Station inspection method and system based on artificial intelligence

Abstract

The invention discloses a station inspection method and system based on artificial intelligence. The method comprises the following steps: generating a first-class inspection task and a second-class inspection task according to a fault prompt and a special prompt of a station operation and maintenance system; for the first-class inspection task, the first-class inspection task is directionally issued through the acquired employee portrait; for the second-class inspection task, the task is issued immediately after being generated, and the task is issued through the order grabbing operation of the inspection personnel; receiving a polling result returned by the polling personnel for the first-class polling task and / or the second-class polling task, wherein the polling result at least comprises a polling video; based on the task content of the first-class inspection task and / or the second-class inspection task, performing inspection operation analysis and / or equipment state analysis on the inspection video, and judging whether an analysis result meets an inspection compliance mechanism or not; if yes, abnormal factors in the station are recognized through the inspection video, and a response mechanism is generated; and if not, correspondingly issuing a re-inspection task to the inspection personnel based on the analysis result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of station inspection, and in particular to a station inspection method and system based on artificial intelligence. BACKGROUND

[0002] At present, the inspection work of traffic sites such as stations and subway stations is mostly that personnel first find faults and then report the faults, and maintenance dispatch personnel receive the reported faults, integrate the fault information into an inspection task, and then send the inspection task to a shift foreman, who then assigns the received inspection task to different maintenance personnel or inspection personnel. If the maintenance personnel or inspection personnel do not find any abnormalities during the inspection, the shift foreman will then randomly check the video data and other data of the inspection, and if the maintenance personnel or inspection personnel find any abnormalities during the inspection, the abnormalities will be returned to the shift foreman, who will then develop an abnormality handling plan and generate a maintenance work order for distribution. Similarly, for special investigation tasks in the station, the tasks also need to be distributed to the shift foreman by a special worker, and then the shift foreman needs to assign the tasks.

[0003] In this process, on the one hand, there are many nodes that require personnel participation, resulting in a large amount of human cost consumed for the station inspection, and the distribution and assignment of tasks all require personnel participation, resulting in low efficiency of task assignment and thus low efficiency of station inspection. On the other hand, the discovery of abnormal conditions in the station is determined by manual judgment of the inspection personnel, which may result in missed inspection or incorrect inspection, and also limits the efficiency of station inspection. SUMMARY

[0004] The present application provides a station inspection method and system based on artificial intelligence to solve at least one of the above technical problems.

[0005] The present application adopts the following technical solutions: In an aspect, the application provides an artificial intelligence-based station inspection method, which comprises: generating a first type of inspection task and a second type of inspection task according to fault prompts and special prompts of a station operation and maintenance system; for the first type of inspection task, obtaining an employee portrait of an inspection personnel, and directing the first type of inspection task to the inspection personnel through the employee portrait, the employee portrait being generated according to the execution of historical inspection tasks by the inspection personnel; for the second type of inspection task, issuing immediately after task generation, and issuing through the inspection personnel's grab operation; receiving the inspection results returned by the inspection personnel for the first type of inspection task and / or the second type of inspection task, the inspection results at least including an inspection video; based on the task content of the first type of inspection task and / or the second type of inspection task, performing inspection operation analysis and / or device state analysis on the inspection video, and determining whether the analysis result meets the inspection compliance mechanism; if yes, identifying abnormal factors in the station through the inspection video, and generating a response mechanism; if not, issuing a re-inspection task to the inspection personnel based on the analysis result.

[0006] In a possible implementation manner of the application, the first type of inspection task and the second type of inspection task are generated according to fault prompts and special prompts of a station operation and maintenance system, which comprises: determining fault points corresponding to the fault prompts of the station operation and maintenance system, and obtaining monitoring videos and / or monitoring images of the fault points; performing fault discrimination on the monitoring videos and / or monitoring images, if it is determined that there is a fault, generating an emergency response task, and determining the emergency response task as the first type of inspection task; obtaining prompt content corresponding to the special prompts of the station operation and maintenance system, the special prompts at least including one of flood season risk prompts, air conditioning season risk prompts and station passenger flow risk prompts; generating a planned inspection task and / or a cross-checking task according to the prompt content, and determining the planned inspection task and / or the cross-checking task as the second type of inspection task.

[0007] In one possible implementation of this application, the employee profile is generated based on the inspector's performance of historical inspection tasks, including: for any inspector whose employee profile is to be generated, obtaining the inspector's completed historical inspection tasks and the inspector's registration information, wherein the registration information includes at least an inspection certificate; extracting the inspection objects of the historical inspection tasks based on the task content of the historical inspection tasks, and calculating the frequency of occurrence of the inspection objects; matching the inspection objects with a frequency higher than a preset frequency threshold with the inspection certificates; if the two match successfully, determining the successfully matched inspection object as the inspector's first label, otherwise, determining the inspection certificate as the inspector's first label; obtaining the inspection scores of the historical inspection tasks, and clustering the task content corresponding to the historical inspection tasks with scores higher than a preset score threshold; generating the inspector's second label based on the clustering results; and generating the inspector's employee profile based on the first label and the second label.

[0008] In one possible implementation of this application, the targeted distribution of the first type of inspection task through the employee profile includes: determining the task content and inspection target of the first type of inspection task; matching the inspection target with a first tag of the employee profile to obtain a first matching degree, and matching the task content with a second tag to obtain a second matching degree; determining the inspection personnel to be assigned the task based on the maximum value of the weighted sum of the first matching degree and the second matching degree; and targeted distribution of the first type of inspection task to the inspection personnel to be assigned the task.

[0009] In one possible implementation of this application, after the type of inspection task is directed and distributed through the employee profile, the method further includes: receiving the execution time feedback returned by the inspection personnel to be dispatched after receiving the type of inspection task; if the execution time feedback indicates immediate execution, then the directed distribution of the type of inspection task is completed; if the execution time feedback indicates pending execution, then it is determined whether the waiting time for pending execution is greater than a preset time threshold; if so, then the type of inspection task is redistributed until the execution time feedback indicates immediate execution or the waiting time for pending execution is less than the preset time threshold.

[0010] In one possible implementation of this application, the second type of inspection task is issued through the order-grabbing operation of the inspection personnel, including: obtaining the average daily order-grabbing volume of the inspection personnel in a historical time period and the number of orders already grabbed by the inspection personnel on the current day; when the number of orders already grabbed on the current day exceeds ±15% of the average daily order-grabbing volume, the order-grabbing operation of the inspection personnel is rejected.

[0011] In one possible implementation of this application, if the number of orders already accepted on the day does not exceed ±15% of the average number of orders accepted on the day, the method further includes: determining whether the inspection personnel have a type I inspection task that has been specifically assigned; if so, determining whether the latest execution time of the type II inspection task corresponding to the order-accepting operation conflicts with the planned execution time period of the type I inspection task; if so, rejecting the order-accepting operation of the inspection personnel.

[0012] In one possible implementation of this application, the inspection result further includes an inspection conclusion, which is the conclusion given by the inspection personnel regarding the presence or absence of abnormalities in the first type of inspection task and / or the second type of inspection task. When the inspection conclusion indicates an abnormality, the method further includes: determining that the inspection video meets the inspection compliance mechanism; identifying the preset inspection actions of the inspection personnel in the inspection video, and determining abnormal factors within the station based on the preset inspection actions, wherein the preset inspection actions are designated actions or instruction actions issued by the inspection personnel when there are abnormalities in the first type of inspection task and / or the second type of inspection task.

[0013] In one possible implementation of this application, when the inspection conclusion is no abnormality, the method further includes: determining that the inspection video meets the inspection compliance mechanism; identifying abnormal factors and maintenance operations in the inspection video according to the task content of the first-class inspection task and / or the second-class inspection task; if the abnormal factors are not identified and / or the maintenance operations are identified, adjusting the fault prompt mechanism of the station operation and maintenance system, and / or adjusting the generation cycle of planned inspection tasks and / or cross-inspection tasks; if the abnormal factors are identified but the maintenance operations are not identified, generating a response mechanism for the abnormal factors, and adjusting the employee profile and / or daily order-grabbing range of the inspection personnel who gave the inspection conclusion.

[0014] On the other hand, this application also provides an artificial intelligence-based station inspection system, the system comprising: a task generation sub-agent, which generates first-class and second-class inspection tasks based on fault prompts and special prompts from the station operation and maintenance system; a task management sub-agent, which, for the first-class inspection tasks, obtains employee profiles of the inspection personnel and distributes the first-class inspection tasks accordingly, the employee profiles being generated based on the inspection personnel's historical inspection task execution; and the task management sub-agent, for the second-class inspection tasks, immediately publishes them after task generation and distributes them through the inspection personnel's order-grabbing operation; and inspection results. The processing sub-agent receives the inspection results returned by the inspection personnel for the first type of inspection task and / or the second type of inspection task, the inspection results including at least inspection video; the inspection result processing sub-agent performs inspection operation analysis and / or equipment status analysis on the inspection video based on the task content of the first type of inspection task and / or the second type of inspection task, and determines whether the analysis results meet the inspection compliance mechanism; if they do, it identifies abnormal factors in the station through the inspection video and generates a response mechanism; if they do not meet the requirements, it issues a re-inspection task to the inspection personnel through the task management sub-agent based on the analysis results.

[0015] The station inspection method and system based on artificial intelligence provided in this application have the following beneficial effects: This application generates two types of inspection tasks for fault alerts and special alerts within the station, respectively. Different task assignment modes apply to different tasks; for example, type I tasks are assigned via targeted assignment, while type II tasks are assigned by inspectors through a bidding process. This task assignment mechanism ensures the effective execution of important inspection tasks while also increasing inspectors' enthusiasm for bidding, positively impacting station inspection efficiency. After inspectors complete their tasks and return the inspection video, this application assesses the video against the compliance mechanism to guarantee the compliance of the inspection process. When the video meets the compliance mechanism, abnormal factors within the station are identified, avoiding the objectivity issues, missed inspections, and incorrect inspections inherent in traditional manual anomaly assessments. Conversely, when the video does not meet the compliance mechanism, a re-inspection task is assigned, requiring inspectors to re-execute the task, thus ensuring the compliance and effectiveness of the inspection results and improving inspection efficiency.

[0016] Furthermore, when allocating Category II inspection tasks, the tasks are assigned based on the order-grabbing actions of the inspectors. However, in order to ensure the effectiveness and efficiency of the inspections, this application limits the number of orders a inspector can grab on a given day to no more than ±15% of the average number of orders they can grab on a given day. This ensures the timely assignment of Category II inspection tasks while also guaranteeing the effectiveness of the inspectors in executing these tasks. It also prevents the passive execution of inspection tasks due to excessive order grabbing, thereby helping to improve the inspection efficiency of the station.

[0017] Furthermore, the inspection results returned by the inspection personnel in this application also include the inspection conclusions given by the inspection personnel. When the inspection conclusion is no abnormality, in order to avoid missed or incorrect inspections, the inspection video will be subjected to abnormal factor identification and maintenance operation identification to determine whether the inspection personnel failed to identify abnormal factors or performed maintenance operations after identifying abnormal factors, thus giving a conclusion of no abnormality. To a certain extent, this can effectively avoid missing abnormal factors and improve inspection efficiency.

[0018] Finally, the station inspection solution proposed in this application is entirely implemented by the corresponding sub-agents, from task generation to task issuance and task execution result review. This avoids human intervention and changes the traditional "human-based business-driven process" to "agent-based data-driven process". This makes the agent the core of the inspection work, responsible for task generation, task issuance, and task execution result review, while humans only need to perform the inspection tasks. Compared with the traditional inspection process, this greatly reduces labor costs and significantly improves the efficiency of station inspection. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart of an artificial intelligence-based station inspection method provided for this application; Figure 2 This application provides a schematic diagram of an AI-based station inspection system architecture. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0021] The method in this application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 A flowchart of an artificial intelligence-based station inspection method is provided for this application, such as... Figure 1 As shown, the station inspection method in this application includes at least the following steps: Step 101: Based on the fault prompts and special prompts from the station operation and maintenance system, generate Class I inspection tasks and Class II inspection tasks.

[0023] The station inspection method described in this application is applicable to transportation stations such as bus stations, subway stations, and train stations. The stations are equipped with conventional operation and maintenance systems that can monitor the station's condition, provide maintenance reminders to inspection personnel, and offer fault indications and specific alerts.

[0024] For example, if the station's maintenance system detects that an escalator has stopped operating, it will issue a fault notification. Another example is that, in summer, to provide a comfortable environment for passengers, the station's air conditioning system should be kept running continuously. Therefore, at regular intervals, the maintenance system will issue a special notification for a check of the air conditioning system.

[0025] Of course, the fault prompts provided by the station's operation and maintenance system can also be generated based on reports of on-site faults submitted by passengers or crew members. For example, if a passenger finds that the ticket machine is not issuing tickets properly when purchasing tickets on-site, they will seek help from the crew. If the crew member confirms that the ticket machine is indeed not issuing tickets, they will upload the fault information to the operation and maintenance system. Upon receiving the fault information, the operation and maintenance system will generate a fault prompt.

[0026] In one possible implementation of this application, an inspection task is generated based on the fault notification from the operation and maintenance system. Specifically, after the station's operation and maintenance system issues a fault notification, it extracts the fault point corresponding to the fault notification. Then, it uses the monitoring function of the operation and maintenance system to determine the monitoring video and / or monitoring image corresponding to the fault point. In the monitoring video and / or monitoring image, it is determined whether the current fault point actually has a fault. If it does, an emergency response task is generated, and this emergency response task is identified as an inspection task. For example, after the operation and maintenance system issues a fault notification that an escalator has stopped operating, the solution in this application will extract the monitoring video and / or monitoring image corresponding to the escalator based on the fault point of the escalator in the fault notification, and identify it to determine whether the escalator in the monitoring video has actually stopped operating. If the escalator is found to have stopped operating and has a fault, an emergency response task corresponding to the escalator is generated to conduct an emergency inspection of the escalator and investigate the abnormal factors or causes that led to the stoppage. It should be noted that although the aforementioned content describes the fault prompts provided by the maintenance system based on the escalator's operational status monitored by its own monitoring functions, this solution does not directly generate inspection tasks upon receiving a fault prompt. Instead, it further verifies the fault prompt, and only generates an inspection task after confirming that the escalator does indeed have a fault. This ensures that all generated inspection tasks are valid, eliminating unnecessary inspections and guaranteeing the effective execution of each task, thus improving station inspection efficiency. Furthermore, it effectively avoids the generation of invalid inspection tasks due to staff misreporting repairs or erroneous monitoring by the maintenance system, preventing the waste of inspection resources. Emergency response tasks are only generated after confirming that a fault exists at the fault point. In other words, this solution performs secondary verification or integration of fault prompts and / or repair reports before generating emergency response tasks to ensure that all generated emergency response tasks are valid, thereby improving station inspection efficiency and avoiding invalid inspections.

[0027] Furthermore, based on the specific prompts from the operation and maintenance system, second-class inspection tasks are generated. Specifically, based on the content of the specific prompts from the operation and maintenance system, planned inspection tasks and / or cross-inspection tasks are generated, and these generated planned inspection tasks and / or cross-inspection tasks are identified as second-class inspection tasks. In one example, the content of the specific prompts includes at least one of the following: flood season risk warning, air conditioning season risk warning, and station passenger flow risk warning. For example, in the hot summer, the air conditioning system in the station is generally in a constantly running state. Therefore, the operation and maintenance system can provide a risk warning about the air conditioning, which is mainly to remind the system to conduct routine inspections. In this case, this application can generate planned inspection tasks based on this specific prompt. These planned inspection tasks are, for example, tasks that are scheduled to inspect the air conditioning system at a certain period, or generate cross-inspection tasks. These cross-inspection tasks are, for example, tasks that periodically inspect the air conditioning system in different locations, such as inspecting the air conditioning system on the station concourse level during the first inspection and inspecting the air conditioning system on the station platform level during the second inspection.

[0028] It should be noted that the examples above are only for more detailed illustration of the generation process of Category I and Category II inspection tasks in this application, and are not intended to limit this application. Of course, in actual station operation scenarios, there are many other situations not listed above that require fault or special notifications, and this application will also generate Category I and Category II inspection tasks for these unlisted situations. In one example, the generation principle for Category I and Category II inspection tasks can be summarized as follows: notifications of faults or maintenance needs within the station correspond to the generation of Category I inspection tasks, while notifications of routine inspections within the station correspond to the generation of Category II inspection tasks.

[0029] Step 102: For a type of inspection task, obtain the employee profile of the inspection personnel, and then distribute the type of inspection task in a targeted manner based on the employee profile.

[0030] This application generates different inspection tasks based on different prompts from the station operation and maintenance system. Different inspection tasks correspond to different task assignment methods. This allows the task assignment process of this application to not only save the manpower costs consumed in the traditional task assignment process, but also to assign first-class and second-class inspection tasks in a targeted manner, ensuring that all inspection tasks can be executed efficiently, thereby improving the station's inspection efficiency.

[0031] In one possible implementation of this application, taking a certain station as an example, the station will be equipped with corresponding inspection personnel, and this application will generate employee profiles for the inspection personnel in the station. Thus, when a type of inspection task is issued, the inspection task can be targeted based on the aforementioned employee profile, so that the inspection task can be executed more efficiently.

[0032] Specifically, when generating employee profiles for inspection personnel, historical inspection tasks completed by the personnel within a past period can be extracted, such as those completed within the past year or six months. Selecting a timeframe that is too long may lead to inaccurate labels that don't accurately reflect the inspector's current status. Conversely, selecting a timeframe that is too short may result in insufficient task completion and potentially inaccurate labels. Therefore, selecting historical inspection tasks from the past six months to a year is optimal. Then, based on the content of these historical inspection tasks, the inspection targets are extracted. For example, if the task was to cut off the power to an escalator, the extracted inspection target would be the escalator. The frequency of occurrence of extracted inspection objects is calculated. Inspection objects with a frequency higher than a preset threshold are matched with inspection certificates uploaded by inspectors. If a match is found, it indicates that the inspector is proficient in inspecting these objects, and the matched object is designated as the inspector's primary tag. However, if no match is found, the inspection certificate may be a newly obtained certificate for other inspection objects, and in this case, it can be used as the inspector's primary tag. Of course, inspectors typically obtain certificates based on their areas of expertise, so this scenario is unlikely. It should also be noted that calculating the match between inspection certificates and inspection objects can be achieved using existing similarity calculation algorithms. For example, calculating the similarity between the inspection object and the object recorded in the inspection certificate. In this case, a successful match is considered to be achieved if the calculated similarity is higher than a preset similarity threshold, such as 80% or higher. Alternatively, one can directly search for objects in the inspection certificate with names matching the inspection object; a successful search indicates a successful match. Both of the above schemes can match inspection certificates with inspection objects, and their implementation process can be achieved through existing algorithms. Therefore, this application will not elaborate further on this, but will focus on the labels determined for inspection personnel.

[0033] Furthermore, after obtaining the historical inspection tasks of the inspectors over the past six months or one year, the inspection score corresponding to each inspection task is extracted. Then, the task content of historical inspection tasks with inspection scores higher than a preset score threshold is extracted. A high score indicates that the inspector performed well in completing the inspection task, that is, the inspector is skilled in performing this type of inspection task. Therefore, the extracted task content is clustered. This clustering process or cluster analysis can be implemented using existing clustering algorithms, which will not be elaborated here. After clustering is completed, a second label for the inspector is generated based on the clustering results. For example, the category corresponding to the largest cluster in the clustering results is used as the second label for the inspector.

[0034] It should also be noted that the above process of determining the first label and the second label is only a different description of the order and is not intended to limit the order of determining the first label and the second label. In actual scenarios, the first label can be determined first based on the inspection object and the inspection certificate, and then the second label can be determined based on the task content of the inspection task, or the second label can be determined first based on the task content of the inspection task, and then the first label can be determined based on the inspection object and the inspection certificate.

[0035] After determining the first and second labels, the first and second labels are used as attribute labels for the inspectors, thus completing the generation of the employee profile corresponding to the inspectors.

[0036] In one possible implementation of this application, a type of inspection task is typically an emergency response task generated upon fault notification. Therefore, it is desirable to assign or distribute this type of inspection task to inspection personnel with stronger inspection capabilities to quickly locate abnormal factors and achieve efficient inspection. Thus, this application preferably implements targeted distribution of this type of inspection task. During distribution, the task content and inspection object of the type of inspection task are first extracted. Then, a first matching degree is calculated between the first tag of the inspection object and the first tag of the employee profile, and a second matching degree is calculated between the second tag of the task content and the second tag of the employee profile. After weighted summation of the first and second matching degrees, the employee profile corresponding to the maximum value is determined, and the type of inspection task is targeted and distributed to the inspection personnel with the employee profile corresponding to that maximum value.

[0037] It should be noted that the calculation of the first and second matching degrees in the above implementation methods can be achieved through existing calculation methods, which will not be elaborated here. Furthermore, the employee profile in this application can also generate other corresponding tags based on other items in historical inspection tasks, and is not limited to generating only the first and second tags. When other tags exist in the employee profile, the matching degree corresponding to the other tags also needs to be calculated when calculating the targeted issuance of a type of inspection task. That is, this application does not limit the calculation of only the first and second matching degrees when determining the targeted inspection personnel, but can adapt and adjust according to the different actual employee profiles.

[0038] In one possible implementation of this application, a type of inspection task is typically an emergency response task. This type of task is expected to be executed as quickly as possible, or even immediately after being assigned. Therefore, after assigning a type of inspection task, an execution time feedback is requested from the receiving inspector. This execution time feedback indicates the execution time given by the receiving inspector. If the received execution time feedback indicates immediate execution, the assignment of the type of inspection task is complete. However, if the received execution time feedback indicates waiting for execution, it means the inspector currently has a task in progress, and the assigned type of inspection task needs to wait for the currently executing task to complete before it can be executed. Therefore, it is determined whether the waiting time exceeds a preset time threshold. If so, it means that the assigned type of inspection task needs to wait a relatively long time before it can be executed, but it is expected to be executed as quickly as possible. Therefore, in this case, the assigned type of inspection task is redistributed until the execution time feedback indicates immediate execution or the waiting time is less than the preset time threshold. In other words, once a type of inspection task is assigned, the execution time reported by the inspection personnel ensures that the task can be executed immediately or after a short wait. This enables the station to quickly inspect faults in the event of a station malfunction, improving inspection efficiency and facilitating rapid fault elimination at the station.

[0039] Step 103: For Category II inspection tasks, release the task immediately after it is generated and distribute it through the order-grabbing operation of inspection personnel.

[0040] Category II inspection tasks typically include planned inspection tasks and / or cross-inspection tasks. These are generally routine inspection tasks, and are released immediately after they are generated. "Relevant release" means that the task is released immediately after it is generated. Unlike Category I inspection tasks, which are distributed to inspection personnel in a targeted manner, inspection personnel bid for the task online through their own terminals. The terminals of the inspection personnel can include at least one of the following: mobile phone, computer, tablet, and maintenance terminal / equipment.

[0041] In one possible implementation of this application, since the second-class inspection tasks are issued through the order-grabbing operation of the inspectors, and the inspectors usually hope to grab as many orders as possible, some inspectors may grab too many orders, exceeding their inspection capacity, resulting in the second-class inspection tasks not being completed in a timely manner, or even the second-class inspection tasks being performed perfunctorily, thereby affecting the efficiency of the inspection tasks. To avoid this situation, when issuing Category II tasks through the order-grabbing operation of inspection personnel, this application, upon receiving an order-grabbing request from an inspection personnel, will obtain the number of orders already grabbed by that inspection personnel that day, and then compare that number with the inspection personnel's average daily order-grabbing volume. In one example, the inspection personnel's average daily order-grabbing volume can be obtained by dividing the inspection personnel's total order-grabbing volume within a historical time period by the corresponding number of days in the historical time period. At the same time, by comparing the inspection personnel's number of orders already grabbed that day with the average daily order-grabbing volume, an allowable range for the inspection personnel's order-grabbing volume can be defined using a range of ±15% of the average daily order-grabbing volume. When the inspection personnel's number of orders already grabbed that day does not exceed this allowable range, the application will respond to the inspection personnel's order-grabbing operation and issue the Category II inspection task to the person who first performed the order-grabbing operation. However, if an inspector's daily order volume exceeds the allowed limit, the inspector's order-grabbing operation will be rejected, which means rejecting the order-grabbing request sent by the inspector's terminal based on the order-grabbing operation.

[0042] Furthermore, because some inspection personnel may be assigned Category I inspection tasks, which are typically emergency response tasks expected to be executed quickly or with priority, these personnel also need to consider the execution time of Category II inspection tasks when bidding for them. Taking planned inspection tasks within Category II as an example, these tasks may be scheduled for a certain period, thus requiring timely execution. Executing them too late may affect the next round of inspections. Therefore, when responding to an inspector's order-grabbing operation, and determining that the inspector's daily order-grabbing volume does not exceed ±15% of the daily average order-grabbing volume (i.e., when allowing the inspector to grab orders), it is first determined whether the inspector has been assigned a Category 1 inspection task. If so, it is determined whether the latest execution time of the Category 2 inspection task the inspector wants to grab conflicts with the planned execution time of the existing Category 1 inspection task. If there is a conflict, the inspector's order-grabbing operation is rejected. However, if the inspector does not have any Category 1 inspection tasks assigned to them, or if the Category 2 inspection task the inspector wants to grab does not have a latest execution time (i.e., no execution time limit), or if the latest execution time of the Category 2 inspection task the inspector wants to grab does not conflict with the planned execution time of the Category 1 inspection task, then the Category 2 inspection task can be assigned to the inspector.

[0043] For example, inspector A wants to accept an order for a type 2 inspection task (a). Inspector A sends an order-accepting request on their terminal. Assume that inspector A is the first to accept an order for this type 2 inspection task, meaning the terminal that first receives the inspection request is inspector A's. In response to this order-accepting action or request, it is determined whether inspector A's daily order-accepting volume is less than the allowed range (assuming inspector A's average daily order-accepting volume is 10, then their allowed range is considered to be 9-12). If it exceeds this range, inspector A's order-accepting action is rejected, and the process continues to inspector B, who is the second to accept an order. If no time limit is exceeded, the process continues to determine if inspector A has been assigned a Category 1 inspection task b. If not, Category 2 inspection task a is assigned to inspector A. If it exists, the planned execution time of Category 1 inspection task b is obtained. Since there may be more than one Category 1 inspection task assigned, it is possible that Category 1 inspection task b is waiting in the queue for execution (inspector A is currently executing other Category 1 inspection tasks). Therefore, there will be a planned execution implementation, as well as obtaining the latest execution time of Category 2 inspection task a. Then, it is determined whether the planned execution time is before the latest execution time. Assuming the planned execution time is 10-20 minutes in the future and the latest execution time is 25 minutes in the future, it means that after inspector A completes the assigned Category 1 inspection task b, there is still time to execute Category 2 inspection task a. In this case, Category 2 inspection task a can be assigned to inspector A.

[0044] It should be noted that the above process describes the issuance of Category I and Category II inspection tasks from the perspective of task assignment, as well as the potential conflicts that may arise during the assignment process. It ensures that Category I inspection tasks are executed with priority while also preventing Category II inspection tasks from being executed too late. Furthermore, the inspection personnel in the above example are generally from a specific station, and the station's area is limited. Therefore, even if Category I inspection task b and Category II inspection task a are located at different locations within the station, the time spent by the inspection personnel traveling to execute these two tasks will be negligible. Of course, when determining the planned execution time of Category I inspection tasks and the latest execution time of Category II inspection tasks, the time required for the transfer between the two tasks can be taken into account. For example, the planned execution time of Category I inspection tasks can be delayed by 5-10 minutes to cover the transfer time for the inspection personnel to reach the execution point of Category II inspection tasks.

[0045] Step 104: Receive the inspection results returned by the inspection personnel for Category I and / or Category II inspection tasks.

[0046] Once inspectors receive a Class I or Class II inspection task, they begin executing the task. After each task is completed, the inspection results are returned. In one example, the inspection results include at least inspection video, which is the video data collected by inspectors using inspection equipment during the execution of the task.

[0047] Step 105: Based on the task content of Category I and / or Category II inspection tasks, perform inspection operation analysis and / or equipment status analysis on the inspection video, and determine whether the analysis results meet the inspection compliance mechanism.

[0048] After receiving the inspection video returned by the inspection personnel, the video is analyzed according to the task content. Specifically, the analysis can focus on the inspection personnel's inspection operations and the status of the equipment to be inspected, such as its operational status. In one example, the inspection personnel are performing a type II inspection task to inspect the station's air conditioning system. Assuming the task is a routine inspection of the air conditioning system, after the inspection personnel return the video, the operational status of the air conditioning system should be identified. Under normal circumstances, the operational status should be "running." Therefore, the operational status of the air conditioning system can be determined by identifying the positional relationship of the air conditioning vent fan blades in consecutive video frames, or by observing the position of instruments / equipment such as wind direction and speed meters when placed at the air conditioning system vents. In this example, the operational status of the air conditioning system and the operational status of the wind direction and speed meters are analyzed. If both indicate that the air conditioning system is operating normally, then the inspection video meets the inspection compliance mechanism, meaning the inspection video is compliant. In addition, when inspectors inspect the air conditioning system, they perform inspection operations on the outdoor units. The inspection compliance mechanism predefines different inspection operations for different equipment. The inspection operation is identified by the inspection video, and then compared with the predetermined inspection operations in the inspection compliance mechanism to obtain the analysis results of the inspection operation. If the comparison determines that the inspection operation matches the predetermined inspection operation, it can also be determined that the inspection video meets the inspection compliance mechanism.

[0049] It should be noted that, in the above example, the inspection video analysis can first be performed by extracting frames from the inspection video to obtain an inspection frame sequence, which is a video frame sequence ordered by time. Then, equipment operation status analysis and / or inspection operation analysis can be performed on the inspection frame sequence. Taking equipment operation status analysis as an example, artificial intelligence can be used to identify the operation status. For instance, the inspection frame sequence can be fed into a pre-trained Transformer model or LSTM model, which can then classify and judge the inspection frame sequence, outputting normal or abnormal. It should be noted that the training and classification processes of the aforementioned two deep learning models can be implemented using existing algorithms, and this application will not elaborate on them.

[0050] Of course, during the inspection, the inspectors in this application can also use other instruments or equipment to inspect the station equipment such as the air conditioning system. In this case, the equipment operating status analyzed when analyzing the inspection video is preferably the equipment used by the inspectors to assist in performing the inspection task. For example, when performing the inspection task, the inspectors can place sensors such as sound, vibration, infrared temperature, power, or cameras at key locations of the station equipment. Thus, when conducting a compliance review of the returned inspection video, the analyzed equipment operating status is the operating status of the aforementioned distance sensor, such as determining whether the sensor is collecting data normally.

[0051] Step 106: If the conditions are met, abnormal factors within the station will be identified through inspection videos, and a response mechanism will be generated.

[0052] If the inspection video meets the inspection compliance mechanism, that is, if the inspection video is compliant, the inspection video is used to identify abnormal factors in the station, and a response mechanism is generated for the abnormal factors. The response mechanism includes at least the maintenance mechanism.

[0053] In one possible implementation of this application, when performing inspection tasks, the inspection personnel may have discovered abnormal factors that could lead to equipment failure. Therefore, this application may require the inspection personnel to return an inspection conclusion along with the inspection video. This inspection conclusion is used to indicate whether there are any abnormalities in the current inspection video.

[0054] Furthermore, if the inspection personnel return an inspection conclusion indicating an anomaly along with the inspection video, the pre-defined inspection actions of the personnel can be directly identified in the video (for example, requiring the personnel performing the inspection task to point to the anomaly upon discovery). Then, based on this pointing-to-anomaly action, the abnormal factors within the station can be determined. It should be noted that identifying specific actions in the inspection video can be achieved through two methods: firstly, direct identification using deep learning models such as convolutional neural networks; and secondly, matching algorithms can be used to match video frames with standard images of pre-defined inspection actions for similarity calculation. Both methods can achieve the identification of pre-defined inspection actions, and both can be implemented using existing algorithms, which will not be elaborated upon here.

[0055] In one example, as illustrated in the related examples above, when inspectors perform their inspection tasks, they can deploy sensors such as sound, vibration, infrared temperature, power, or cameras at key locations on station equipment such as air conditioning systems for auxiliary inspection. Therefore, when using inspection videos to identify abnormal factors, if the preset inspection actions issued by the inspectors are not detected, the data collected by these sensors (which can be read directly from the video or received from the sensors) can be used to identify the normal or abnormal status of the station equipment to be inspected using data analysis algorithms.

[0056] Furthermore, if the inspection personnel return a conclusion of "no abnormalities" along with the inspection video, it means that the inspection personnel did not find any abnormalities during the inspection task. Alternatively, it means that the inspection personnel found an abnormality during the inspection task but repaired it, such as finding that a light in the station was not working and discovering a short circuit during the inspection, which was then repaired directly. In this case, the judgment is made by identifying the abnormal factors and repair operations through the inspection video. If no abnormal factors, or no repair operations, or neither, are identified in the inspection video, it means that no abnormalities were found during the inspection task. In this case, the fault indication mechanism of the station operation and maintenance system and / or the generation cycle of planned inspection tasks and / or cross-inspection tasks can be adjusted, that is, the inspection cycle of Category I and / or Category II inspection tasks can be adjusted. If an anomaly is identified in the inspection video but no maintenance operation is detected, the inspection personnel will conclude that there is no anomaly. This means that the inspection personnel may have missed the anomaly during the inspection, resulting in missed or incorrect inspections. To solve this problem, this application generates a response mechanism for anomalies, such as generating and dispatching maintenance work orders. At the same time, it will adjust the employee profile of the inspection personnel to reduce the likelihood of them being assigned Category 1 inspection tasks, and / or adjust the daily order-grabbing range of the inspection personnel, prioritizing reducing their daily order-grabbing range to minimize the number of Category 2 inspection tasks they perform, thereby ensuring the efficient execution of inspection tasks.

[0057] Step 107: If the conditions are not met, a re-inspection task will be issued to the corresponding inspection personnel based on the analysis results.

[0058] If the inspection videos returned by the inspectors do not meet the inspection compliance mechanism, meaning the execution of the inspection task does not meet compliance requirements, then a re-inspection task will be issued, requiring the inspectors to redo / re-execute the inspection task. Alternatively, the non-compliant inspection task can be reissued to other inspectors.

[0059] The above is a description of the method in this application, which mainly focuses on three stages: task generation, task issuance, and task execution result processing. All three stages avoid human intervention. From the station operation and maintenance system issuing fault prompts or special prompts, to the generation and issuance of Class I and Class II inspection tasks, and then to the review and anomaly identification after the inspection results are returned, all are realized based on artificial intelligence technology. Compared with traditional station inspection solutions, this method not only reduces labor costs but also helps to improve the efficiency of station inspections and achieve high-efficiency inspections.

[0060] Based on the same inventive concept, this application also provides an artificial intelligence-based station inspection system, the architecture of which is as follows: Figure 2 As shown.

[0061] Figure 2 This application provides a schematic diagram of an AI-based station inspection system architecture. Figure 2 As shown, the AI-based station inspection system 200 in this application specifically includes: a task generation sub-intelligent agent 201, which generates a first-class inspection task and a second-class inspection task based on fault prompts and special prompts from the station operation and maintenance system; a task management sub-intelligent agent 202, which, for the first-class inspection task, obtains the employee profile of the inspection personnel and distributes the first-class inspection task accordingly, the employee profile being generated based on the inspection personnel's historical inspection task execution; the task management sub-intelligent agent 202, for the second-class inspection task, immediately publishes it after task generation and distributes it through the inspection personnel's order-grabbing operation; and an inspection result processing sub-intelligent agent. Intelligent agent 203 receives inspection results returned by the inspection personnel for the first type of inspection task and / or the second type of inspection task, the inspection results including at least inspection video; the inspection result processing sub-intelligent agent 203, based on the task content of the first type of inspection task and / or the second type of inspection task, performs inspection operation analysis and / or equipment status analysis on the inspection video, and determines whether the analysis results meet the inspection compliance mechanism; if they do, it identifies abnormal factors in the station through the inspection video and generates a response mechanism; if they do not meet the requirements, it issues a re-inspection task to the inspection personnel through the task management sub-intelligent agent 202 based on the analysis results. It should be noted that each sub-agent in the above system is preferably an artificial intelligence agent.

[0062] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0063] The system and method provided in this application are one-to-one correspondences. Therefore, the system also has similar beneficial technical effects as its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system will not be repeated here.

[0064] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0066] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0067] The above are merely embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An artificial intelligence-based station inspection method, characterized by, The method comprises: According to the fault prompt and special prompt of the station operation and maintenance system, a first type of inspection task and a second type of inspection task are generated; For the first type of inspection task, an employee portrait of the inspection personnel is obtained, and the first type of inspection task is directedly issued through the employee portrait, and the employee portrait is generated according to the execution of the historical inspection task by the inspection personnel; For the second type of inspection task, the task is published immediately after generation and is issued through the grab operation of the inspection personnel; The inspection result returned by the inspection personnel for the first type of inspection task and / or the second type of inspection task is received, and the inspection result at least includes an inspection video; Based on the task content of the first type of inspection task and / or the second type of inspection task, the inspection operation analysis and / or the equipment state analysis are performed on the inspection video, and it is judged whether the analysis result meets the inspection compliance mechanism; If yes, the abnormal factors in the station are identified through the inspection video, and a response mechanism is generated; If not, a re-inspection task is issued to the inspection personnel based on the analysis result.

2. The station inspection method based on artificial intelligence according to claim 1, characterized in that, According to the fault prompt and special prompt of the station operation and maintenance system, a first type of inspection task and a second type of inspection task are generated, comprising: The fault points corresponding to the fault prompts of the station operation and maintenance system are determined, and the monitoring video and / or monitoring image of the fault points are obtained; The monitoring video and / or monitoring image are subjected to fault discrimination, and if it is determined that there is a fault, an emergency response task is generated, and the emergency response task is determined as the first type of inspection task; The prompt content corresponding to the special prompt of the station operation and maintenance system is obtained, and the special prompt at least includes one of the flood season risk prompt, the air conditioning season risk prompt and the station passenger flow risk prompt; According to the prompt content, a planned inspection task and / or a cross-checking task are generated, and the planned inspection task and / or the cross-checking task are determined as the second type of inspection task. 3.The station inspection method based on artificial intelligence according to claim 1, characterized in that, The employee portrait is generated according to the execution of the historical inspection task by the inspection personnel, comprising: For any inspection personnel to be generated employee portrait, the historical inspection tasks completed by the inspection personnel and the registration information of the inspection personnel are obtained, and the registration information at least includes an inspection certificate; According to the task content of the historical inspection task, the inspection object of the historical inspection task is extracted, and the appearance frequency of the inspection object is calculated; The inspection object with an appearance frequency higher than a preset frequency threshold and the inspection certificate are matched; If the two are matched successfully, the matched inspection object is determined as the first label of the inspection personnel, otherwise, the inspection certificate is determined as the first label of the inspection personnel; The inspection score of the historical inspection task is obtained, and the task content corresponding to the historical inspection task with a score higher than a preset score threshold is clustered; According to the clustering result, the second label of the inspection personnel is generated; Based on the first label and the second label, the employee portrait of the inspection personnel is generated.

4. The station inspection method based on artificial intelligence according to claim 3, characterized in that, The first type of inspection task is directedly issued through the employee portrait, comprising: The task content and the inspection object of the first type of inspection task are determined; match the first label of the staff portrait with the inspection object to obtain a first matching degree, and match the second label with the task content to obtain a second matching degree; determine the to-be-assigned inspection personnel based on a maximum value in a weighted summation result of the first matching degree and the second matching degree; directly assign the first type of inspection task to the to-be-assigned inspection personnel.

5. The station inspection method based on artificial intelligence according to claim 4, characterized in that, After the first type of inspection task is directly assigned through the staff portrait, the method further includes: receive an execution time feedback returned by the to-be-assigned inspection personnel after receiving the first type of inspection task; if the execution time feedback is instant execution, the direct assignment of the first type of inspection task is completed; if the execution time feedback is waiting execution, it is determined whether a waiting time length of the waiting execution is greater than a preset time threshold; if yes, the first type of inspection task is re-directed until the execution time feedback is instant execution or the waiting time length of the waiting execution is less than the preset time threshold.

6. The station inspection method based on artificial intelligence according to claim 1, characterized in that, For the second type of inspection task, the assignment is performed through the inspection personnel's order grabbing operation, including: obtain a daily order grabbing average amount of the inspection personnel in a historical time period and a daily order grabbed amount of the inspection personnel; when the daily order grabbed amount is within a ±15% range of the daily order grabbing average amount, the order grabbing operation of the inspection personnel is rejected.

7. The station inspection method based on artificial intelligence according to claim 6, characterized in that, If the daily order grabbed amount is not within the ±15% range of the daily order grabbing average amount, the method further includes: determine whether the inspection personnel has a first type of inspection task that is directly assigned; if yes, it is determined whether a latest execution time point of the second type of inspection task corresponding to the order grabbing operation conflicts with a planned execution time period of the first type of inspection task; if yes, the order grabbing operation of the inspection personnel is rejected. 8.The station inspection method based on artificial intelligence according to claim 1, wherein, The inspection result further includes an inspection conclusion, which is a conclusion given by the inspection personnel on whether there is an abnormality for the first type of inspection task and / or the second type of inspection task. When the inspection conclusion is that there is an abnormality, the method further includes: determining that the inspection video meets the inspection compliance mechanism; identifying a preset inspection action of the inspection personnel in the inspection video, and determining an abnormal factor in the station based on the preset inspection action. The preset inspection action is a specified action or an indicated action issued by the inspection personnel when there is an abnormality in the first type of inspection task and / or the second type of inspection task. 9.The station inspection method based on artificial intelligence according to claim 8, characterized in that, When the inspection conclusion is that there is no abnormality, the method further includes: determining that the inspection video meets the inspection compliance mechanism; performing abnormal factor identification and maintenance operation identification on the inspection video according to the task content of the first type of inspection task and / or the second type of inspection task; if the abnormal factor is not identified and / or the maintenance operation is identified, adjusting a fault prompt mechanism of the station operation and maintenance system, and / or adjusting a generation period of a planned inspection task and / or a cross-checking task; if the abnormal factor is identified and the maintenance operation is not identified, generating a response mechanism for the abnormal factor, and adjusting a staff portrait and / or a daily order grabbing amount range of the inspection personnel who gives the inspection conclusion.

10. An artificial intelligence-based station inspection system, characterized by, The system includes: A task generation sub-agent generates a first type of inspection task and a second type of inspection task according to fault prompts and special prompts of a station operation and maintenance system; A task management sub-agent obtains an employee profile of an inspection personnel for the first type of inspection task, and directs the first type of inspection task to the inspection personnel through the employee profile, wherein the employee profile is generated according to execution of historical inspection tasks by the inspection personnel; The task management sub-agent instantaneously publishes the second type of inspection task after generation, and issues the second type of inspection task through a grab operation of the inspection personnel; An inspection result processing sub-agent receives inspection results returned by the inspection personnel for the first type of inspection task and / or the second type of inspection task, wherein the inspection results at least include an inspection video; The inspection result processing sub-agent performs inspection operation analysis and / or device state analysis on the inspection video based on task content of the first type of inspection task and / or the second type of inspection task, and determines whether the analysis results meet an inspection compliance mechanism; If yes, the inspection result processing sub-agent identifies abnormal factors in the station through the inspection video, and generates a response mechanism; If no, the task management sub-agent issues a re-inspection task to the inspection personnel based on the analysis results.

Images