each image analysis strategy to basic configuration parameters of the basic model frame to obtain an AI analysis model;
Inputting the intercepted image into an AI analysis model, extracting image characteristics of the intercepted image based on the AI analysis model, and determining a target monitoring type based on the image characteristics;
invoking a target image evaluation index based on a target monitoring type, and analyzing the intercepted image based on the target image evaluation index to obtain a target difference value of the intercepted image and a standard monitoring image;
when the target difference value is larger than a preset threshold value, judging that the intercepted image is an abnormal image, and meanwhile, adding a type identifier to the abnormal image based on the target monitoring type to obtain a final abnormal image.
6. The method for storing and rapidly searching surveillance videos based on AI analysis of claim 1, wherein in step 3, the abnormal image is marked, comprising:
reading the abnormal image, determining scene characteristics of the abnormal image, and determining an image identifier of the abnormal image according to the scene characteristics of the abnormal image;
inputting the image identification of the abnormal image into a preset image library for matching, and obtaining a reference image consistent with the image identification;
comparing the abnormal image with a reference image, determining a sensitive area of the abnormal image based on a comparison result, identifying the sensitive area, and determining pixel point distribution of the sensitive area and pixel point edges of the sensitive area;
And marking the abnormal image based on the pixel point distribution of the sensitive area and the pixel point edge of the sensitive area.
7. The method for storing and rapidly querying surveillance videos based on AI analysis according to claim 1, wherein in step 3, the marked abnormal image is returned to a corresponding target preset channel in the streaming media server, comprising:
obtaining a cache address of an abnormal image, and reading channel addresses of different preset channels;
respectively matching the cache address of the abnormal image with the channel addresses of different preset channels, and determining a target preset channel corresponding to the abnormal image based on a matching result;
reading a target address corresponding to a target preset channel, generating a data return request based on the cache address and the target address, sending the data return request to a streaming media server for reading, simultaneously, calling cache logs of the video stream in different preset channels of the streaming media server for queue cache, verifying the data return request based on the cache logs, and judging whether the abnormal image is consistent with the target preset channel;
when the data return request does not have a corresponding matching relation in the cache log, judging that the abnormal image is inconsistent with the target preset channel, outputting a verification failure report, and simultaneously re-matching the target preset channel based on the verification failure report;
When the data feedback request has a corresponding matching relation in the cache log, the abnormal image is judged to be consistent with the target preset channel, meanwhile, the running state of the current target preset channel is read based on the streaming media server, when the current target preset channel outputs an idle signal, a verification success report is generated, and meanwhile, the marked abnormal image is transmitted back to the corresponding target preset channel in the streaming media server based on the verification success report.
8. The method for storing and rapidly querying surveillance videos based on AI analysis according to claim 1, wherein in step 4, video cached in a target preset channel is stored based on a streaming media server, comprising:
determining the video quantity in a target preset channel based on a preset unit video cache frame, and simultaneously determining the video information cached in the target preset channel;
generating a first monitoring factor in the streaming media server based on the video amount in the target preset channel, and generating a second monitoring factor in the streaming media server based on the video information cached in the target preset channel;
when the first monitoring factor and the second monitoring factor are generated, receiving and storing the video cached in the target preset channel based on the streaming media server, and simultaneously, performing first monitoring on the video storage cached in the target preset channel based on the first monitoring factor and performing second monitoring on the video storage cached in the target preset channel based on the second monitoring factor;
When the first monitoring result reaches a first preset standard and the second monitoring result reaches a second preset standard, judging that the streaming media server finishes storing the cache video in the target preset channel;
and after the storage of the video cached in the target preset channel is completed, reading the channel information of the target preset channel, generating a data storage identifier based on the channel information of the target preset channel, constructing a target storage data block based on the data storage identifier, and storing the video cached in the target preset channel into the corresponding target storage data block.
9. The method for storing and rapidly querying surveillance videos based on AI analysis according to claim 1, wherein retrieving and feeding back the stored videos to the management terminal according to the query request of the management terminal based on the storage result comprises:
generating a query requirement based on a management terminal, performing first reading on the query requirement, determining contents to be queried, and simultaneously reading user information of a query user;
inputting the content to be queried to a management terminal to call a user with the queriable authority, matching the user information in the user with the queriable authority, and determining whether the user information is in the user with the queriable authority;
When the user information is in the inquireable right, inputting the inquiry requirement to the streaming media server, simultaneously, carrying out second reading on the inquiry requirement based on the streaming media server, determining inquiry keywords of the inquiry requirement, and simultaneously, generating a video inquiry index based on the inquiry keywords;
and matching the video query indexes in the stored videos, calling the video segment of interest based on the matching result, and feeding the video segment of interest back to the management terminal.
10. The method for storing and rapidly searching surveillance videos based on AI analysis of claim 1, wherein in step 2, further comprises:
when the intercepted image is transmitted to an AI analysis model for analysis, the method comprises the steps of monitoring personnel climbing in a target monitoring area, wherein the specific process is as follows:
acquiring adjacent frame numbers in the intercepted images, acquiring corresponding first target images and second target images based on the adjacent frame numbers of the screenshot images, and respectively determining a human body detection frame of the first target images and a human body detection frame of the second target images;
calculating a target aspect ratio change rate between the human body detection frame of the first target image and the human body detection frame of the second target image and a centroid change rate of the human body detection frame of the first target image and the second target image based on the change conditions between the human body detection frame of the first target image and the human body detection frame of the second target image;
Acquiring a preset length-width ratio change rate threshold and a preset centroid change rate threshold;
comparing the target length-width ratio change rate with a preset length-width ratio change rate threshold, and simultaneously comparing the centroid change rate with a preset centroid change rate threshold to judge whether personnel climbing behaviors exist in a target detection area;
when the target length-width ratio change rate is equal to or greater than a preset length-width ratio change rate threshold value and the centroid change rate is equal to or greater than a preset centroid change rate threshold value, determining that a person climbing behavior exists in the target detection area;
otherwise, judging that no personnel climbing behavior exists in the target detection area.
CN202310542271.9A 2023-05-15 2023-05-15 Monitoring video storage and quick query method based on AI analysis Active CN116610834B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310542271.9A CN116610834B (en) 2023-05-15 2023-05-15 Monitoring video storage and quick query method based on AI analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310542271.9A CN116610834B (en) 2023-05-15 2023-05-15 Monitoring video storage and quick query method based on AI analysis

Publications (2)

Publication Number Publication Date
CN116610834A true CN116610834A (en) 2023-08-18
CN116610834B CN116610834B (en) 2024-04-12

Family

ID=87679328

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310542271.9A Active CN116610834B (en) 2023-05-15 2023-05-15 Monitoring video storage and quick query method based on AI analysis

Country Status (1)

Country Link
CN (1) CN116610834B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117061794A (en) * 2023-10-12 2023-11-14 深圳市迪威泰实业有限公司 Bank monitoring camera with data encryption protection function and monitoring method thereof
CN117155902A (en) * 2023-08-24 2023-12-01 深圳市乔安科技有限公司 Method, system, device and storage medium for detecting video image input equipment
CN117294903A (en) * 2023-10-25 2023-12-26 深蓝(深圳)感知科技有限公司 Multi-channel video stream key frame marking and stream pushing method
CN117931098A (en) * 2024-03-22 2024-04-26 江西云眼视界科技股份有限公司 Edge side event storage method, system, storage medium and computer

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102025714A (en) * 2009-09-17 2011-04-20 ***通信集团公司 Implementation method, device and system for multimedia broadcast multicast service (MBMS)
US20140085480A1 (en) * 2008-03-03 2014-03-27 Videolq, Inc. Content-aware computer networking devices with video analytics for reducing video storage and video communication bandwidth requirements of a video surveillance network camera system
CN104077522A (en) * 2014-06-30 2014-10-01 江苏华大天益电力科技有限公司 Process integrity detection method of operation system
CN110443109A (en) * 2019-06-11 2019-11-12 万翼科技有限公司 Abnormal behaviour monitor processing method, device, computer equipment and storage medium
CN111949737A (en) * 2020-09-01 2020-11-17 宁波富万信息科技有限公司 Block chain-based data integrity verification method, terminal device and storage medium
CN112511650A (en) * 2020-12-22 2021-03-16 湖南新云网科技有限公司 Video uploading method, device, equipment and readable storage medium
WO2021151279A1 (en) * 2020-06-17 2021-08-05 平安科技（深圳）有限公司 Method and apparatus for cloud monitoring based on edge computing, electronic device, and storage medium
CN114222166A (en) * 2021-09-29 2022-03-22 重庆创通联达智能技术有限公司 Multi-path video code stream real-time processing and on-screen playing method and related system
CN115082825A (en) * 2022-06-16 2022-09-20 中新国际联合研究院 Video-based real-time human body falling detection and alarm method and device
CN115984967A (en) * 2023-01-05 2023-04-18 北京轩宇空间科技有限公司 Human body falling detection method, device and system based on deep learning

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140085480A1 (en) * 2008-03-03 2014-03-27 Videolq, Inc. Content-aware computer networking devices with video analytics for reducing video storage and video communication bandwidth requirements of a video surveillance network camera system
CN102025714A (en) * 2009-09-17 2011-04-20 ***通信集团公司 Implementation method, device and system for multimedia broadcast multicast service (MBMS)
CN104077522A (en) * 2014-06-30 2014-10-01 江苏华大天益电力科技有限公司 Process integrity detection method of operation system
CN110443109A (en) * 2019-06-11 2019-11-12 万翼科技有限公司 Abnormal behaviour monitor processing method, device, computer equipment and storage medium
WO2021151279A1 (en) * 2020-06-17 2021-08-05 平安科技（深圳）有限公司 Method and apparatus for cloud monitoring based on edge computing, electronic device, and storage medium
CN111949737A (en) * 2020-09-01 2020-11-17 宁波富万信息科技有限公司 Block chain-based data integrity verification method, terminal device and storage medium
CN112511650A (en) * 2020-12-22 2021-03-16 湖南新云网科技有限公司 Video uploading method, device, equipment and readable storage medium
CN114222166A (en) * 2021-09-29 2022-03-22 重庆创通联达智能技术有限公司 Multi-path video code stream real-time processing and on-screen playing method and related system
CN115082825A (en) * 2022-06-16 2022-09-20 中新国际联合研究院 Video-based real-time human body falling detection and alarm method and device
CN115984967A (en) * 2023-01-05 2023-04-18 北京轩宇空间科技有限公司 Human body falling detection method, device and system based on deep learning

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117155902A (en) * 2023-08-24 2023-12-01 深圳市乔安科技有限公司 Method, system, device and storage medium for detecting video image input equipment
CN117061794A (en) * 2023-10-12 2023-11-14 深圳市迪威泰实业有限公司 Bank monitoring camera with data encryption protection function and monitoring method thereof
CN117061794B (en) * 2023-10-12 2024-02-13 深圳市迪威泰实业有限公司 Bank monitoring camera with data encryption protection function and monitoring method thereof
CN117294903A (en) * 2023-10-25 2023-12-26 深蓝(深圳)感知科技有限公司 Multi-channel video stream key frame marking and stream pushing method
CN117931098A (en) * 2024-03-22 2024-04-26 江西云眼视界科技股份有限公司 Edge side event storage method, system, storage medium and computer
CN117931098B (en) * 2024-03-22 2024-06-07 江西云眼视界科技股份有限公司 Edge side event storage method, system, storage medium and computer

Also Published As

Publication number Publication date
CN116610834B (en) 2024-04-12

Similar Documents

Publication Publication Date Title
CN116610834B (en) Monitoring video storage and quick query method based on AI analysis
CN111835585B (en) Inspection method and device for Internet of things equipment, computer equipment and storage medium
CN111917707B (en) Grading method, client and system in network security level protection
CN110738178A (en) Garden construction safety detection method and device, computer equipment and storage medium
CN111934906B (en) Artificial intelligence assessment method, client and system for grade protection
CN111429304A (en) Food safety supervision platform
CN111444072A (en) Client abnormality identification method and device, computer equipment and storage medium
CN113269081A (en) System and method for automatic personnel identification and video track query
CN111586432B (en) Method and device for determining air-broadcast live broadcast room, server and storage medium
CN110991979A (en) Attendance management method, device and equipment for intelligent construction site and computer storage medium
CN113190700B (en) Face snapshot, screening and storage method and system for real estate transaction
CN116594801A (en) Fault analysis system for big data cloud computing
CN117319066A (en) Specific network-oriented risk assessment method and device for Internet of things equipment
CN115660610A (en) Decentralized cooperative office system and method
CN108388672A (en) Lookup method, device and the computer readable storage medium of video
CN110225009A (en) It is a kind of that user's detection method is acted on behalf of based on communication behavior portrait
CN117539739A (en) User continuous behavior anomaly monitoring method based on double features
CN108234484A (en) For tracing the wooden horse source traceability system of the computer readable storage medium in wooden horse source and the application medium
CN112148920A (en) Data management method
US9742641B2 (en) System and method for identifying real users behind application servers
CN112039936A (en) Data transmission method, first data processing equipment and monitoring system
CN212990113U (en) Data acquisition equipment is dressed to intelligence
Ze Design and implementation of core modules of WEB application vulnerability detection model
KR20020081561A (en) System and method for golf swing selfclinic
CN113407502B (en) Data flow classification management method and 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
GR01 Patent grant
GR01 Patent grant