Patrol method and system

Abstract

The invention belongs to the field of inspection tour, and particularly relates to an inspection tour method and system. Comprising the steps of judging whether a manual attention label is preset at an inspection point; if the manual attention label is preset at the inspection point position, determining an inspection result of the inspection point position through manual auditing and filing the inspection result; otherwise, the inspection result of the inspection point location is identified through a corresponding algorithm, and whether the inspection point location is preset with a label which does not need manual auditing or not is judged; the label does not need to be manually audited and determined according to whether the number of continuous errors of an inspection result recognized by an algorithm corresponding to an inspection point location within a set operation duration exceeds a set upper limit number of times or not; if a label which does not need to be checked manually is preset at the inspection point position, automatically confirming the inspection result to obtain the inspection result after automatic confirmation and filing the inspection result; and otherwise, manually checking the patrol result obtained by the algorithm identification corresponding to the patrol point location, so as to obtain the patrol result after manual checking, and archiving the patrol result.

Description

Technical Field

[0001] The present invention belongs to the field of patrol inspections, and particularly relates to a patrol method and system. Background Art

[0002] As Figure 1 shown, a patrol inspection system is a management system integrating a photographing function and a patrol inspection process. Such a system usually combines modern technologies and devices, such as smart phones, drones, and inspection robots, as well as corresponding software platforms to achieve efficient, accurate, and real-time inspection work. The system can analyze the uploaded photo or video data, identify potential problems, and generate inspection reports. This helps managers understand the operating status of equipment and inspection situations, and make decisions and adjustments in a timely manner. The system can also preset inspection plans and inspection items, and automatically plan and allocate inspection tasks according to the inspection plans. This ensures the timely execution of inspection tasks and effectively avoids the occurrence of missed inspections or late inspections.

[0003] Currently, in practical applications of the patrol inspection system, there are often situations where the camera captures unclear images due to on-site light factors, which cannot meet the requirements of algorithm recognition; or the algorithm cannot correctly identify due to factors such as blurred or reflective instrument dials; and other on-site environmental factors result in the algorithm ultimately being unable to correctly identify (that is, the accuracy rate of the recognition algorithm is relatively low); and the situation of incorrect identification includes two cases: unable to identify and incorrect identification results. Therefore, after algorithm recognition, it is necessary to manually review the recognition results obtained for each patrol point after algorithm recognition, so as to ensure the accuracy rate of the final patrol results to a certain extent.

[0004] The patrol system usually has a large number of patrol points. Currently, for all patrol points, regardless of whether they are correctly identified during the algorithm recognition process, manual review is required (that is, corresponding manual review of the inspection results is required after each inspection). Repeated manual review of some patrol points that are correctly identified by the algorithm or have almost no problems in historical reviews seriously increases the workload of the operation and maintenance personnel of the patrol system. Summary of the Invention

[0005] The purpose of the present invention is to provide a patrol method and system, which are used to solve the problem that in the prior art, the patrol inspection system conducts manual review on all patrol points in various situations, seriously increasing the workload of the operation and maintenance personnel of the patrol system.

[0006] To achieve the above purpose, the present invention provides a patrol method, which includes: determining whether a manual attention label is preset for a patrol point; If the patrol point is preset with a manual attention label, the patrol result of the patrol point is determined through manual review and archived; otherwise, the patrol result of the patrol point is obtained through the corresponding algorithm recognition, and it is determined whether the patrol point is preset with a label that does not require manual review; the label that does not require manual review is determined based on whether the number of consecutive errors in the patrol result obtained by the algorithm recognition corresponding to the patrol point within the set running time exceeds the set upper limit; if the patrol point is preset with a label that does not require manual review, the patrol result is automatically confirmed to obtain the automatically confirmed patrol result and archived; otherwise, the patrol result obtained by the algorithm recognition corresponding to the patrol point is manually reviewed to obtain the manually reviewed patrol result and archived.

[0007] Beneficial effects: The purpose of the present invention is to provide a new patrol method, which actually classifies patrol points by using preset labels, and then adopts corresponding patrol methods for different types of patrol points to realize patrol of all points.

[0008] First, determine whether each patrol point is a patrol point with a preset manual attention label; the processing method for each patrol point with a preset manual attention label is: there is no need to identify it through an algorithm, but directly enter the manual review link for processing. Such patrol points of manual attention may be points that the algorithm cannot identify or have been identified incorrectly. When using the algorithm to identify these points, it usually takes a long time and cannot give a correct identification result. This method distinguishes such points through the manual attention label, so as to save the step of algorithm identification during the patrol process, and directly perform manual identification; it can not only guarantee the accuracy of the identification results to a certain extent, but also reduce the duration of the patrol process, thereby saving some workload.

[0009] Secondly, the remaining inspection points are then divided into two categories by means of preset tags according to whether they are "points that do not require manual review". Whether manual review is required is judged based on historical recognition situations, that is, within the set operation duration, it is judged whether a point requires manual review according to whether the number of consecutive errors in the inspection results obtained by the algorithm corresponding to the inspection point within a certain period of historical time exceeds the set upper limit. For such points with a preset tag of not requiring manual review, since there are few or almost no recognition errors during algorithm recognition, they are recognized as points with absolutely correct inspection results and directly archived. In the prior art, a large number of inspection results that are not easily recognized incorrectly during algorithm recognition will increase a lot of unnecessary workload due to the processing measure of "regardless of whether the point is a point with high quality in algorithm recognition, repeated manual review is carried out on it"; by adopting the method of setting points that do not require manual review, the situation of repeated manual review of inspection results that are consistent after algorithm recognition and manual review can be avoided, thus saving a part of the workload.

[0010] Finally, for each point without a preset tag of not requiring manual review, it can be known from this tag whether the number of consecutive errors in the inspection results obtained by the algorithm corresponding to the inspection point within the set operation duration exceeds the set upper limit, that is, it can be reflected that such points often have recognition errors during algorithm recognition. Therefore, manual recognition is still required for them to ensure the correctness of the final inspection results of such points.

[0011] In summary, by reasonably classifying and inspecting all inspection points through this method, it is possible to avoid repeated manual review of correctly recognized points while ensuring a high accuracy rate of the inspection results corresponding to each inspection point, saving a large amount of the inspection workload to a great extent, and thus reducing the workload of operation and maintenance personnel.

[0012] Furthermore, the inspection points with a preset manual attention tag include: inspection points where the proportion of incorrect inspection results in the historical data that meet the inspection results obtained by the corresponding algorithm is greater than or equal to the set threshold and inspection points with inaccurate algorithm recognition set manually according to actual needs.

[0013] Furthermore, the method of determining the inspection results of inspection points through manual review includes: combining the real-time video corresponding to the inspection points with a preset manual attention tag, observing the situation of the inspection points manually, and determining the inspection results of the inspection points according to the observation results.

[0014] Furthermore, the method of automatically confirming the inspection results includes: directly marking the inspection results as the automatically confirmed state to obtain the automatically confirmed inspection results.

[0015] Furthermore, the method for manually reviewing the inspection results identified by the algorithm corresponding to the inspection point includes: manually checking the picture corresponding to the inspection point to obtain the inspection value corresponding to the manual review. If the inspection value in the inspection result identified by the algorithm corresponding to the inspection point is inconsistent with the inspection value corresponding to the manual review, then the inspection value in the inspection result identified by the algorithm is corrected according to the inspection value corresponding to the manual review.

[0016] The present invention also provides an inspection system, which includes a processor for executing a computer program to implement the steps of the above-mentioned inspection method.

[0017] This inspection system can achieve the same beneficial effects as the above-mentioned inspection method. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the system architecture of the inspection and patrol system in the background technology of the present invention; Figure 2 It is a schematic diagram of the flow of the inspection method in the embodiment of the inspection method of the present invention. Detailed Embodiments

[0019] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments.

[0020] Embodiment of the Inspection Method This embodiment provides a technical solution for an inspection method. The method is to preset labels for inspection points, divide the inspection points into several categories, and then perform reasonable planning and adaptive inspection processing on different types of points according to the preset labels. The inspection points classified according to the labels then execute their set processes, in which many unnecessary processing steps (which may be algorithm recognition operations or manual review operations) may be omitted. In summary, inspecting the inspection points after the planning process greatly reduces the workload of the maintenance personnel of the inspection system.

[0021] As Figure 2 shown, this embodiment specifically includes: determining whether the inspection point is preset with a manual attention label; the manual attention label is a label added to the inspection point model in advance; among them, the inspection points preset with the manual attention label include: 1) Inspection points where the proportion of incorrect inspection results in the historical data that meet the inspection results identified by the corresponding algorithm is greater than or equal to the set threshold in all the inspection results in the historical data; Taking a certain inspection point as an example, for instance, if the set threshold is 0.5, and the proportion of incorrect inspection results in the historical data of the inspection results identified by the algorithm corresponding to a certain inspection point is 0.6 in the historical data of all inspection results, that is, 60% of the historical inspection results are incorrect, and this proportion value is greater than the set threshold of 0.5, then an artificial attention label needs to be set for this point to focus on it.

[0022] 2) Inspection points with inaccurate algorithm recognition set manually according to actual requirements; Taking a certain inspection point as an example, for example, if the recognition result of a certain inspection point after algorithm recognition is incorrect, then an artificial attention label can be added to this inspection point according to the requirements to focus on it. In other embodiments, other conditions can also be used to preset artificial attention labels for inspection points.

[0023] In this embodiment, if an artificial attention label is preset for an inspection point, the inspection result of the inspection point is determined through manual review and archived; the methods for determining the inspection result of the inspection point through manual review include: combining the real-time video corresponding to the inspection point with the preset artificial attention label (that is, during the inspection process, taking pictures and capturing images through inspection devices such as cameras, robots, or drones to obtain real-time video), observing the situation of the inspection point manually, and then determining the inspection result of the inspection point according to the observation result; and when conducting manual review, the inspection points with this artificial attention label will be preferentially displayed, and each will be combined with the real-time video for manual confirmation, and the observation result will be entered into the inspection result to complete the archiving of the inspection result. The entire process does not require defect recognition and meter recognition through algorithms. It supports the business requirements of manual remote inspection and helps improve the performance of algorithm analysis.

[0024] Otherwise (that is, if no artificial attention label is set for the inspection point), the inspection result of the inspection point is obtained through the corresponding algorithm recognition, and it is judged whether the inspection point has a label indicating no need for manual review; the label indicating no need for manual review is determined according to whether the number of consecutive incorrect inspection results obtained by the algorithm recognition corresponding to the inspection point within the set operation duration exceeds the set upper limit number.

[0025] The requirement for the recognition accuracy of the algorithm is preset to be higher than 80%. Taking the set running duration of one week and the set upper limit of 3 times as an example; for the inspection point A, if the number of errors in the inspection results obtained by the algorithm corresponding to the inspection point A within one week is greater than 3 times, exceeding the set upper limit of 3 times, it indicates that the recognition result of the algorithm for the inspection point A is inaccurate and the risk of error is relatively high. Therefore, it is determined as a problem point. However, since the algorithm recognition cannot guarantee sufficient accuracy, manual meticulous review of this point is required. Therefore, the label of "no need for manual review" cannot be added to it. By this method, the inspection process corresponding to this part of the inspection points with the label of "no need for manual review" can be reduced, thus saving a part of the workload.

[0026] Similarly, taking the set running duration of one week and the set upper limit of 3 times as an example; for the inspection point B, if the number of errors in the inspection results obtained by the algorithm corresponding to the inspection point B within one week is 1 time, not exceeding the set upper limit of 3 times, that is, the recognition result of the algorithm for the inspection point B is relatively accurate and the risk of error is relatively low. In this case, the label of "no need for manual review" can be added to it.

[0027] If the inspection point is preset with the label of "no need for manual review", the inspection result is automatically confirmed to obtain the automatically confirmed inspection result and archived; the methods for automatically confirming the inspection result include: directly marking the inspection result as the automatically confirmed status to obtain the automatically confirmed inspection result. Otherwise, the inspection result obtained by the algorithm corresponding to this inspection point is manually reviewed to obtain the manually reviewed inspection result and archived.

[0028] Specifically, taking the above-mentioned inspection point B as an example, this point B is preset with the label of "no need for manual review"; the inspection result obtained for this point B is automatically confirmed; in this embodiment, the method for automatically confirming the inspection result includes: directly setting the automatically confirmed status label for the inspection result; when the inspection of the inspection point B enters the next cycle, the generated inspection result is automatically confirmed according to this automatically confirmed status label to obtain the automatically confirmed inspection result, and the result archiving is completed. In other embodiments, it is also possible to first set the automatically confirmed status label for result archiving, and then extract the inspection result after archiving for confirmation (that is, there is no need to confirm immediately currently, and it is also possible to make the label first and then confirm later).

[0029] In summary, since no other processing is required for this type of inspection results with the automatically confirmed status label set, subsequent automatic confirmation can be directly carried out according to this label without manual judgment and confirmation. Therefore, this setting can save a part of the workload.

[0030] Taking the inspection point C as an example, no tag for bypassing manual review is preset at this point C; then, it is necessary to conduct a manual review of the inspection results obtained by algorithm recognition corresponding to this inspection point C; in this embodiment, the method for conducting a manual review of the inspection results obtained by algorithm recognition corresponding to this inspection point includes: manually viewing the pictures corresponding to the inspection point (i.e., photos, videos, screenshots, etc. taken by inspection devices such as cameras, robots, or drones), so as to obtain the inspection results after manual review. In this embodiment, specifically, the inspection results after manual review are compared with the inspection values included in the inspection results obtained by algorithm recognition corresponding to the inspection point, and the inspection values in the inspection results obtained by algorithm recognition are corrected according to the inspection values corresponding to the manual review to obtain the inspection results after manual review, thereby realizing the manual review. Taking the process of conducting a manual review of the inspection point C as an example, if the inspection value in the inspection results obtained by algorithm recognition corresponding to this inspection point C is inconsistent with the inspection value corresponding to the manual review result obtained by manually viewing the pictures corresponding to this inspection point C, then the inspection value in the inspection results obtained by algorithm recognition is corrected according to the inspection value corresponding to the manual review, and the inspection results including this inspection value are archived.

[0031] Embodiment of the inspection system This embodiment provides an inspection system, which includes a processor. The processor stores executable program instructions for implementing the inspection method in the inspection method embodiment as described above.

[0032] Since the specific working mode and working principle of the inspection system in this embodiment have been described in detail in the above inspection method embodiment, they will not be elaborated here.

[0033] It should be understood that the above specific embodiments of the present invention are only used for exemplary illustration or explanation of the principle of the present invention, and do not constitute a limitation to the present invention.

Claims

1. A patrol method, characterized in that: include: Determine whether the patrol point has a preset manual attention label; If the patrol point is preset with a manual attention label, the patrol result of the patrol point is determined through manual review and archived; otherwise, the patrol result of the patrol point is obtained through the corresponding algorithm recognition, and it is determined whether the patrol point is preset with a label that does not require manual review; the label that does not require manual review is determined based on whether the number of consecutive errors in the patrol result obtained by the algorithm recognition corresponding to the patrol point within the set running time exceeds the set upper limit; if the patrol point is preset with a label that does not require manual review, the patrol result is automatically confirmed to obtain the automatically confirmed patrol result and archived; otherwise, the patrol result obtained by the algorithm recognition corresponding to the patrol point is manually reviewed to obtain the manually reviewed patrol result and archived.

2. The patrol method according to claim 1, characterized in that: The inspection points preset with manual attention labels include: inspection points where the proportion of erroneous inspection results in the historical data of the inspection results obtained by the corresponding algorithm recognition is greater than or equal to the set threshold, and inspection points where the algorithm manually set according to actual needs identifies inaccurate inspection points.

3. The patrol method according to claim 1 or 2, characterized in that: The method of determining the inspection results of the inspection points through manual review includes: combining the real-time video corresponding to the inspection points with preset manual attention tags, manually observing the conditions of the inspection points, and determining the inspection results of the inspection points based on the observation results.

4. The patrol method according to claim 1 or 2, characterized in that: The method of automatically confirming the inspection result includes: directly marking the inspection result as an automatic confirmation state, and obtaining the inspection result after automatic confirmation.

5. The patrol method according to claim 1 or 2, characterized in that: The method of manually reviewing the inspection results obtained by the algorithm recognition corresponding to the inspection point includes: manually checking the pictures corresponding to the inspection points to obtain the inspection values ​​corresponding to the manual review; if the inspection values ​​in the inspection results obtained by the algorithm recognition corresponding to the inspection point are inconsistent with the inspection values ​​corresponding to the manual review, then correcting the inspection values ​​in the inspection results obtained by the algorithm recognition according to the inspection values ​​corresponding to the manual review.

6. A patrol system, comprising a processor, characterized in that: The processor is used to execute the computer program to implement the steps of the patrol method described in any one of claims 1-5.

Images