Intelligent monitoring analysis method and device based on context awareness, equipment and medium

Abstract

The invention discloses an intelligent monitoring analysis method, device and equipment based on context awareness and a medium, and belongs to the field of intelligent monitoring, and the method comprises the steps: obtaining video data of a monitoring scene, and extracting time attribute information, space attribute information, object attribute information and environment attribute information of the video data in parallel to construct context information; matching in a preset rule base by taking the context information as a query condition to obtain a corresponding algorithm scheduling rule; selecting a plurality of monitoring analysis algorithms from a preset algorithm library according to the algorithm scheduling rule, and instantiating each monitoring analysis algorithm according to a workflow template corresponding to the algorithm scheduling rule to generate an analysis workflow; and scheduling the computing resources to execute the analysis workflow, and outputting a monitoring analysis result obtained after the analysis workflow processes the video data. Therefore, by implementing the method, the resource utilization rate and the scene adaptability of a monitoring analysis algorithm can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring, and in particular to a context-aware intelligent monitoring and analysis method, device, equipment, and medium. Background Technology

[0002] To balance the limited computing power of edge devices with the complex demands of video analytics, existing technologies generally employ multi-level cascaded detection schemes. A typical workflow for this scheme is as follows: one or more primary algorithms, such as motion detection algorithms, reside on the device to detect motion changes in the scene in real time. When a primary algorithm detects a moving target, it triggers one or more intermediate algorithms, such as human detection algorithms, to filter out non-target interference such as leaves, shadows, or animals. If the intermediate algorithms confirm a valid target (such as a human figure), they further trigger advanced algorithms, such as face recognition or behavior analysis, for in-depth analysis. The entire process is a fixed, linear trigger chain, like a multi-level filter, progressively filtering and ultimately completing the analysis task.

[0003] While the above scheme saves computing power to some extent, its inherent architecture brings the following significant drawbacks: First, the cascaded scheme is a pre-defined, fixed pipeline (e.g., movement-human-face). It cannot flexibly adjust to the complexities of the actual scene (such as time, weather, lighting, and regional importance). For example, in extremely poor lighting conditions at night, it will still mechanically trigger the face recognition algorithm, resulting in a very low success rate and wasting valuable computing resources. Second, due to a lack of contextual understanding, the system often loads and runs unnecessary algorithms. Each trigger starts a fixed analysis chain, regardless of whether the current scene requires all the links in the chain. This leads to an "average distribution" of computing power rather than "precise allocation on demand." Third, basic algorithms (such as motion detection) are very sensitive and prone to generating a large number of false alarms due to environmental changes (light and shadow or swaying leaves). These false alarms will unnecessarily trigger subsequent complex analysis processes, not only consuming resources but also potentially generating invalid alarms, reducing the overall reliability of the system. Fourth, adding new analytical functions or modifying analytical logic usually requires redesigning and deploying the entire cascading process, which is not flexible enough and difficult to cope with rapidly changing application requirements. Summary of the Invention

[0004] This invention provides a context-aware intelligent monitoring and analysis method, device, equipment, and medium that can improve the resource utilization and scenario adaptability of monitoring and analysis algorithms.

[0005] This invention provides a context-aware intelligent monitoring and analysis method, including: Acquire video data from the monitored scene, and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information; The context information is used as a query condition to match the results in a pre-set rule base, thereby obtaining the corresponding algorithm scheduling rule. According to the algorithm scheduling rules, multiple monitoring and analysis algorithms are selected from the preset algorithm library, and each monitoring and analysis algorithm is instantiated according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow. The system schedules computing resources to execute the analysis workflow and outputs the monitoring and analysis results obtained after the analysis workflow processes the video data.

[0006] This invention, through acquiring and extracting multi-dimensional attributes of time, space, objects, and environment to construct contextual information, comprehensively perceives the real-time status of the monitoring scene, providing a basis for subsequent intelligent decision-making. By matching the contextual information with a pre-set rule base to obtain algorithm scheduling rules, the understanding of the scene can be transformed into analysis instructions. By selecting algorithms according to the scheduling rules and instantiating them according to the corresponding workflow templates to generate analysis workflows, the analysis workflows can be dynamically assembled according to scene requirements, achieving precise matching and combination of algorithm resources. By scheduling resources to execute the analysis workflows and outputting processing results, the analysis workflows can be ensured to be executed efficiently, ultimately producing monitoring and analysis results. Compared with the problems of rigid decision-making, resource waste, and high false alarm rates in existing technologies, this application can improve the resource utilization and scene adaptability of monitoring and analysis algorithms.

[0007] Furthermore, the process of acquiring video data from the monitored scene and extracting, in parallel, temporal, spatial, object, and environmental attribute information from the video data to construct context information includes: Video data is input to a distributor, which then pushes the video data to a temporal attribute analyzer, a spatial attribute analyzer, an object attribute analyzer, and an environmental attribute analyzer. The time attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data are extracted by the time attribute analyzer, spatial attribute analyzer, object attribute analyzer, and environmental attribute analyzer, respectively. The context aggregator aggregates the time attribute information, spatial attribute information, object attribute information, and environmental attribute information to obtain context information.

[0008] The embodiments of the present invention employ a distributor to push information in parallel to each dedicated attribute analyzer, and an aggregator to aggregate the information, thereby achieving efficient extraction and unified construction of multi-dimensional contextual information and improving the processing efficiency of the perception process.

[0009] Further, the step of using the context information as a query condition to perform matching in a preset rule base to obtain the corresponding algorithm scheduling rule includes: The context information is matched against each predefined rule in the rule base, and the rule whose matching result meets the preset matching conditions is determined as the algorithm scheduling rule.

[0010] This invention enables precise rule triggering by matching context information with preset rules and determining the rules that meet the conditions, thus ensuring that scheduling decisions are highly relevant to the current scenario.

[0011] Furthermore, the step of selecting multiple monitoring and analysis algorithms from a pre-set algorithm library according to the algorithm scheduling rules includes: Based on the algorithm identifier indicated in the algorithm scheduling rules, select multiple corresponding monitoring and analysis algorithms from the algorithm library.

[0012] This invention enables accurate access to algorithm resources by selecting algorithms based on the identifiers indicated in the algorithm scheduling rules, thus ensuring the precision of component calls during dynamic orchestration.

[0013] Further, the step of instantiating each of the monitoring and analysis algorithms according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow includes: According to the algorithm scheduling rules, the corresponding workflow template is invoked; Based on the predefined algorithm combination logic in the workflow template, each monitoring and analysis algorithm is instantiated and connected according to the algorithm dependency relationship defined by the algorithm combination logic to obtain the analysis workflow; wherein, the analysis workflow is represented as a directed acyclic graph in memory.

[0014] This invention enables the structured definition of the analysis process by calling a workflow template and instantiating the algorithm according to its predefined combinatorial logic, and connecting them into a directed acyclic graph according to dependencies. This lays the foundation for subsequent intelligent scheduling.

[0015] Furthermore, the scheduled computing resources execute the analysis workflow and output the monitoring analysis results obtained after processing the video data by the analysis workflow, including: The analysis workflow is loaded into the computing unit, and the directed acyclic graph is topologically sorted to obtain the execution sequence of each algorithm instance; The computing unit executes the monitoring and analysis algorithms corresponding to each algorithm instance in sequence according to the execution sequence; Obtain the processing result output by the final-level algorithm instance in the execution sequence, and output the processing result as the monitoring and analysis result.

[0016] This invention, through topological sorting of a directed acyclic graph to determine and sequentially execute node sequences, can parse and process dependencies between algorithms, ensuring that the analysis workflow is executed efficiently.

[0017] Furthermore, after scheduling computing resources to execute the analysis workflow and outputting the monitoring analysis results obtained by the analysis workflow after processing the video data, the method further includes: The context information is updated in real time. When the context information meets preset conditions, the analysis workflow is unloaded and computing resources are released.

[0018] This invention, by unloading the workflow and releasing resources when conditions are met based on updated context information, can avoid resource idleness and waste, and improve overall resource utilization.

[0019] Another embodiment of the present invention provides a context-aware intelligent monitoring and analysis device, including: a context information module, an algorithm scheduling rule module, an analysis workflow module, and a monitoring and analysis result module; The context information module is used to acquire video data of the monitoring scene and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information; The algorithm scheduling rule module is used to use the context information as a query condition to match it in a preset rule base to obtain the corresponding algorithm scheduling rule. The analysis workflow module is used to select multiple monitoring and analysis algorithms from a preset algorithm library according to the algorithm scheduling rules, and instantiate each monitoring and analysis algorithm according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow. The monitoring and analysis results module is used to schedule computing resources to execute the analysis workflow and output the monitoring and analysis results obtained after the analysis workflow processes the video data.

[0020] Another embodiment of the present invention provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps of a context-aware intelligent monitoring and analysis method as described in the present invention.

[0021] Another embodiment of the present invention provides a computer-readable storage medium item, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to perform steps of a context-aware intelligent monitoring and analysis method as described in the present invention. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating an embodiment of the context-aware intelligent monitoring and analysis method provided by the present invention. Figure 2 This is a schematic diagram of the structure of an embodiment of the context-aware intelligent monitoring and analysis system provided by the present invention; Figure 3 A schematic diagram of the structure of one embodiment of the context building module provided by the present invention; Figure 4 A flowchart illustrating another embodiment of the context-aware intelligent monitoring and analysis method provided by the present invention; Figure 5 This is a schematic diagram of one embodiment of the context-aware intelligent monitoring and analysis device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0025] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0028] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0029] See Figure 1 To address the problems of rigid decision-making, resource waste, and high false alarm rates in existing technologies, an embodiment of the present invention provides a context-aware intelligent monitoring and analysis method, including steps S101 to S104: Step S101: Acquire video data of the monitored scene, and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information.

[0030] It should be noted that acquiring video data from a monitored scene and extracting its temporal, spatial, object, and environmental attributes in parallel to construct contextual information involves: a distribution unit synchronously pushing the video data to multiple independent attribute analyzers. Specifically, the temporal attribute analyzer parses timestamps to determine time periods and date attributes; the spatial attribute analyzer determines the location of moving targets based on preset area coordinates; the object attribute analyzer uses motion detection and lightweight classification to identify target types and track their trajectories; the environmental attribute analyzer calculates global image statistical features to assess illumination and sharpness; and finally, a context aggregator integrates the outputs from all dimensions to form structured contextual information.

[0031] Preferably, the step of acquiring video data of the monitored scene and extracting temporal attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data in parallel to construct context information includes: Video data is input to a distributor, which then pushes the video data to a temporal attribute analyzer, a spatial attribute analyzer, an object attribute analyzer, and an environmental attribute analyzer. The time attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data are extracted by the time attribute analyzer, spatial attribute analyzer, object attribute analyzer, and environmental attribute analyzer, respectively. The context aggregator aggregates the time attribute information, spatial attribute information, object attribute information, and environmental attribute information to obtain context information.

[0032] In one embodiment, the processing steps of the time attribute analyzer include: obtaining the current timestamp from the system clock; determining and generating time attribute information based on the timestamp using a calendar algorithm; wherein, one form of time attribute information is: dividing 24 hours into several time periods, such as morning, forenoon, afternoon, evening, night, and late night; another form of time attribute information is: dividing a week into weekdays and weekends; yet another form of time attribute information is: returning yes or no by querying a preset list of holidays.

[0033] In one embodiment, the processing steps of the spatial attribute analyzer include: loading the polygon coordinates of user- or system-preset regions of interest (ROIs) and restricted areas from the configuration file; receiving the initial motion detection results (i.e., the bounding boxes of moving blocks) from the object attribute analyzer; for each moving block, using algorithms such as ray casting to determine whether its centroid falls within a certain ROI or restricted area; generating spatial attribute information; wherein, one form of spatial attribute information is: a dictionary recording the active state of each ROI, such as {"gateway": true, "parking lot": false}, and another form of spatial attribute information is: a Boolean value, which is true as long as any moving block enters the restricted area.

[0034] In one embodiment, the processing steps of the object attribute analyzer include: using frame difference or background subtraction model (such as MOG2) to separate all moving pixel regions from the static background and generate bounding boxes of the moving target; performing a small scaling on the bounding boxes and then feeding them into a binary neural network (BNN) or a mini convolutional network; tracking the same bounding box in several consecutive frames (through IOU matching) to generate object attribute information; wherein, the object attribute information includes motion speed, motion trajectory vector, and the width and height of the bounding box.

[0035] In one embodiment, the processing steps of the environmental attribute analyzer include: calculating the grayscale average value or brightness histogram of the entire frame image, mapping the calculation result to several preset levels, such as bright light, normal, dim light, and extremely dark light; calculating the variance of the image using the Laplacian operator, the larger the variance, the clearer the image; and mapping the result to several preset levels, such as clear, slightly blurry, and severely blurry.

[0036] In one embodiment, the context aggregator collects the output of all attribute analyzers at a high frequency (e.g., 5-10 times per second) and packages it into a uniform structured data object (e.g., JSON or Protobuf format).

[0037] Step S102: Use the context information as a query condition and match it in a preset rule base to obtain the corresponding algorithm scheduling rule.

[0038] It should be noted that using the context information as a query condition and matching it in a pre-defined rule base to obtain the corresponding algorithm scheduling rule means: taking the real-time generated context information as input and matching it with the predefined rules in the rule base; when the context information meets the triggering condition of a certain rule, then that rule is used as the algorithm scheduling rule; this rule not only indicates the analysis algorithm to be activated, but may also be associated with a specific workflow template, thereby forming an executable scheduling instruction in the current scenario.

[0039] Preferably, the step of using the context information as a query condition to perform matching in a preset rule base to obtain the corresponding algorithm scheduling rule includes: The context information is matched against each predefined rule in the rule base, and the rule whose matching result meets the preset matching conditions is determined as the algorithm scheduling rule.

[0040] In one embodiment, the rule base stores a large number of rules representing the "context-algorithm" mapping relationship. These rules can be preset by experts or automatically generated and optimized through machine learning. Rules include various forms, such as: IF {lighting: weak} AND {object: humanoid} THEN {disable: face recognition algorithm}; IF {time: weekday daytime} AND {space: entrance} THEN {load: face recognition, vehicle recognition}; IF {object trajectory: loitering} AND {duration: >30 seconds} THEN {load: abnormal behavior analysis algorithm}. Each rule is also associated with a specific workflow template, which precisely describes the algorithm combination, execution order, and data flow required to complete a specific analysis task in a data-driven manner.

[0041] Step S103: According to the algorithm scheduling rules, select multiple monitoring and analysis algorithms from the preset algorithm library, and instantiate each monitoring and analysis algorithm according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow.

[0042] It should be noted that, according to the algorithm scheduling rules, selecting multiple monitoring and analysis algorithms from the pre-set algorithm library means: obtaining a clear list of algorithm requirements from the results of rule matching, and then finding and preparing the corresponding multiple algorithm modules from the pre-stored, independently callable algorithm library based on this list.

[0043] Preferably, the step of selecting multiple monitoring and analysis algorithms from a preset algorithm library according to the algorithm scheduling rules includes: Based on the algorithm identifier indicated in the algorithm scheduling rules, select multiple corresponding monitoring and analysis algorithms from the algorithm library.

[0044] In one embodiment, the algorithm library stores various independent, dynamically loadable and unloadable heterogeneous monitoring and analysis algorithm modules, such as face recognition, vehicle recognition, license plate recognition, behavior analysis, sound detection, and target tracking. To achieve plug-and-play and dynamic combination of algorithms, each algorithm module in the library follows a set of predefined standardized interfaces, which uniformly define the methods for algorithm initialization, data processing, parameter configuration, and resource release.

[0045] It should be noted that, according to the workflow template corresponding to the algorithm scheduling rules, instantiating each of the monitoring and analysis algorithms to generate an analysis workflow means that the algorithm orchestration engine, in a data-driven manner, dynamically creates algorithm instances and establishes data transfer relationships between them based on the workflow template obtained through rule matching, thereby assembling a computational task flow specifically for the current monitoring scenario in real time on-site. For example, for the "nighttime loitering in restricted areas" scenario, the generated analysis workflow might be a human detection algorithm + a target tracking algorithm + a loitering behavior analysis algorithm, but would not include a face recognition algorithm, which is ineffective at night.

[0046] Preferably, the step of instantiating each of the monitoring and analysis algorithms according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow includes: According to the algorithm scheduling rules, the corresponding workflow template is invoked; Based on the predefined algorithm combination logic in the workflow template, each monitoring and analysis algorithm is instantiated and connected according to the algorithm dependency relationship defined by the algorithm combination logic to obtain the analysis workflow; wherein, the analysis workflow is represented as a directed acyclic graph in memory.

[0047] In one embodiment, the algorithm orchestration engine sends instructions to the operating system or container management platform to dynamically load and instantiate the corresponding algorithm modules from the algorithm library, and construct a directed acyclic graph in memory, where nodes represent algorithm instances (including algorithm name, instance ID and initialization parameters) and edges represent data flow (i.e., which node's output should be used as input to which node).

[0048] Step S104: Schedule computing resources to execute the analysis workflow and output the monitoring analysis results obtained after the analysis workflow processes the video data.

[0049] It should be noted that scheduling computing resources to execute the analysis workflow and outputting the monitoring and analysis results obtained after the analysis workflow processes the video data means: allocating appropriate CPU / GPU or other hardware computing power to the dynamically generated analysis workflow, driving each algorithm in the analysis workflow to process the video data sequentially according to predetermined logic, and taking the valid conclusions produced by the final algorithm as the valid output of this monitoring and analysis.

[0050] Preferably, the scheduling computing resources execute the analysis workflow and output the monitoring analysis results obtained after the analysis workflow processes the video data, including: The analysis workflow is loaded into the computing unit, and the directed acyclic graph is topologically sorted to obtain the execution sequence of each algorithm instance; The computing unit executes the monitoring and analysis algorithms corresponding to each algorithm instance in sequence according to the execution sequence; Obtain the processing result output by the final-level algorithm instance in the execution sequence, and output the processing result as the monitoring and analysis result.

[0051] In one embodiment, the system performs topological sorting on the directed acyclic graph, schedules CPU / GPU resources in sequence to execute each algorithm instance, and data flows between nodes through pre-established channels.

[0052] Preferably, after the scheduling computing resources execute the analysis workflow and output the monitoring analysis results obtained by the analysis workflow after processing the video data, the method further includes: The context information is updated in real time. When the context information meets preset conditions, the analysis workflow is unloaded and computing resources are released.

[0053] It should be noted that updating the context information in real time, and unloading the analysis workflow and releasing computing resources when the context information meets the preset conditions, means that during the execution of the monitoring task, the scene state changes are continuously monitored. Once the target task is detected to have ended or the scene has returned to normal, the cleanup mechanism is automatically triggered: all algorithms involved in the current analysis workflow are unloaded and the memory and computing power they occupy are immediately released, so that the system returns to a low-power standby monitoring state.

[0054] In one embodiment, when the context information indicates that the event has ended, the algorithm orchestration engine destroys the directed acyclic graph, unloads the algorithm instance, and releases resources.

[0055] This invention, through acquiring and extracting multi-dimensional attributes of time, space, objects, and environment to construct contextual information, can comprehensively perceive the real-time status of the monitoring scene, providing a basis for subsequent intelligent decision-making. By matching the contextual information with a pre-set rule base to obtain algorithm scheduling rules, the understanding of the scene can be transformed into analysis instructions. By selecting algorithms according to the scheduling rules and instantiating them according to the corresponding workflow template to generate analysis workflows, the analysis workflows can be dynamically assembled according to scene requirements, achieving precise matching and combination of algorithm resources. By scheduling resources to execute the analysis workflows and output processing results, the analysis workflows can be executed efficiently, ultimately producing monitoring analysis results.

[0056] Optionally, in this embodiment of the invention, the step of acquiring video data of the monitored scene and extracting in parallel the temporal attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data to construct context information includes: Video data is input to a distributor, which then pushes the video data to a temporal attribute analyzer, a spatial attribute analyzer, an object attribute analyzer, and an environmental attribute analyzer. The time attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data are extracted by the time attribute analyzer, spatial attribute analyzer, object attribute analyzer, and environmental attribute analyzer, respectively. The context aggregator aggregates the time attribute information, spatial attribute information, object attribute information, and environmental attribute information to obtain context information.

[0057] The embodiments of the present invention employ a distributor to push information in parallel to each dedicated attribute analyzer, and an aggregator to aggregate the information, thereby achieving efficient extraction and unified construction of multi-dimensional contextual information and improving the processing efficiency of the perception process.

[0058] Optionally, in this embodiment of the invention, the step of using the context information as a query condition to perform matching in a preset rule base to obtain the corresponding algorithm scheduling rule includes: The context information is matched against each predefined rule in the rule base, and the rule whose matching result meets the preset matching conditions is determined as the algorithm scheduling rule.

[0059] This invention enables precise rule triggering by matching context information with preset rules and determining the rules that meet the conditions, thus ensuring that scheduling decisions are highly relevant to the current scenario.

[0060] Optionally, in this embodiment of the invention, selecting multiple monitoring and analysis algorithms from a preset algorithm library according to the algorithm scheduling rules includes: Based on the algorithm identifier indicated in the algorithm scheduling rules, select multiple corresponding monitoring and analysis algorithms from the algorithm library.

[0061] This invention enables accurate access to algorithm resources by selecting algorithms based on the identifiers indicated in the algorithm scheduling rules, thus ensuring the precision of component calls during dynamic orchestration.

[0062] Optionally, in this embodiment of the invention, the step of instantiating each of the monitoring and analysis algorithms according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow includes: According to the algorithm scheduling rules, the corresponding workflow template is invoked; Based on the predefined algorithm combination logic in the workflow template, each monitoring and analysis algorithm is instantiated and connected according to the algorithm dependency relationship defined by the algorithm combination logic to obtain the analysis workflow; wherein, the analysis workflow is represented as a directed acyclic graph in memory.

[0063] This invention enables the structured definition of the analysis process by calling a workflow template and instantiating the algorithm according to its predefined combinatorial logic, and connecting them into a directed acyclic graph according to dependencies. This lays the foundation for subsequent intelligent scheduling.

[0064] Optionally, in this embodiment of the invention, the scheduling computing resources execute the analysis workflow and output the monitoring analysis results obtained after the analysis workflow processes the video data, including: The analysis workflow is loaded into the computing unit, and the directed acyclic graph is topologically sorted to obtain the execution sequence of each algorithm instance; The computing unit executes the monitoring and analysis algorithms corresponding to each algorithm instance in sequence according to the execution sequence; Obtain the processing result output by the final-level algorithm instance in the execution sequence, and output the processing result as the monitoring and analysis result.

[0065] This invention, through topological sorting of a directed acyclic graph to determine and sequentially execute node sequences, can parse and process dependencies between algorithms, ensuring that the analysis workflow is executed efficiently.

[0066] Optionally, in this embodiment of the invention, after the scheduled computing resources execute the analysis workflow and output the monitoring analysis results obtained by the analysis workflow after processing the video data, the method further includes: The context information is updated in real time. When the context information meets preset conditions, the analysis workflow is unloaded and computing resources are released.

[0067] This invention, by unloading the workflow and releasing resources when conditions are met based on updated context information, can avoid resource idleness and waste, and improve overall resource utilization.

[0068] like Figure 2 As shown, based on the above method implementation examples, corresponding system implementation examples are provided, including: a context building module, an algorithm orchestration engine, a rule base, and an algorithm library; The context building module can execute the action in step S101, which decomposes the video stream into multiple independent, computationally lightweight subtasks, processes them in parallel using multithreading or asynchronous event-driven methods, and finally aggregates the results into structured context information. The structure of the context building module is as follows: Figure 3 As shown, it includes: an input dispatcher, a temporal attribute analysis submodule, a spatial attribute analysis submodule, an object attribute analysis submodule, an environmental attribute analysis submodule, and a context aggregator. The processing steps of the context construction module include: the video stream is input to the input dispatcher, which simultaneously pushes each frame of video data (or keyframe data) to multiple parallel lightweight analysis submodules; each analysis submodule runs independently and continuously outputs its analysis results to a shared context data buffer; the context aggregator periodically reads the context data buffer, integrates all information, and generates the final multidimensional context information.

[0069] The rule base, which is equivalent to the rule base described in step S102, is used to store rules that represent the mapping relationship between "context-algorithm".

[0070] The algorithm library, equivalent to the algorithm library described in step S103, is used to store various independent heterogeneous monitoring and analysis algorithm modules that can be dynamically loaded and unloaded.

[0071] The algorithm orchestration engine can perform the actions of steps S102 to S104, which are used to receive real-time context information from the context building module, match the optimal rule from the rule base according to the context information, generate an analysis workflow according to the optimal rule, schedule system computing resources, load and execute the analysis workflow, and unload the current analysis workflow and release resources when the event ends or the context information changes.

[0072] The embodiments of the present invention adopt a modular architecture. By adding an algorithm or modifying a rule, the system can be given new capabilities, enabling rapid iteration and functional customization without changing the core architecture.

[0073] like Figure 4 As shown, based on the above method embodiments, another embodiment of the context-aware intelligent monitoring and analysis method is provided, including steps S1 to S5; Step S1: Continuously acquire multi-dimensional context information of the monitoring scene; wherein, executing step S1 is equivalent to executing step S101.

[0074] Step S2: Match the multidimensional context information with preset rules; wherein, executing step S2 is equivalent to executing step S102.

[0075] Step S3: Based on the matching results, dynamically generate an analysis workflow containing at least two heterogeneous algorithms; wherein, executing step S3 is equivalent to executing step S103.

[0076] Step S4: Schedule computing resources to execute the analysis workflow; wherein, executing step S4 is equivalent to executing step S104.

[0077] Step S5: After the event ends, unload the analysis workflow and release resources; wherein, executing step S5 is equivalent to the action of "updating the context information in real time after executing step S104, and unloading the analysis workflow and releasing computing resources when the context information meets the preset conditions".

[0078] The embodiments of the present invention can improve adaptability and analysis accuracy in different scenarios by making decisions based on contextual information; by loading the most suitable algorithm combination, it can avoid any form of computing power waste, making it possible to run complex intelligent analysis on low-power edge devices; through multi-dimensional context cross-validation, it can effectively filter false triggers caused by a single event (such as movement), significantly reduce invalid alarms, and improve overall reliability.

[0079] like Figure 5 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided; An embodiment of the present invention provides a context-aware intelligent monitoring and analysis device, including: a context information module 501, an algorithm scheduling rule module 502, an analysis workflow module 503, and a monitoring and analysis result module 504; The context information module 501 is used to acquire video data of the monitoring scene and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information. The algorithm scheduling rule module 502 is used to use the context information as a query condition to match in a preset rule base to obtain the corresponding algorithm scheduling rule. The analysis workflow module 503 is used to select multiple monitoring and analysis algorithms from a preset algorithm library according to the algorithm scheduling rules, and to instantiate each monitoring and analysis algorithm according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow. The monitoring and analysis result module 504 is used to schedule computing resources to execute the analysis workflow and output the monitoring and analysis results obtained after the analysis workflow processes the video data.

[0080] Optionally, in this embodiment of the invention, the context information module 501 includes: a data push submodule, an information extraction submodule, and an information aggregation submodule; The data push submodule is used to input video data to the distributor, so that the distributor pushes the video data to the time attribute analyzer, spatial attribute analyzer, object attribute analyzer and environmental attribute analyzer; The information extraction submodule is used to extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data through the time attribute analyzer, spatial attribute analyzer, object attribute analyzer and environmental attribute analyzer respectively; The information aggregation submodule is used to aggregate the time attribute information, spatial attribute information, object attribute information and environmental attribute information through a context aggregator to obtain context information.

[0081] The embodiments of the present invention employ a distributor to push information in parallel to each dedicated attribute analyzer, and an aggregator to aggregate the information, thereby achieving efficient extraction and unified construction of multi-dimensional contextual information and improving the processing efficiency of the perception process.

[0082] Optionally, in this embodiment of the invention, the algorithm scheduling rule module 502 includes: a condition matching submodule; The condition matching submodule is used to perform condition matching between the context information and each predefined rule in the rule base, and to determine the rule whose matching result meets the preset matching conditions as the algorithm scheduling rule.

[0083] This invention enables precise rule triggering by matching context information with preset rules and determining the rules that meet the conditions, thus ensuring that scheduling decisions are highly relevant to the current scenario.

[0084] Optionally, in this embodiment of the invention, the analysis workflow module 503 includes: an algorithm selection submodule; The algorithm selection submodule is used to select multiple corresponding monitoring and analysis algorithms from the algorithm library according to the algorithm identifier indicated in the algorithm scheduling rules.

[0085] This invention enables accurate access to algorithm resources by selecting algorithms based on the identifiers indicated in the algorithm scheduling rules, thus ensuring the precision of component calls during dynamic orchestration.

[0086] Optionally, in this embodiment of the invention, the analysis workflow module 503 further includes: a template calling submodule and a workflow generation submodule; The template invocation submodule is used to invoke the corresponding workflow template according to the algorithm scheduling rules; The workflow generation submodule is used to instantiate each of the monitoring and analysis algorithms according to the algorithm combination logic predefined in the workflow template and connect them according to the algorithm dependency relationship defined by the algorithm combination logic to obtain the analysis workflow; wherein, the analysis workflow is represented as a directed acyclic graph in memory.

[0087] This invention enables the structured definition of the analysis process by calling a workflow template and instantiating the algorithm according to its predefined combinatorial logic, and connecting them into a directed acyclic graph according to dependencies. This lays the foundation for subsequent intelligent scheduling.

[0088] Optionally, in this embodiment of the invention, the monitoring and analysis result module 504 includes: an execution sequence submodule, an algorithm execution submodule, and a result output submodule; The execution sequence submodule is used to load the analysis workflow into the computing unit and perform topological sorting on the directed acyclic graph to obtain the execution sequence of each algorithm instance; The algorithm execution submodule is used by the computing unit to sequentially execute the monitoring and analysis algorithms corresponding to each of the algorithm instances according to the execution sequence. The result output submodule is used to obtain the processing result output by the final-level algorithm instance in the execution sequence, and output the processing result as the monitoring and analysis result.

[0089] This invention, through topological sorting of a directed acyclic graph to determine and sequentially execute node sequences, can parse and process dependencies between algorithms, ensuring that the analysis workflow is executed efficiently.

[0090] Optionally, in this embodiment of the invention, a resource release submodule is further included after the monitoring and analysis result module 504; The resource release submodule is used to update the context information in real time. When the context information meets preset conditions, the analysis workflow is unloaded and the computing resources are released.

[0091] This invention, by unloading the workflow and releasing resources when conditions are met based on updated context information, can avoid resource idleness and waste, and improve overall resource utilization.

[0092] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement a context-aware intelligent monitoring and analysis method provided by any of the above-described method embodiments of the present invention.

[0093] In this embodiment of the invention, the context information module 501 acquires and extracts multi-dimensional attributes of time, space, object, and environment to construct context information, enabling comprehensive perception of the real-time status of the monitored scene and providing a basis for subsequent intelligent decision-making. The algorithm scheduling rule module 502 matches the context information with a preset rule base to obtain algorithm scheduling rules, which can transform scene understanding into analysis instructions. The analysis workflow module 503 selects algorithms according to the scheduling rules and instantiates them according to the corresponding workflow template to generate analysis workflows, which can dynamically assemble analysis workflows according to scene requirements and achieve accurate matching and combination of algorithm resources. The monitoring and analysis result module 504 schedules resources to execute the analysis workflow and outputs processing results, which can ensure that the analysis workflow is executed efficiently and ultimately produce monitoring and analysis results.

[0094] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0095] Based on the above embodiment of a context-aware intelligent monitoring and analysis method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a context-aware intelligent monitoring and analysis method according to any embodiment of the present invention.

[0096] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0097] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0098] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0099] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute a context-aware intelligent monitoring and analysis method as described in any of the above-described method embodiments of the present invention.

[0100] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0101] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A context-aware based intelligent surveillance analysis method, characterized in that, include: Acquire video data from the monitored scene, and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information; The context information is used as a query condition to match the results in a pre-set rule base, thereby obtaining the corresponding algorithm scheduling rule. According to the algorithm scheduling rules, multiple monitoring and analysis algorithms are selected from the preset algorithm library, and each monitoring and analysis algorithm is instantiated according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow. The system schedules computing resources to execute the analysis workflow and outputs the monitoring and analysis results obtained after the analysis workflow processes the video data.

2. A context-aware based intelligent monitoring analysis method as claimed in claim 1, wherein, The process involves acquiring video data from the monitored scene and extracting its temporal, spatial, object, and environmental attributes in parallel to construct contextual information, including: Video data is input to a distributor, which then pushes the video data to a temporal attribute analyzer, a spatial attribute analyzer, an object attribute analyzer, and an environmental attribute analyzer. The time attribute information, spatial attribute information, object attribute information, and environmental attribute information of the video data are extracted by the time attribute analyzer, spatial attribute analyzer, object attribute analyzer, and environmental attribute analyzer, respectively. The context aggregator aggregates the time attribute information, spatial attribute information, object attribute information, and environmental attribute information to obtain context information.

3. The context-aware intelligent monitoring and analysis method as described in claim 1, characterized in that, The step of using the context information as a query condition to match within a pre-set rule base to obtain the corresponding algorithm scheduling rule includes: The context information is matched against each predefined rule in the rule base, and the rule whose matching result meets the preset matching conditions is determined as the algorithm scheduling rule.

4. The context-aware intelligent monitoring and analysis method as described in claim 1, characterized in that, The step of selecting multiple monitoring and analysis algorithms from a pre-set algorithm library according to the algorithm scheduling rules includes: Based on the algorithm identifier indicated in the algorithm scheduling rules, select multiple corresponding monitoring and analysis algorithms from the algorithm library.

5. The context-aware intelligent monitoring and analysis method as described in claim 4, characterized in that, The step of instantiating each of the monitoring and analysis algorithms according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow includes: According to the algorithm scheduling rules, the corresponding workflow template is invoked; Based on the predefined algorithm combination logic in the workflow template, each monitoring and analysis algorithm is instantiated and connected according to the algorithm dependency relationship defined by the algorithm combination logic to obtain the analysis workflow; wherein, the analysis workflow is represented as a directed acyclic graph in memory.

6. The context-aware intelligent monitoring and analysis method as described in claim 5, characterized in that, The scheduled computing resources execute the analysis workflow and output the monitoring analysis results obtained after processing the video data by the analysis workflow, including: The analysis workflow is loaded into the computing unit, and the directed acyclic graph is topologically sorted to obtain the execution sequence of each algorithm instance; The computing unit executes the monitoring and analysis algorithms corresponding to each algorithm instance in sequence according to the execution sequence; Obtain the processing result output by the final-level algorithm instance in the execution sequence, and output the processing result as the monitoring and analysis result.

7. The context-aware intelligent monitoring and analysis method as described in claim 6, characterized in that, After the scheduled computing resources execute the analysis workflow and output the monitoring analysis results obtained by the analysis workflow after processing the video data, the method further includes: The context information is updated in real time, and when the context information meets preset conditions, the analysis workflow is unloaded and computing resources are released.

8. A context-aware intelligent monitoring and analysis device, characterized in that, include: Context information module, algorithm scheduling rules module, analysis workflow module, and monitoring and analysis results module; The context information module is used to acquire video data of the monitoring scene and extract the time attribute information, spatial attribute information, object attribute information and environmental attribute information of the video data in parallel to construct context information; The algorithm scheduling rule module is used to use the context information as a query condition to match in a preset rule base to obtain the corresponding algorithm scheduling rule; The analysis workflow module is used to select multiple monitoring and analysis algorithms from a preset algorithm library according to the algorithm scheduling rules, and instantiate each monitoring and analysis algorithm according to the workflow template corresponding to the algorithm scheduling rules to generate an analysis workflow. The monitoring and analysis results module is used to schedule computing resources to execute the analysis workflow and output the monitoring and analysis results obtained after the analysis workflow processes the video data.

9. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a context-aware intelligent monitoring and analysis method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform a context-aware intelligent monitoring and analysis method as described in any one of claims 1-7.

Images