Algorithm task scheduling method, device, electronic device and storage medium

Abstract

The embodiments of the present invention provide an algorithm task scheduling method, device, electronic device and storage medium. The algorithm task scheduling method includes obtaining first association information of each device point; for each image information, performing parsing processing through a preset algorithm set to obtain corresponding parsing content; for each image information, forming an algorithm label with the parsing content corresponding to the image information and the corresponding preset algorithm and adding it to the corresponding first association information through a point identifier to form second association information; based on the second association information, constructing a relationship model of the point label and / or condition label with respect to the algorithm label, and executing the scheduling of the algorithm task through the relationship model. The present invention improves the scheduling efficiency of the algorithm task.

Description

Technical Field

[0001] The present invention relates to the field of task management, and in particular to an algorithmic task scheduling method, device, electronic equipment and storage medium. Background Art

[0002] With the development of smart cities, AI machine vision technology has been used to efficiently extract effective information from video data in various scenarios, and to use this information to serve the needs of urban management such as urban security, social governance, and safe production. This has become an important development direction for AI vision technology to be applied to urban innovation. AI vision technology can be used to monitor various urban safety hazards and violations, operate around the clock and uninterruptedly, and effectively solve the problems of "difficult to discover, difficult to obtain evidence, and difficult to prevent" in urban governance. The application of AI vision technology in the field of urban management requires the following necessary technical links: 1. A series of algorithms that meet the needs of urban scenarios; 2. An algorithm warehouse platform that carries the operation of related algorithms, through which videos can be analyzed to extract effective information; 3. A system for scheduling algorithm tasks to match algorithm tasks with scenario requirements. Among them, the relevant algorithms and algorithm warehouse platform technical solutions and related technical architectures for the current urban scene requirements have become increasingly mature, but in the field of algorithm task scheduling systems, the manual orchestration method is currently mainly used. In this method, the orchestration of algorithm task scheduling mainly relies on personal experience, and the algorithm and the front-end points need to be coupled in advance; when the scale of front-end device access is large, the manual orchestration workload is extremely large and the efficiency is low; because it is manually orchestrated in advance, it is difficult to meet the timeliness requirements in emergencies. Therefore, the existing algorithm task scheduling method has the problem of low scheduling efficiency under urban scene requirements. Summary of the invention

[0003] The embodiment of the present invention provides a task scheduling method, which does not require manual arrangement of algorithm tasks and improves the scheduling efficiency of algorithm tasks.

[0004] In a first aspect, an embodiment of the present invention provides an algorithm task scheduling method, the method comprising:

[0005] Acquire first association information of each device point, the first association information includes image information, point labels and condition labels, the image information, the point labels and the condition labels are associated through point identifiers, and each device point corresponds to a point identifier;

[0006] For each of the image information, a preset algorithm set is used to perform analysis processing to obtain corresponding analysis content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs an analysis content for the image information;

[0007] For each piece of image information, the analysis content corresponding to the image information and the corresponding preset algorithm are used to form an algorithm tag, and the tag is added to the corresponding first associated information through the point identifier to form second associated information, where each algorithm tag corresponds to one preset algorithm;

[0008] Based on the second association information, a relationship model between the point label and / or the condition label and the algorithm label is constructed, and the scheduling of the algorithm task is performed through the relationship model.

[0009] Optionally, the obtaining of first associated information of each device point includes:

[0010] Acquire the point information corresponding to each of the equipment points, and form a corresponding point label according to the point information;

[0011] Acquire task condition information corresponding to each equipment point, and form corresponding condition labels according to the task condition information;

[0012] Obtaining image information corresponding to each device point;

[0013] The image information, the point label and the condition label are associated with each other through the point identification to obtain the first associated information.

[0014] Optionally, the acquiring task condition information corresponding to each equipment point and forming a corresponding condition label according to the task condition information includes:

[0015] Obtain information from government websites or the Internet;

[0016] The task condition information corresponding to each equipment point is extracted from the government network information or the Internet information, and a corresponding condition label is formed according to the task condition information.

[0017] Optionally, constructing a relationship model between the point label and / or the condition label and the algorithm label based on the second association information includes:

[0018] Establishing a knowledge graph based on the second association information by performing data mining on the point label, the condition label needle and the algorithm label in the second association information;

[0019] Through the knowledge graph, a relationship model between the point label and / or the condition label and the algorithm label is formed.

[0020] Optionally, the scheduling of executing the algorithm task through the relational model includes:

[0021] When a trigger condition of an algorithm task is received, the relationship model is retrieved;

[0022] According to the task triggering condition, converting the relational model into an executable scheduling instruction;

[0023] The scheduling of the algorithm tasks is performed by the scheduling instructions.

[0024] Optionally, after the scheduling of executing the algorithm task through the relational model, the method further includes:

[0025] Evaluate the execution of the algorithm task by using a preset evaluation method;

[0026] When the evaluation of the execution status does not satisfy a preset evaluation condition, the relationship model is updated and iterated, and the scheduling of subsequent algorithm tasks is performed through the updated and iterated relationship model.

[0027] Optionally, the iterative updating of the relationship model includes:

[0028] Updating the second association information of the current iteration cycle by using the first association information of the previous iteration cycle;

[0029] Updating the relationship model of the current iteration period by using the second association information of the current iteration period;

[0030] Scheduling of algorithm tasks is performed through the relational model of the current iteration cycle;

[0031] If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle does not meet the preset evaluation conditions, then enter the next iteration cycle;

[0032] If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle meets the preset evaluation condition, the iteration is stopped and the relational model of the current iteration cycle is stored.

[0033] In a second aspect, an embodiment of the present invention provides an algorithm task scheduling device, the device comprising:

[0034] An acquisition module, used to acquire first associated information of each equipment point, wherein the first associated information includes image information, a point label and a condition label, wherein the image information, the point label and the condition label are associated with each other through a point identifier, and each equipment point corresponds to a point identifier;

[0035] A parsing module, used for parsing each of the image information by using a preset algorithm set to obtain corresponding parsing content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs one parsing content for one image information;

[0036] An association module, configured to form an algorithm tag by combining the analysis content and the corresponding preset algorithm corresponding to each image information and adding the algorithm tag to the corresponding first association information through the point identifier to form second association information, wherein each algorithm tag corresponds to one preset algorithm;

[0037] A processing module is used to construct a relationship model between the point label and / or the condition label and the algorithm label based on the second association information, and execute the scheduling of the algorithm task through the relationship model.

[0038] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the algorithm task scheduling method provided in the embodiment of the present invention when executing the computer program.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the algorithmic task scheduling method provided in the embodiment of the invention are implemented.

[0040] In an embodiment of the present invention, first association information of each device point is obtained, the first association information includes image information, point labels and conditional labels, the image information, point labels and conditional labels are associated through point identifiers, and each device point corresponds to a point identifier; for each image information, parsing is performed through a preset algorithm set to obtain corresponding parsing content, the preset algorithm set includes multiple preset algorithms, and each preset algorithm outputs a parsing content for the image information; for each image information, the parsing content corresponding to the image information and the corresponding preset algorithm are formed into an algorithm label and added to the corresponding first association information through the point identifier to form second association information, and each algorithm label corresponds to a preset algorithm; based on the second association information, a relationship model of the point label and / or conditional label with respect to the algorithm label is constructed, and scheduling of the algorithm task is performed through the relationship model. The present invention associates the point label, condition label and image information of each equipment point through point identification to obtain the first association information of each equipment point, parses the first association information of each equipment point through a preset algorithm set, adds the parsed content and the preset algorithm used for parsing to the first association information to form second association information, and then forms a corresponding relationship model for the second association information. Since the preset algorithm is added to the second association information, the relationship model formed by the second association information records the relationship between each equipment point and each preset algorithm under different point labels and condition labels, so that the scheduling of algorithm tasks can be executed through the relationship model, without the need for manual arrangement of algorithm tasks, thereby improving the scheduling efficiency of algorithm tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a flow chart of an algorithm task scheduling method provided by an embodiment of the present invention;

[0043] Figure 2 It is a structural diagram of an algorithm task scheduling device provided by an embodiment of the present invention;

[0044] Figure 3 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] See also Figure 1 , Figure 1 is a flowchart of a task scheduling method provided by an embodiment of the present invention, such as Figure 1 As shown, the task scheduling method includes the following steps:

[0047] 101. Obtain first associated information of each device point.

[0048] In the embodiment of the present invention, the first association information includes image information, point labels and condition labels. The image information, point labels and condition labels are associated with each other through point identifiers, and each device point corresponds to a point identifier.

[0049] Furthermore, the device point refers to the device installation location, and each device point corresponds to an installed device. The above-mentioned device is an imaging device, and the above-mentioned imaging device is used to capture video or capture images, so as to realize monitoring of the target area at the device point.

[0050] The above-mentioned point label can be understood as the point information describing the device point, such as point scene, point location, point longitude and latitude, point industry and other attribute information. The point scene, for example, the point is in the scene corresponding to the algorithm task such as the traffic monitoring scene and the pedestrian monitoring scene. The point location can be a traffic intersection, supermarket, customs and other places where image equipment can be installed. The point longitude and latitude can be the longitude and latitude of the installation location. The point industry can be determined according to the place where the image equipment is installed. For example, if the place where the image equipment is installed is a bank, the corresponding point industry is the financial industry. If the place where the image equipment is installed is a construction site, the corresponding point industry is the construction industry.

[0051] The above-mentioned conditional labels can be understood as the task condition information triggered by the algorithm task. The above-mentioned task condition information is used to determine under what conditions the image device performs what algorithm task, such as what algorithm task to perform under weather conditions, what algorithm task to perform under holiday conditions, what algorithm task to perform under preset time period conditions, what algorithm task to perform under emergency conditions, what algorithm task to perform under security conditions, what algorithm task to perform under seasonal conditions, etc.

[0052] The above-mentioned image information may be picture data or video data collected by an imaging device. Furthermore, the above-mentioned image information may include a point identification of the imaging device, a shooting time, a storage address or a storage link of the picture data and the video data.

[0053] In an embodiment of the present invention, the above-mentioned point labels and image information can be associated based on the point identification, and combined and associated according to the conditional label items, such as weather conditions, holiday conditions, preset time period conditions, emergency conditions, security conditions, seasonal conditions and other conditional items, to obtain the first association information of each image device, and the above-mentioned first association information includes the point identification, video data and the storage address or storage link of the video data, the point label and the conditional label.

[0054] Furthermore, the algorithm task has a time period attribute, that is, the algorithm task includes a corresponding task execution time, that is, the task execution time of each device point corresponding to the algorithm task under different condition labels, such as the task execution time of weather conditions, the task execution time of holidays, the task execution time of preset time period conditions, the task execution time of emergency conditions, the task execution time of security conditions, the task execution time of seasonal conditions, etc. The task execution time corresponding to the algorithm task is related to the duration of the task condition. For example, the longer the duration of a certain weather, the longer the task execution time of the corresponding algorithm task under the weather condition. The first associated information includes the point identification, the video data and the storage address or storage link of the video data, the task execution time corresponding to the algorithm task (also called the task time period), the point label and the condition label.

[0055] Optionally, the point information corresponding to each of the above-mentioned equipment points can be obtained, and corresponding point tags can be formed based on the above-mentioned point information; the task condition information corresponding to each of the above-mentioned equipment points can be obtained, and corresponding condition tags can be formed based on the above-mentioned task condition information; the image information corresponding to each of the above-mentioned equipment points can be obtained; and the above-mentioned image information, the above-mentioned point tags and the above-mentioned condition tags can be associated through the above-mentioned point identification to obtain the above-mentioned first associated information.

[0056] When the imaging device is installed, information can be collected for the installed equipment points, so as to obtain the equipment point identification, point scene, point location, point longitude and latitude, point industry and other point information. Among them, the above-mentioned point identification can be a point ID, and the point ID can be a device ID, or an ID generated according to the device ID. Furthermore, the acquired point information can be data-managed to form a point basic information record, and a corresponding point tag can be formed according to the point basic information record. The above-mentioned point basic information record includes the equipment point identification, point scene, point location, point longitude and latitude, point industry and other information. Specifically, the point scene, point location, point longitude and latitude, and point industry corresponding to the point identification can be formed into a point tag.

[0057] Government network information or Internet information can be obtained; task condition information corresponding to each equipment point is extracted from the above government network information or Internet information, and corresponding condition labels are formed according to the above task condition information.

[0058] Specifically, the network information corresponding to a certain area can be obtained from the government network information or Internet information, and the weather, holidays, time periods, seasons, emergency incidents, security incidents and other information can be extracted from the above network information. The certain area will correspond to one or more equipment points, so the weather, holidays, time periods, seasons, emergency incidents, security incidents and other information corresponding to each equipment point can be extracted. Data governance is performed based on the weather, holidays, time periods, seasons, emergency incidents, security incidents and other information to obtain task condition information such as weather conditions, holiday conditions, preset time period conditions, emergency conditions, security conditions, and seasonal conditions, and corresponding condition labels are formed based on the task condition information.

[0059] In an embodiment of the present invention, by forming point information into corresponding point tags and task condition information into corresponding condition tags, the spatiotemporal attributes of equipment points can be intuitively labeled, so that algorithm tasks can be scheduled in the spatiotemporal dimensions based on point tags and condition tags, thereby more flexibly and accurately scheduling algorithm tasks under urban scenario requirements, thereby improving the scheduling efficiency of algorithm tasks.

[0060] 102. For each piece of image information, a preset algorithm set is used to perform analysis processing to obtain corresponding analysis content.

[0061] In an embodiment of the present invention, the above-mentioned preset algorithm set includes multiple preset algorithms, and each of the above-mentioned preset algorithms outputs a parsing content for one image information. Specifically, the above-mentioned algorithm set may include all preset algorithms under urban scene requirements, such as vehicle detection algorithm, face detection algorithm, rainy day vehicle detection algorithm, rainy day puddle detection algorithm, epidemic body temperature detection algorithm, epidemic mask detection algorithm, Monday morning rush hour traffic violation detection algorithm, spring urban greening plant detection algorithm, fishing recognition algorithm, water accumulation recognition algorithm, fireworks recognition algorithm and helmet recognition algorithm, etc. It should be noted that the above-mentioned algorithms are examples and should not be regarded as limitations of the present invention. All the above-mentioned preset algorithms should include all algorithms corresponding to the combination of different point labels and different conditional labels.

[0062] It should be noted that all the preset algorithms in the above algorithm set are pre-trained, and the pre-trained preset algorithms can directly perform inference results on image data and video data.

[0063] In a possible embodiment, each piece of image information may be attributed and classified according to task execution time, point label, and condition label, which is equivalent to adding attributes such as task execution time, point label, and condition label to each piece of image information.

[0064] Furthermore, the above algorithm set can be used to parse part or all of the image information. For each image information, the above algorithm set is only parsed once to avoid repeated parsing. Each image information will output a parsed content through a preset algorithm. Furthermore, for one image information, an algorithm set containing N preset algorithms will output N parsed contents.

[0065] The above analysis content may include confidence, recall rate, accuracy rate and the like.

[0066] 103. For each piece of image information, the analysis content corresponding to the image information and the corresponding preset algorithm are formed into an algorithm label, and are added to the corresponding first associated information through a point identifier to form second associated information.

[0067] In an embodiment of the present invention, each algorithm tag corresponds to a preset algorithm, indicating the preset algorithm used when parsing image information. The above algorithm tag can be the name of the corresponding preset algorithm or the algorithm ID, etc., which can form a unique identifier of the individual tag.

[0068] Since the above image information includes point identification, the obtained parsing content and the corresponding preset algorithm can also be added to the first associated information through the point identification to obtain the second associated information. Specifically, the second associated information includes the point identification, the task execution time corresponding to the algorithm task, the point label, the condition label, the algorithm label, and the parsing content.

[0069] In the embodiment of the present invention, the acquisition of the second association information may be started periodically or manually, which is equivalent to starting the construction of the relationship model periodically or manually, and then starting the scheduling of the algorithm task periodically or manually.

[0070] 104. Based on the second association information, a relationship model between the point label and / or the condition label and the algorithm label is constructed, and the algorithm task is scheduled through the relationship model.

[0071] In an embodiment of the present invention, the second association information includes point identification, task execution time corresponding to the algorithm task, point label, condition label, algorithm label, and parsing content. A relationship model with labels as nodes and parsing content as node connection relationships can be constructed through the above-mentioned point labels, condition labels, algorithm labels, and parsing content. In the above-mentioned relationship model, point labels, condition labels, and algorithm labels are nodes, and parsing content is a connection relationship. For example, the higher the confidence, the higher the recall rate, and the higher the accuracy rate in the parsing content, the stronger the connection relationship. Conversely, the lower the confidence, the lower the recall rate, and the lower the accuracy rate in the parsing content, the weaker the connection relationship.

[0072] Optionally, a knowledge graph based on the second association information may be established by performing data mining on the point labels, condition labels and algorithm labels in the second association information; and a relationship model of the point labels and / or condition labels with respect to the algorithm labels may be formed through the knowledge graph.

[0073] The above-mentioned data mining can be understood as big data mining and analysis, which mines and analyzes the relationship between the above-mentioned point labels and the above-mentioned conditional labels and the above-mentioned algorithm labels. It should be noted that in the above-mentioned knowledge graph, the point labels, conditional labels and algorithm labels are used as nodes, and the parsed content is used as an edge connecting the two nodes. The edge has a relationship value, and the relationship value is related to the parsed content. In a possible embodiment, of the two nodes of an edge, one is a node corresponding to the algorithm label, and the other is a point label or a conditional label. A node can be connected to multiple corresponding nodes through multiple edges.

[0074] In a possible embodiment, through the second association information, big data mining and analysis technology can be used to establish a knowledge graph based on point labels, condition labels and algorithm labels, form a relationship model based on the association of various types of labels with algorithm labels in different time periods, and store the relationship model in a relationship model library.

[0075] In an embodiment of the present invention, when a trigger condition of an algorithm task is received, the above relationship model can be called; according to the above task trigger condition, the above relationship model can be converted into an executable scheduling instruction; and the scheduling of the above algorithm task can be executed through the above scheduling instruction.

[0076] Specifically, the trigger condition can be a trigger condition formed by manual call or external condition triggering. When the trigger condition is received, the corresponding relationship model is retrieved from the relationship model library according to the trigger condition, and the relationship model is converted into an executable scheduling instruction, and the scheduling of the corresponding algorithm task is executed through the scheduling instruction. The above scheduling instructions include algorithm tasks, and the above algorithm tasks are tasks executed based on algorithms. For example, the scheduling instructions for vehicle detection tasks may include vehicle detection algorithm parameters, point identification for executing algorithm tasks, and paths for returning algorithm task results.

[0077] Taking rain as an example, when rain is detected at point A, the trigger condition rain is formed. After receiving the trigger condition of rain at point A, the relationship model corresponding to rain at point A is retrieved from the relationship model library, and the relationship model is converted into the corresponding scheduling instruction. The algorithm corresponding to rain (such as puddle detection algorithm, image rain removal algorithm, rainy vehicle detection algorithm, etc.) is formed into an algorithm task and scheduled to the image device corresponding to point A, and the algorithm task is executed by the image device.

[0078] Optionally, after step 104, the execution status of the above-mentioned algorithm task can be evaluated regularly or manually. Specifically, the execution status of the above-mentioned algorithm task can be evaluated by a preset evaluation method; when the evaluation of the above-mentioned execution status does not meet the preset evaluation conditions, the above-mentioned relationship model is updated and iterated, and the scheduling of subsequent algorithm tasks is performed through the above-mentioned updated and iterated relationship model.

[0079] In the embodiment of the present invention, the effect of algorithm task scheduling can be evaluated by the execution status of the algorithm task. If the execution status of the algorithm task is good, the scheduling effect of the algorithm task is good; if the execution status of the algorithm task is poor, the scheduling effect of the algorithm task is poor.

[0080] The above-mentioned execution status may be a comparison between the algorithm task result and the parsed content. The above-mentioned algorithm task result may include confidence, recall rate, and accuracy rate. The confidence, recall rate, and accuracy rate in the algorithm task result and the confidence, recall rate, and accuracy rate in the parsed content may be subjected to error calculation to obtain an error result. The above-mentioned preset evaluation condition may be that the error result of the confidence, recall rate, and accuracy rate is within a preset range, such as within a range of ±10%. When the error result is within a range of ±10%, it can be considered that the execution of the algorithm task is good, and thus the scheduling effect of the algorithm task is good, and the relationship model does not need to be updated. When the error result exceeds the range of ±10%, it can be considered that the execution of the algorithm task is poor, and thus the scheduling effect of the algorithm task is poor, and the relationship model needs to be updated and iterated.

[0081] In a possible embodiment, when the above-mentioned algorithm task is an algorithm task based on a target detection algorithm, the above-mentioned execution status may also be a situation where the number of detected targets changes. When the difference between the number of detected targets and the actual number of targets reaches a preset threshold, it can be considered that the evaluation of the execution status does not meet the preset evaluation conditions.

[0082] Optionally, when the relationship model needs to be updated and iterated, the second association information of the current iteration cycle can be updated through the first association information of the previous iteration cycle; the relationship model of the current iteration cycle can be updated through the second association information of the current iteration cycle; the algorithm task scheduling can be executed through the relationship model of the current iteration cycle; if the execution status of the algorithm task corresponding to the relationship model of the current iteration cycle does not meet the preset evaluation conditions, enter the next iteration cycle; if the execution status of the algorithm task corresponding to the relationship model of the current iteration cycle meets the preset evaluation conditions, stop the iteration and store the relationship model of the current iteration cycle.

[0083] It should be noted that when the relationship model needs to be updated and iterated, it can be iterated according to the above steps 101 to 104 to obtain a relationship model that meets the conditions, thereby improving the scheduling effect of the algorithm task.

[0084] In an embodiment of the present invention, first association information of each device point is obtained, the first association information includes image information, point labels and conditional labels, the image information, point labels and conditional labels are associated through point identifiers, and each device point corresponds to a point identifier; for each image information, parsing is performed through a preset algorithm set to obtain corresponding parsing content, the preset algorithm set includes multiple preset algorithms, and each preset algorithm outputs a parsing content for one image information; for each image information, the parsing content corresponding to the image information and the corresponding preset algorithm are formed into an algorithm label, and are added to the corresponding first association information through the point identifier to form second association information, and each algorithm label corresponds to a preset algorithm; based on the second association information, a relationship model of the point label and / or conditional label with respect to the algorithm label is constructed, and scheduling of the algorithm task is performed through the relationship model. The present invention associates the point label, condition label and image information of each equipment point through point identification to obtain the first association information of each equipment point, parses the first association information of each equipment point through a preset algorithm set, adds the parsed content and the preset algorithm used for parsing to the first association information to form second association information, and then forms a corresponding relationship model for the second association information. Since the preset algorithm is added to the second association information, the relationship model formed by the second association information records the relationship between each equipment point and each preset algorithm under different point labels and condition labels, so that the scheduling of algorithm tasks can be executed through the relationship model, without the need for manual arrangement of algorithm tasks, thereby improving the scheduling efficiency of algorithm tasks.

[0085] It should be noted that the algorithm task scheduling method provided in the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. that can perform algorithm task scheduling.

[0086] Optional, see Figure 2 , Figure 2 is a schematic diagram of the structure of an algorithm task scheduling device provided by an embodiment of the present invention, such as Figure 2 As shown, the device comprises:

[0087] The acquisition module 201 is used to acquire first associated information of each equipment point, wherein the first associated information includes image information, a point label and a condition label, wherein the image information, the point label and the condition label are associated with each other through a point identifier, and each equipment point corresponds to a point identifier;

[0088] The parsing module 202 is used to perform parsing processing on each of the image information by using a preset algorithm set to obtain corresponding parsing content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs a parsing content for one image information;

[0089] The association module 203 is used to form an algorithm tag with the analysis content corresponding to the image information and the corresponding preset algorithm for each of the image information, and add the point identifier to the corresponding first association information to form second association information, wherein each algorithm tag corresponds to one preset algorithm;

[0090] The processing module 204 is used to construct a relationship model between the point label and / or the condition label and the algorithm label based on the second association information, and execute the scheduling of the algorithm task through the relationship model.

[0091] Optionally, the acquisition module 201 includes:

[0092] A first acquisition submodule is used to acquire the point information corresponding to each device point and form a corresponding point label according to the point information;

[0093] A second acquisition submodule is used to acquire task condition information corresponding to each equipment point, and form a corresponding condition label according to the task condition information;

[0094] The third acquisition submodule is used to acquire the image information corresponding to each device point;

[0095] The association submodule is used to associate the image information, the point label and the condition label through the point identification to obtain the first association information.

[0096] Optionally, the second acquisition submodule includes:

[0097] An acquisition unit, used to acquire government network information or Internet information;

[0098] The extraction unit is used to extract the task condition information corresponding to each equipment point from the government network information or the Internet information, and form a corresponding condition label according to the task condition information.

[0099] Optionally, the processing module 204 includes:

[0100] A first establishing submodule is used to establish a knowledge graph based on the second association information by performing data mining on the point label, the condition label needle and the algorithm label in the second association information;

[0101] The second establishment submodule is used to form a relationship model between the point label and / or the condition label and the algorithm label through the knowledge graph.

[0102] Optionally, the processing module 204 further includes:

[0103] A calling submodule, used for calling the relationship model when receiving the triggering condition of the algorithm task;

[0104] A conversion submodule, used for converting the relational model into executable scheduling instructions according to the task triggering condition;

[0105] The first scheduling submodule is used to execute the scheduling of the algorithm task through the scheduling instruction.

[0106] Optionally, the device further comprises:

[0107] An evaluation module, used to evaluate the execution of the algorithm task by a preset evaluation method;

[0108] The updating module is used to update and iterate the relationship model when the evaluation of the execution status does not meet the preset evaluation conditions, and to schedule subsequent algorithm tasks through the updated and iterated relationship model.

[0109] Optionally, the update module includes:

[0110] A first updating submodule, used to update the second association information of the current iteration cycle by using the first association information of the previous iteration cycle;

[0111] A second updating submodule, used to update the relationship model of the current iteration period according to the second association information of the current iteration period;

[0112] A second scheduling submodule, used for scheduling algorithm tasks executed through the relational model of the current iteration cycle;

[0113] The iteration submodule is used to enter the next iteration cycle if the execution status of the algorithm task corresponding to the relational model of the current iteration cycle does not meet the preset evaluation conditions;

[0114] The stop submodule is used to stop the iteration if the execution status of the algorithm task corresponding to the relationship model of the current iteration cycle meets the preset evaluation condition, and store the relationship model of the current iteration cycle.

[0115] It should be noted that the algorithm task scheduling device provided in the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. that can perform algorithm task scheduling.

[0116] The algorithm task scheduling device provided in the embodiment of the present invention can implement each process implemented by the task scheduling method in the above method embodiment and can achieve the same beneficial effect. To avoid repetition, it will not be described here.

[0117] See also Figure 3 , Figure 3is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, it includes: a memory 302, a processor 301, and a computer program of a task scheduling method stored in the memory 302 and executable on the processor 301, wherein:

[0118] The processor 301 is used to call the computer program stored in the memory 302 and execute the following steps:

[0119] Acquire first association information of each device point, the first association information includes image information, point labels and condition labels, the image information, the point labels and the condition labels are associated through point identifiers, and each device point corresponds to a point identifier;

[0120] For each of the image information, a preset algorithm set is used to perform analysis processing to obtain corresponding analysis content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs one analysis content for one image information;

[0121] For each piece of image information, the analysis content corresponding to each piece of image information and the corresponding preset algorithm are used to form an algorithm label, and the label is added to the corresponding first associated information through the point identifier to form second associated information, and each algorithm label corresponds to one preset algorithm;

[0122] Based on the second association information, a relationship model between the point label and / or the condition label and the algorithm label is constructed, and the scheduling of the algorithm task is performed through the relationship model.

[0123] Optionally, the acquiring of first associated information of each device point performed by the processor 301 includes:

[0124] Acquire the point information corresponding to each of the equipment points, and form a corresponding point label according to the point information;

[0125] Acquire task condition information corresponding to each equipment point, and form corresponding condition labels according to the task condition information;

[0126] Obtaining image information corresponding to each device point;

[0127] The image information, the point label and the condition label are associated with each other through the point identification to obtain the first associated information.

[0128] Optionally, the processor 301 executes the acquiring of the task condition information corresponding to each equipment point and forming a corresponding condition label according to the task condition information, including:

[0129] Obtain information from government websites or the Internet;

[0130] The task condition information corresponding to each equipment point is extracted from the government network information or the Internet information, and a corresponding condition label is formed according to the task condition information.

[0131] Optionally, the step of constructing a relationship model between the point label and / or the condition label and the algorithm label based on the second association information, performed by the processor 301, includes:

[0132] Establishing a knowledge graph based on the second association information by performing data mining on the point label, the condition label needle and the algorithm label in the second association information;

[0133] Through the knowledge graph, a relationship model between the point label and / or the condition label and the algorithm label is formed.

[0134] Optionally, the scheduling of executing the algorithm task by using the relational model performed by the processor 301 includes:

[0135] When a trigger condition of an algorithm task is received, the relationship model is retrieved;

[0136] According to the task triggering condition, converting the relational model into an executable scheduling instruction;

[0137] The scheduling of the algorithm tasks is performed by the scheduling instructions.

[0138] Optionally, after the scheduling of executing the algorithm task through the relational model, the method executed by the processor 301 further includes:

[0139] Evaluate the execution of the algorithm task by using a preset evaluation method;

[0140] When the evaluation of the execution status does not satisfy a preset evaluation condition, the relationship model is updated and iterated, and the scheduling of subsequent algorithm tasks is performed through the updated and iterated relationship model.

[0141] Optionally, the updating and iterating of the relationship model performed by the processor 301 includes:

[0142] Updating the second association information of the current iteration cycle by using the first association information of the previous iteration cycle;

[0143] Updating the relationship model of the current iteration period by using the second association information of the current iteration period;

[0144] Scheduling of algorithm tasks is performed through the relational model of the current iteration cycle;

[0145] If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle does not meet the preset evaluation conditions, then enter the next iteration cycle;

[0146] If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle meets the preset evaluation condition, the iteration is stopped and the relational model of the current iteration cycle is stored.

[0147] It should be noted that the electronic device provided by the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. that can perform algorithm task scheduling.

[0148] The electronic device provided by the embodiment of the present invention can implement each process implemented by the algorithm task scheduling method in the above method embodiment, and can achieve the same beneficial effect. To avoid repetition, it will not be described here.

[0149] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the algorithm task scheduling method or the application-side algorithm task scheduling method provided in the embodiment of the present invention are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0150] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0151] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. An algorithmic task scheduling method, characterized in that: The following steps are involved: Acquire first association information of each device point, the first association information includes image information, point labels and condition labels, the image information, the point labels and the condition labels are associated through point identifiers, and each device point corresponds to a point identifier; For each of the image information, a preset algorithm set is used to perform analysis processing to obtain corresponding analysis content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs an analysis content for the image information; For each piece of image information, the analysis content corresponding to the image information and the corresponding preset algorithm are used to form an algorithm tag, and the tag is added to the corresponding first associated information through the point identifier to form second associated information, where each algorithm tag corresponds to one preset algorithm; Based on the second association information, a relationship model between the point label and / or the condition label and the algorithm label is constructed, and the scheduling of the algorithm task is performed through the relationship model.

2. The method according to claim 1, characterized in that The obtaining of first associated information of each device point includes: Acquire the point information corresponding to each of the equipment points, and form a corresponding point label according to the point information; Acquire task condition information corresponding to each equipment point, and form corresponding condition labels according to the task condition information; Obtaining image information corresponding to each device point; The image information, the point label and the condition label are associated with each other through the point identification to obtain the first associated information.

3. The method according to claim 2, characterized in that The obtaining of the task condition information corresponding to each equipment point and forming a corresponding condition label according to the task condition information includes: Obtain information from government websites or the Internet; The task condition information corresponding to each equipment point is extracted from the government network information or the Internet information, and a corresponding condition label is formed according to the task condition information.

4. The method according to claim 1, characterized in that The step of constructing a relationship model between the point label and / or the condition label and the algorithm label based on the second association information includes: Establishing a knowledge graph based on the second association information by performing data mining on the point label, the condition label needle and the algorithm label in the second association information; Through the knowledge graph, a relationship model between the point label and / or the condition label and the algorithm label is formed.

5. The method according to claim 1, characterized in that The scheduling of executing the algorithm task through the relational model includes: When a trigger condition of an algorithm task is received, the relationship model is retrieved; According to the task triggering condition, converting the relational model into an executable scheduling instruction; The scheduling of the algorithm tasks is performed by the scheduling instructions.

6. The method according to claim 1, characterized in that After the scheduling of executing the algorithm task through the relational model, the method further includes: Evaluate the execution of the algorithm task by using a preset evaluation method; When the evaluation of the execution status does not satisfy a preset evaluation condition, the relationship model is updated and iterated, and the scheduling of subsequent algorithm tasks is performed through the updated and iterated relationship model.

7. The method according to claim 6, characterized in that The updating and iterating of the relationship model includes: Updating the second association information of the current iteration cycle by using the first association information of the previous iteration cycle; Updating the relationship model of the current iteration period by using the second association information of the current iteration period; Scheduling of algorithm tasks is performed through the relational model of the current iteration cycle; If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle does not meet the preset evaluation conditions, then enter the next iteration cycle; If the execution status of the algorithm task corresponding to the relational model of the current iteration cycle meets the preset evaluation condition, the iteration is stopped and the relational model of the current iteration cycle is stored.

8. An algorithm task scheduling device, characterized in that: The device comprises: An acquisition module, used to acquire first associated information of each equipment point, wherein the first associated information includes image information, a point label and a condition label, wherein the image information, the point label and the condition label are associated with each other through a point identifier, and each equipment point corresponds to a point identifier; A parsing module, used for parsing each of the image information by using a preset algorithm set to obtain corresponding parsing content, wherein the preset algorithm set includes a plurality of preset algorithms, and each of the preset algorithms outputs a parsing content for the image information; an association module, configured to form an algorithm tag with the analysis content corresponding to the image information and the corresponding preset algorithm for each of the image information, and add the point identifier to the corresponding first association information to form second association information, wherein each algorithm tag corresponds to one preset algorithm; A processing module is used to construct a relationship model between the point label and / or the condition label and the algorithm label based on the second association information, and execute the scheduling of the algorithm task through the relationship model.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the algorithmic task scheduling method as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the algorithmic task scheduling method according to any one of claims 1 to 7 are implemented.

Images