Tower crane maintenance detection method and system based on image recognition

Abstract

The invention relates to the technical field of tower crane maintenance and detection, in particular to a tower crane maintenance and detection method and system based on image recognition, and aims to solve the problems of high omission ratio and high false detection rate when a tower crane is maintained and detected through an image recognition technology. The invention provides a tower crane maintenance detection method based on image recognition. The method comprises the following steps: determining a correlation coefficient between correlation targets according to the working age limit of a tower crane; determining a control standard of the detection target according to the historical maintenance data, and identifying the image data according to the control standard; obtaining an abnormal target according to an identification result, and recording an associated target of the abnormal target as a key target; the development rate of the abnormal target is acquired, and the operation state of the key target is pre-judged according to the development rate and the correlation coefficient; rechecking the key target according to a detection deviation result; and updating the historical maintenance data according to the recheck result and the pre-judgment state.

Description

Technical Field

[0001] This invention relates to the field of tower crane maintenance and inspection technology, and in particular to a tower crane maintenance and inspection method and system based on image recognition. Background Technology

[0002] Tower cranes, as core heavy equipment in the engineering construction field, directly determine construction efficiency and the safety of personnel and property through their structural stability and operational safety. Regular and accurate maintenance and inspection, as well as reliable alarms, are key safeguards against major accidents such as collapses and falling loads. Non-contact tower crane inspection solutions based on image recognition are gradually replacing traditional manual climbing inspections, becoming the mainstream technology due to their high efficiency and safety advantages. However, their alarm accuracy is still limited by the core design flaw of "isolated component inspection," making it difficult to meet the core engineering requirements of "accurate early warning, fewer false alarms and fewer missed alarms." Existing visual inspection technologies mostly focus on the defect identification of single components (such as detecting loose bolts, weld cracks, and broken wires in steel cables), failing to fully explore the structural characteristics of tower cranes, including "mechanical transmission correlation," "functional linkage correlation," and "visual complementary correlation." On the one hand, core components of tower cranes exhibit strong correlations. For example, loose bolts in standard sections can cause flange deformation, leading to stress concentration and cracking in welds. Existing technologies, when detecting welds in isolation, are prone to missing such "correlated derivative defects," or may only detect loose bolts without recognizing the resulting risk of tower verticality deviation, resulting in "missed alarms." On the other hand, false features in complex operating environments (such as crack-like textures formed by shadow occlusion or sudden changes in lighting) are easily misjudged as real defects by isolated image recognition methods. However, their falsehood can be verified by combining the status of related components (such as the absence of loose bolts and structural deformation around a weld that is judged as a "crack"). Existing technologies lack such correlation verification mechanisms, leading to frequent "false alarms." Therefore, how to overcome the limitations of "isolated detection" and optimize the defect identification and alarm logic of image recognition by mining and utilizing the correlation between tower crane structures, so as to improve alarm accuracy and balance "safety coverage" and "operation and maintenance efficiency", has become the core bottleneck for the practical application of image recognition-based tower crane maintenance and inspection technology. Summary of the Invention

[0003] The problem solved by this invention is the high rate of missed detections and high rate of false detections when using image recognition technology to perform maintenance and inspection of tower cranes.

[0004] To address the aforementioned problems, this invention provides a tower crane maintenance and inspection method based on image recognition. The method includes: acquiring the tower crane's maintenance and inspection task; obtaining inspection targets based on the maintenance and inspection task; determining correlation standards based on historical maintenance data; obtaining associated targets based on the inspection targets and correlation standards; determining correlation coefficients between associated targets based on the tower crane's service life; determining control standards for the inspection targets based on historical maintenance data; acquiring image data of the inspection targets; recognizing the image data according to the control standards; obtaining recognition results; identifying abnormal targets based on the recognition results; and marking associated targets of abnormal targets as key targets; acquiring the development rate of abnormal targets; predicting the operating status of key targets based on the development rate and correlation coefficient; obtaining detection deviation results based on the identification results and prediction status of key targets; re-inspecting key targets based on the detection deviation results; and updating historical maintenance data based on the re-inspection results and prediction status.

[0005] Compared with existing technologies, the technical effects achieved by this solution are as follows: By determining the correlation standards between inspection targets using historical maintenance data, key targets associated with abnormal targets are identified. This fully utilizes the mechanical transmission and functional linkages between tower crane structures, allowing the inspection process to go beyond single components and comprehensively consider the mutual influence between them. The correlation coefficients between related targets are determined based on the tower crane's service life, taking into account the performance changes of each component over time during operation, making the determination of correlation coefficients more reasonable and further improving inspection accuracy. After identifying abnormal targets, not only are their related targets highlighted, but the development rate of the abnormal targets is also obtained, combined with the correlation coefficients, to predict the operating status of key targets, effectively preventing the expansion of potential risks and providing strong support for timely maintenance measures. By comparing the identification results and predicted status of key targets, inspection deviation results are obtained, and key targets are re-inspected based on these deviation results. This re-inspection mechanism ensures the reliability of the inspection results. Simultaneously, historical maintenance data is updated based on the re-inspection results and predicted status, allowing for continuous optimization of historical maintenance data and providing more accurate data support for subsequent inspections.

[0006] In one embodiment of the present invention, the association criteria are determined based on historical maintenance data, and the associated targets are obtained based on the detection targets and the association criteria. Specifically, this includes: obtaining historical maintenance data from a data management platform; obtaining the location information, co-occurrence frequency, and transmission time of each detection target based on the historical maintenance data; determining the association criteria between each detection target based on the location information, co-occurrence frequency, and transmission time; and recording the detection targets that meet the association criteria as associated targets.

[0007] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: By deeply analyzing the location information, co-occurrence frequency, and transmission timeliness of the detection targets, the correlation standards between each detection target can be accurately determined. It not only considers the physical positional relationship between components, but also combines their co-occurrence in actual maintenance and the timeliness of information transmission, thereby ensuring the practicality of the correlation standards. Detection targets that meet these correlation standards are marked as associated targets, providing a solid foundation for subsequent key detection and prediction, and helping to more comprehensively grasp the mutual influence and potential risks between the various components of the tower crane.

[0008] In one embodiment of the present invention, the association criteria between each detection target are determined based on location information, co-occurrence frequency, and transmission time. Specifically, this includes: recording detection targets with co-occurrence frequency greater than a frequency threshold or transmission time less than a time threshold as a target group to be associated; obtaining location information between detection targets within the target group to be associated, and obtaining the location distance and location relationship between detection targets based on the location information; and filtering the target group to be associated based on the location distance and location relationship to obtain associated targets.

[0009] Compared with existing technologies, the technical effects achieved by this solution are as follows: By setting thresholds for co-occurrence frequency and transmission time, potential related detection targets can be preliminarily screened to form a group of targets to be associated, effectively narrowing the scope of subsequent analysis and improving processing efficiency. By obtaining detailed positional information between detection targets within the group of targets to be associated and calculating positional distances and relationships accordingly, the spatial layout and mutual influence between components can be more accurately depicted. Based on the screening of positional distances and relationships, truly related detection targets can be accurately identified, providing strong support for subsequent maintenance and inspection, and helping to more accurately assess the overall condition and potential risks of the tower crane.

[0010] In one embodiment of the present invention, determining the correlation coefficient between related targets based on the working years of the tower crane specifically includes: determining the working years range of the tower crane and dividing the working years range into multiple working years intervals; calculating the co-occurrence frequency and transmission time of the related targets in each working years interval based on the frequency of their occurrence in each working years interval; and determining the correlation coefficient between the related targets based on the co-occurrence frequency and transmission time.

[0011] Compared with existing technologies, the technical effects achieved by adopting this solution are as follows: By meticulously dividing the working years of the tower crane and analyzing the co-occurrence frequency and transmission timeliness of related targets in each interval, the dynamic changes in the correlation between components under different working years can be captured more accurately. This not only considers the characteristics of the performance degradation of each component over time during tower crane use, but also combines the timeliness of the correlation strength between components reflected in actual maintenance data, thereby ensuring the rationality of the correlation coefficient determination. Based on the correlation coefficient, the degree of mutual influence between related targets can be assessed more accurately, providing more reliable data support for the formulation of subsequent maintenance and inspection strategies, and helping to further improve the accuracy and efficiency of tower crane maintenance and inspection.

[0012] In one embodiment of the present invention, the control standards for the detection targets are determined based on historical maintenance data, image data of the detection targets are acquired, and the image data is identified according to the control standards to obtain identification results. Specifically, this includes: obtaining risk characteristics of each detection target based on historical maintenance data; determining control standards for each detection target based on the risk characteristics; identifying image data according to the control standards to obtain risk probabilities; identifying abnormal targets when the risk probability is greater than a first risk threshold; and identifying image data according to a preset method when the risk probability is less than the first risk threshold but greater than a second risk threshold to obtain identification results.

[0013] Compared with existing technologies, the technical effects achieved by this solution are as follows: By deeply analyzing historical maintenance data, the risk characteristics of each detection target can be accurately extracted, and scientific and reasonable control standards can be formulated accordingly. This not only considers the inherent properties of the components but also their performance in historical maintenance, making the control standards more practical and targeted. Using these control standards to intelligently identify image data helps to accurately calculate the risk probability of each detection target, providing a reliable basis for subsequent anomaly judgment. For detection targets with risk probabilities between the first and second risk thresholds, further identification is performed through preset methods, ensuring the accuracy and comprehensiveness of the identification results.

[0014] In one embodiment of the present invention, the development rate of the abnormal target is obtained, and the operational status of the key target is predicted based on the development rate and the correlation coefficient to obtain the predicted status. Specifically, this includes: obtaining image data of the abnormal target at different time points, determining the development rate of the abnormal target based on the image data at different time points; obtaining the predicted status based on the development rate and the correlation coefficient; and determining whether the key target has an abnormal risk based on the predicted status and the status threshold.

[0015] Compared with existing technologies, the technical effects achieved by this solution are as follows: By acquiring image data of abnormal targets at different time points, it is possible to accurately track their development and changes, and then accurately calculate the development rate, providing key data support for subsequent predictions. This makes the prediction results more realistic and forward-looking. Combined with the correlation coefficient, it is possible to comprehensively consider the mutual influence between abnormal targets and their related targets, thereby more comprehensively assessing the operational status of key targets. By comparing the predicted status with the preset status threshold, it is possible to promptly identify abnormal risks existing in key targets, providing a strong basis for timely maintenance measures.

[0016] In one embodiment of the present invention, a detection deviation result is obtained based on the identification result and the predicted state of the key target. The key target is then re-examined based on the detection deviation result to obtain a re-examination result. Specifically, this includes: when there is an abnormal risk to the key target, comparing the identification result and the predicted state of the key target to obtain a similarity, and recording the similarity as the detection deviation result; when the detection deviation result is greater than the deviation threshold, the key target is re-examined to generate a re-examination result.

[0017] Compared with existing technologies, the technical effects achieved by this solution are as follows: By comparing the identification results of key targets with the predicted status and calculating the similarity, the degree of deviation between the initial detection and the prediction can be quantitatively assessed, providing a scientific basis for subsequent re-inspection decisions. When the detection deviation exceeds the preset threshold, the system automatically triggers the re-inspection process, effectively avoiding the risk of misjudgment that may be caused by a single detection result, ensuring the rigor of maintenance and inspection. The introduction of the re-inspection process not only improves the reliability of the detection results, but also further reduces the probability of missed detection and false detection through a secondary verification mechanism, providing dual protection for the safe operation of tower cranes.

[0018] In one embodiment of the present invention, updating historical maintenance data based on re-inspection results and predicted status specifically includes: when there is a difference between the re-inspection results and the predicted status, determining the type of difference and updating the historical maintenance data accordingly based on the type of difference; when there is no difference between the re-inspection results and the predicted status, updating the re-inspection results to the historical maintenance data.

[0019] Compared with existing technologies, the technical effects achieved by adopting this solution are as follows: By comparing the re-inspection results with the predicted status, the system can accurately identify the types of differences between the two and update the historical maintenance data accordingly. This not only ensures the timeliness and accuracy of historical data but also enables subsequent inspections to be conducted based on a more comprehensive data foundation, thereby improving the overall reliability and effectiveness of the inspection. When there is no difference between the re-inspection results and the predicted status, the system automatically incorporates the re-inspection results into the historical maintenance data, realizing the dynamic accumulation and optimization of data.

[0020] In one embodiment of the present invention, a tower crane maintenance and inspection system based on image recognition is also provided. The tower crane maintenance and inspection method based on image recognition described in the above embodiment is applied to the inspection system. The inspection system includes: an image acquisition module for acquiring image data; an image recognition module for recognizing the image data according to control standards to obtain recognition results; a data judgment module for obtaining predicted status and abnormal targets; and a data management module for storing and updating historical maintenance data. The inspection system has all the technical features of the above inspection method, which will not be described in detail here. Attached Figure Description

[0021] Figure 1 This is one of the flowcharts for a tower crane maintenance and inspection method based on image recognition according to the present invention; Figure 2 This is the second flowchart of a tower crane maintenance and inspection method based on image recognition according to the present invention; Figure 3 This is the third flowchart of a tower crane maintenance and inspection method based on image recognition according to the present invention; Figure 4 This is the fourth flowchart of a tower crane maintenance and inspection method based on image recognition according to the present invention; Figure 5 This is a schematic diagram of a tower crane maintenance and inspection system based on image recognition according to the present invention; Explanation of reference numerals in the attached figures: 100 - Detection system; 110 - Image acquisition module; 120 - Image recognition module; 130 - Data judgment module; 140 - Data management module. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Example 1 See Figure 1 In one specific embodiment, the present invention provides a tower crane maintenance and inspection method based on image recognition, the inspection method comprising: S100: Obtain the tower crane's maintenance and inspection tasks, obtain the inspection targets based on the maintenance and inspection tasks, determine the correlation standards based on historical maintenance data, and obtain the correlation targets based on the inspection targets and correlation standards; S200. Determine the correlation coefficient between related targets based on the working years of the tower crane; S300: Determine the control standards for the inspection targets based on historical maintenance data, acquire image data of the inspection targets, identify the image data according to the control standards, obtain the identification results, identify abnormal targets based on the identification results, and record the related targets of the abnormal targets as key targets; S400. Obtain the development rate of abnormal targets, and predict the operational status of key targets based on the development rate and correlation coefficient to obtain the predicted status. S500. Based on the identification results and predicted status of key targets, the detection deviation results are obtained. Based on the detection deviation results, the key targets are re-inspected to obtain the re-inspection results. S600: Update historical maintenance data based on re-inspection results and predicted status.

[0024] In step S100, tower crane maintenance and inspection tasks typically originate from daily inspection plans, sudden fault reports, or periodic maintenance requirements. Upon receiving a maintenance and inspection task, the system first parses the task content to identify the specific monitoring targets to be inspected. These targets may include key structural components, transmission systems, and electrical components of the tower crane. Historical maintenance data covers past maintenance records, fault phenomena, and replaced parts information. Typically, the correlation criteria between various inspection targets may involve the physical connections between components, functional dependencies, and fault propagation paths.

[0025] In step S200, generally speaking, as the service time increases, the various components of the tower crane will experience different degrees of wear and aging, causing the relationship between components to evolve from the design expectation or simple mechanical transmission to a damage-driven, complex or nonlinear risk network. Therefore, when determining the correlation coefficient between the related targets, the service life of the tower crane must be fully considered. For example, in a 2-year-old building tower crane, an inspection found that one bolt on a standard section was exposed with 3 threads (loose by 0.5mm), while the surrounding bolts were not loose, the flange was not deformed, and the weld was not cracked. The defect was limited to a single component and there was no related diffusion, verifying a weak correlation. On the other hand, in a 10-year-old tower crane, an inspection found that the outer gear ring of the slewing bearing had a local broken tooth, and at the same time, it was found that the tooth surface of the corresponding part of the pinion meshing with it was severely worn. After analysis, the two were found to be directly related, verifying a strong correlation.

[0026] In step S300, the control standards for different tower cranes and tower cranes in different working environments are generally different. Therefore, it is necessary to determine them through historical maintenance data. The historical maintenance data records the risk performance of each inspection target in different maintenance scenarios. These risk characteristics may include common fault types of components, the frequency of fault occurrence, and the severity of the consequences of the fault. Based on these risk characteristics, a set of scientific and reasonable control standards are tailored for each inspection target. The control standards not only clarify the parameter range of the components under normal conditions, but also set the judgment threshold for abnormal conditions. For example, for a tower crane that has been in operation for 3 years, attention should be paid when the crack length of the tower crane's weld is identified as greater than or equal to 3mm. For a tower crane that has been in operation for 10 years, attention should be paid when the crack length of the tower crane's weld is identified as greater than or equal to 2mm.

[0027] Target detection (such as YOLOv8, SSD, or Faster R-CNN) can detect and locate abnormal targets. Image segmentation (such as U-Net and Mask R-CNN) can perform pixel-level area and morphology analysis on image data, such as identifying the area ratio of rusted regions, the maximum length and direction of cracks, and the diameter of steel wire ropes after wear.

[0028] In step S400, image data can be collected by fixed cameras and non-fixed cameras (such as cameras mounted on drones). These cameras can capture real-time images of the tower crane from different angles and distances. After obtaining image data of the abnormal target at different time points, image processing techniques, such as edge detection and feature extraction, are used to accurately measure key parameters such as the shape and size of the abnormal target at different times, thereby determining its development rate.

[0029] In step S500, the identification results of key targets usually cover multi-dimensional information such as their risk characteristics and current operating status, while the predicted status is a prediction of the current risk characteristics and current operating status of key targets based on factors such as correlation and development rate. By comparing the two, the similarity between them can be calculated. This similarity is used as the detection deviation result. When the detection deviation result exceeds the preset deviation threshold, it means that there is a large deviation between the initial detection and the prediction, and there may be a risk of misjudgment or missed detection.

[0030] It should be noted that components that are more concealed can be re-inspected through manual testing.

[0031] In step S600, the re-inspection results may differ from the predicted state. When the re-inspection results differ from the predicted state, this difference may stem from various factors, such as environmental interference during the inspection process, special wear conditions of components, or errors in data recording. The types of differences may include data recording errors, abnormal wear of components, and the influence of environmental factors. According to different types of differences, the historical maintenance data is updated accordingly. For example, if it is a data recording error, the system will correct the erroneous data recording; if it is abnormal wear of components, the system will record this abnormal wear condition and update the wear abnormality type of the relevant components.

[0032] By establishing correlation standards between inspection targets using historical maintenance data, key targets associated with abnormal targets are identified. This fully leverages the mechanical transmission and functional linkages between tower crane structures, allowing the inspection process to go beyond single components and comprehensively consider the mutual influence between them. Correlation coefficients between related targets are determined based on the tower crane's service life, taking into account the performance changes of each component over time during operation, making the determination of correlation coefficients more reasonable and further improving inspection accuracy. After identifying abnormal targets, not only are their related targets highlighted, but the development rate of the abnormal targets is also obtained. Combined with the correlation coefficients, the operating status of key targets is predicted, effectively preventing the expansion of potential risks and providing strong support for timely maintenance measures. By comparing the identification results and predicted status of key targets, inspection deviation results are obtained, and key targets are re-inspected based on the deviation results. This re-inspection mechanism ensures the reliability of the inspection results. Simultaneously, historical maintenance data is updated based on the re-inspection results and predicted status, allowing for continuous optimization of historical maintenance data and providing more accurate data support for subsequent inspections.

[0033] Example 2 See Figure 2 In one specific embodiment, the association criteria are determined based on historical maintenance data, and the association target is obtained based on the detection target and the association criteria, specifically including: S110. Obtain historical maintenance data from the data management platform, and obtain the location information, co-occurrence frequency and transmission time of each detection target based on the historical maintenance data; S120. Determine the correlation criteria between each detection target based on location information, co-occurrence frequency, and transmission time. S130. Record the detection targets that meet the correlation criteria as correlation targets.

[0034] In step S110, the data management platform includes a data management system built by the tower crane manufacturer, a database built by a third-party operation and maintenance service company, and a shared information platform built under the leadership of the industry regulatory department. It gathers a massive amount of historical maintenance data, covering detailed information such as all maintenance records, fault phenomena, and component replacements of the tower crane from its commissioning to the present moment.

[0035] In step S120, location information is one of the important factors to consider in determining the association criteria. Generally speaking, there is a closer association between detection targets that are physically close, because they are more likely to influence and interact with each other during the operation of the tower crane. For example, key components located in the same structural layer are subject to similar mechanical and working environments. Once one component fails, it is likely to quickly affect the adjacent components. The co-occurrence frequency reflects the degree of association between detection targets from another perspective. The higher the co-occurrence frequency, the more likely the two detection targets will have problems at the same time when a fault or abnormality occurs, and the stronger the association between them. The transmission time further reveals the dynamic characteristics of fault propagation. The shorter the transmission time, the faster the fault propagates from one detection target to another, and the closer the association between the two.

[0036] By deeply analyzing the location information, co-occurrence frequency, and transmission timeliness of the detection targets, the correlation standards between each detection target can be accurately determined. This not only considers the physical positional relationship between components but also their co-occurrence in actual maintenance and the timeliness of information transmission, thereby ensuring the practicality of the correlation standards. Detection targets that meet these correlation standards are marked as associated targets, providing a solid foundation for subsequent key inspections and predictions. This helps to more comprehensively grasp the mutual influence and potential risks between the various components of the tower crane.

[0037] Example 3 In a specific embodiment, the association criteria between various detection targets are determined based on location information, co-occurrence frequency, and transmission time, specifically including: S121. Record the detection targets whose co-occurrence frequency is greater than the frequency threshold or whose transmission time is less than the time threshold as the target group to be associated. S122. Obtain the position information between the detected targets within the target group to be associated, and obtain the positional distance and positional relationship between the detected targets based on the position information; S123. Filter the target group to be associated by location distance and location relationship to obtain associated targets.

[0038] In step S121, under normal circumstances, by setting frequency thresholds and time thresholds, potential correlations in detection target combinations can be quickly screened out. Frequency thresholds and time thresholds can be obtained by statistical analysis of historical maintenance data, or they can be set based on industry experience and expert advice. Different tower crane types, working environments, and service life factors may affect the reasonable values ​​of frequency thresholds and time thresholds. For example, for tower cranes with a short service life and relatively stable working environments, the propagation of faults between their components is usually slower, and the co-occurrence frequency is relatively low. Therefore, a lower frequency threshold and a higher time threshold can be set to more accurately capture potential correlations. However, for tower cranes with a long service life or complex working environments, due to increased component wear and increased external interference, fault propagation may be faster, and the co-occurrence frequency will also increase accordingly. In this case, the frequency threshold should be appropriately increased and the time threshold decreased.

[0039] In step S122, the specific coordinates or relative positional relationships of the detection targets in the tower crane structure can be obtained through tower crane design drawings, on-site measurements, or 3D scanning. Then, the positional distance between them can be calculated. The distance directly reflects the density of the detection targets in physical space, while the positional relationship reveals their relative layout and interaction. For example, some detection targets may be located on the same structural layer or adjacent support points. This information is crucial for understanding the mechanical transmission and fault propagation paths between them.

[0040] In step S123, generally speaking, there is a stronger correlation between detection targets that are close in location and closely related in location, because they are more likely to be affected by the same or similar factors during tower crane operation. For example, two key components located on the same beam are subject to the same vibration and load. Once one component fails, it is likely to quickly affect the other component. Therefore, in the screening process, these detection targets that are close in location and closely related in location should be given priority and identified as the final associated targets.

[0041] By setting thresholds for co-occurrence frequency and transmission time, potentially related detection targets can be preliminarily screened, forming a target group to be associated. This effectively narrows the scope of subsequent analysis and improves processing efficiency. By obtaining detailed location information between detection targets within the target group to be associated, and calculating the location distance and relationship accordingly, the spatial layout and mutual influence between components can be more accurately depicted. Based on the screening of location distance and relationship, truly related detection targets can be accurately identified, providing strong support for subsequent maintenance and inspection, and helping to more accurately assess the overall condition and potential risks of the tower crane.

[0042] Example 4 See Figure 3 In one specific embodiment, the correlation coefficient between related targets is determined based on the service life of the tower crane, specifically including: S210. Determine the service life range of the tower crane and divide the service life range into multiple service life intervals; S220. Based on the frequency of occurrence of the associated targets within each working year interval, calculate the co-occurrence frequency and transmission time of the associated targets within each working year interval. S230. Determine the correlation coefficient between related targets based on co-occurrence frequency and transmission time.

[0043] In steps S210 and S220, the working years of the tower crane are reasonably divided, for example, into different intervals such as 0 to 5 years, 5 to 10 years, and more than 10 years. For each interval, the frequency of occurrence of related targets is statistically analyzed, and then the co-occurrence frequency and transmission time are calculated.

[0044] In step S230, the correlation coefficient can be calculated by assigning different weights to co-occurrence frequency and transmission time according to actual needs and experience, and then performing a weighted calculation to obtain the correlation coefficient, so as to better reflect the degree of influence of different factors on the correlation. The calculation formula is as follows: K = a × P + b × T; Where a and b are the weighting coefficients for co-occurrence frequency and propagation time, respectively. These two weighting coefficients can be flexibly adjusted according to the specific type of tower crane, working environment, and characteristics of historical maintenance data. P is the co-occurrence frequency, and T is the propagation time.

[0045] By meticulously dividing the service life of tower cranes and analyzing the co-occurrence frequency and transmission timeliness of related targets in each interval, the dynamic changes in the correlation between components under different service lifespans can be captured more accurately. This not only considers the characteristics of the performance degradation of each component over time during tower crane use, but also combines the timeliness of the correlation strength between components reflected in actual maintenance data, thereby ensuring the rationality of the correlation coefficient determination. Based on the correlation coefficient, the degree of mutual influence between related targets can be assessed more accurately, providing more reliable data support for the formulation of subsequent maintenance and inspection strategies, and helping to further improve the accuracy and efficiency of tower crane maintenance and inspection.

[0046] Example 5 See Figure 4 In one specific embodiment, control standards for the detection target are determined based on historical maintenance data, image data of the detection target is acquired, and the image data is identified according to the control standards to obtain the identification result, specifically including: S310. Based on historical maintenance data, obtain the risk characteristics of each inspection target, and determine the control standards for each inspection target based on the risk characteristics; S320. Identify the image data according to the control standards to obtain the risk probability; S330. When the risk probability is greater than the first risk threshold, an abnormal target is obtained; S340. When the risk probability is less than the first risk threshold and greater than the second risk threshold, the image data is identified according to a preset method to obtain the identification result.

[0047] In step S310, the historical maintenance data records in detail the types of failures, failure frequencies, and maintenance measures that occurred in various components of the tower crane at different stages of use. This information is an important basis for determining the risk characteristics of the detection target. For example, some components may be prone to cracking due to long-term exposure to large loads, and the frequency and length changes of cracks can be used as one of the risk characteristics of that component.

[0048] In step S320, after acquiring the image data of the target to be detected, an advanced image recognition algorithm is used to compare and analyze the image data with the predetermined control standards. Typically, the image recognition algorithm can accurately capture various features of the target to be detected in the image, such as size, shape, and surface defects, and match these features one by one with the normal parameter range and abnormal judgment threshold specified in the control standards. Through this comparison and analysis, the algorithm can calculate the probability of the target to be detected being at risk, i.e., the risk probability.

[0049] In step S330, the first risk threshold is a pre-set critical limit value, which represents the critical point at which the detection target may have a serious risk. It can be determined by statistical analysis of historical maintenance data, combined with industry experience and expert advice. When the calculated risk probability exceeds this first risk threshold, it indicates that the detection target has a high probability of serious risk. At this time, it will be judged as an abnormal target so that corresponding maintenance and detection measures can be taken in time to prevent the risk from escalating further.

[0050] In step S340, the second risk threshold is another key boundary value that is lower than the first risk threshold. When the risk probability is between the first risk threshold and the second risk threshold, it indicates that the detected target has a certain risk, but has not yet reached a serious level. At this time, the system will re-identify the image data according to a preset method, such as using a more refined image recognition algorithm or combining other relevant data for comprehensive analysis, in order to obtain a more accurate identification result and determine whether the detected target really has a risk and the type and degree of the risk.

[0051] By deeply analyzing historical maintenance data, the risk characteristics of each detection target can be accurately extracted, and scientific and reasonable control standards can be formulated accordingly. These standards not only consider the inherent properties of the components but also their performance in historical maintenance, making them more practical and targeted. Using these control standards to intelligently identify image data helps to accurately calculate the risk probability of each detection target, providing a reliable basis for subsequent anomaly judgment. For detection targets with risk probabilities between the first and second risk thresholds, further identification is performed using preset methods, ensuring the accuracy and comprehensiveness of the identification results.

[0052] Example 6 In one specific embodiment, the development rate of abnormal targets is obtained, and the operational status of key targets is predicted based on the development rate and correlation coefficient to obtain the predicted status, specifically including: S410. Obtain image data of the abnormal target at different time points, and determine the development rate of the abnormal target based on the image data at different time points; S420. Based on the development rate and correlation coefficient, the predicted state is obtained; S430. Based on the predicted state and state threshold, determine whether there are any abnormal risks for key targets.

[0053] In step S410, generally speaking, an abnormal state of a certain structure will accelerate the development rate of abnormal states of related structures associated with the threshold. For example, when a crack appears in a key support structure of a tower crane and gradually expands, other support structures or connecting parts closely connected to it may accelerate their wear or damage process due to changes in force. Generally, the time node can be flexibly set according to the actual situation, such as every certain number of working hours, every day, or every week. Through advanced image processing and analysis technology, image data at different time nodes are carefully compared to accurately measure the changes in the size, shape, and degree of defects of the abnormal target, and then calculate its development rate. For example, for crack-type anomalies, parameters such as the length and width of the crack at different time nodes can be measured, and the development rate of the crack can be determined by calculating the change per unit time. For wear-type anomalies, parameters such as the wear depth and area of ​​the component can be measured to calculate the wear development rate.

[0054] In steps S420 and S430, the development rate reflects the current trend of the abnormal target, while the correlation coefficient reflects the degree of mutual influence between the abnormal target and other related targets. Combining these two for comprehensive analysis allows for a more accurate prediction of the operational status of key targets. Typically, the predicted status is a comprehensive evaluation level, used to intuitively reflect the operational status of key targets within the current timeframe. For example, the evaluation level can be divided into four levels: normal, slightly abnormal, moderately abnormal, and severely abnormal. Based on the magnitude of the development rate and the strength of the correlation coefficient, the evaluation level of the key target is determined according to a pre-set state threshold. For example, the development rate can be multiplied by the correlation coefficient, and the product can be divided into multiple level ranges, each corresponding to an evaluation level. When the product falls within a certain level range, the key target is determined to be in the corresponding evaluation level. The status threshold can be obtained by statistical analysis of the development trend and correlation impact of similar abnormal situations in historical maintenance data, or it can be reasonably set by combining industry experience and expert advice. For example, for some tower cranes with extremely high safety requirements, their status thresholds should be set more strictly in order to detect potential risks earlier; while for some tower cranes with relatively stable working environments and low usage frequency, the status thresholds can be appropriately relaxed.

[0055] By acquiring image data of abnormal targets at different time points, we can accurately track their development and changes, and then accurately calculate their development rate, providing key data support for subsequent predictions. This makes the prediction results more realistic and forward-looking. Combined with correlation coefficients, we can comprehensively consider the mutual influence between abnormal targets and their related targets, thereby more comprehensively assessing the operational status of key targets. By comparing the predicted status with the preset status thresholds, we can promptly identify abnormal risks in key targets, providing a strong basis for timely maintenance measures.

[0056] Example 7 In a specific embodiment, a detection deviation result is obtained based on the identification result and the predicted state of the key target. The key target is then re-inspected based on the detection deviation result to obtain a re-inspection result, specifically including: S510. When there is an abnormal risk to a key target, compare the identification result of the key target with the predicted state to obtain the similarity, and record the similarity as the detection deviation result. S520. When the detection deviation result is greater than the deviation threshold, the key target is re-inspected and the re-inspection result is generated.

[0057] In step S510, the operating status level of the key target can be obtained based on the recognition result, and the similarity between the two can be judged to determine whether there is a detection bias. The detection bias refers to the possible false detection and missed detection when performing image recognition on the tower crane.

[0058] In step S520, the deviation threshold is a key parameter. It is determined based on the tower crane's safety standards, historical maintenance data, and actual operating experience. When the detected deviation result is greater than this deviation threshold, it means that the deviation between the predicted state and the actual identification result exceeds the acceptable range. There may be abnormal situations that have not been accurately captured or there may be deficiencies in the prediction system.

[0059] By comparing the identification results of key targets with the predicted status and calculating the similarity, the degree of deviation between the initial detection and the prediction can be quantitatively assessed, providing a scientific basis for subsequent re-inspection decisions. When the detection deviation exceeds the preset threshold, the system automatically triggers the re-inspection process, effectively avoiding the risk of misjudgment that may be caused by a single detection result, and ensuring the rigor of maintenance and inspection. The introduction of the re-inspection process not only improves the reliability of the detection results, but also further reduces the probability of missed detection and false detection through the secondary verification mechanism, providing double protection for the safe operation of the tower crane.

[0060] Example 8 In one specific embodiment, updating historical maintenance data based on the re-inspection results and the predicted status specifically includes: S610. When there is a difference between the re-inspection result and the predicted status, determine the type of difference and update the historical maintenance data accordingly based on the type of difference. S620. When there is no difference between the re-inspection result and the predicted state, update the re-inspection result to the historical maintenance data.

[0061] In step S610, the discrepancies may be caused by various reasons, such as errors in the initial inspection, imperfections in the prediction system, or new changes to the tower crane components before the re-inspection. The types of discrepancies can be further subdivided into detection error type, system deviation type, and component change type. For detection error type discrepancies, it is necessary to correct the records of the corresponding detection targets in the historical maintenance data, add the correct detection results, and record the error situation so that the detection algorithm or equipment can be optimized later. For system deviation type discrepancies, it is necessary to analyze in which aspects the prediction system is deficient, adjust the system parameters or algorithm structure according to the re-inspection results to make the system more accurate, and then update the system-related parts in the historical maintenance data to ensure the accuracy of subsequent predictions. For component change type discrepancies, it is necessary to record in detail the new changes that occurred to the components before the re-inspection, such as new fault types and fault degrees, and update this information in the historical maintenance data to enrich the historical status information of the components.

[0062] In step S620, when there is no difference between the re-inspection result and the predicted state, it indicates that the initial inspection and prediction results are relatively accurate. At this time, the re-inspection result is directly updated to the historical maintenance data, which can further improve the completeness and timeliness of the historical maintenance data and provide more comprehensive and accurate data support for subsequent maintenance and inspection work.

[0063] By comparing the re-inspection results with the predicted status, the system can accurately identify the types of differences between the two and update the historical maintenance data accordingly. This not only ensures the timeliness and accuracy of the historical data but also enables subsequent inspections to be conducted based on a more comprehensive data foundation, thereby improving the overall reliability and effectiveness of the inspection. When there is no difference between the re-inspection results and the predicted status, the system automatically incorporates the re-inspection results into the historical maintenance data, realizing the dynamic accumulation and optimization of data.

[0064] Example 9 See Figure 5 In one specific embodiment, the present invention also provides a tower crane maintenance and inspection system 100 based on image recognition. The tower crane maintenance and inspection method based on image recognition described in the above embodiment is applied to the inspection system 100. The inspection system 100 includes: an image acquisition module 110 for acquiring image data; an image recognition module 120 for recognizing the image data according to control standards to obtain recognition results; a data judgment module 130 for obtaining predicted status and abnormal targets; and a data management module 140 for storing and updating historical maintenance data. The inspection system 100 has all the technical features of the above inspection method, which will not be described in detail here.

[0065] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A tower crane maintenance and inspection method based on image recognition, characterized in that, The detection method includes: Obtain the tower crane's maintenance and inspection tasks, and derive the inspection targets based on the maintenance and inspection tasks; The correlation criteria are determined based on historical maintenance data, the correlation targets are obtained based on the detection targets and the correlation criteria, and the correlation coefficients between the correlation targets are determined based on the working years of the tower crane. Based on the historical maintenance data, the control standards for the detection target are determined, image data of the detection target is acquired, and the image data is identified according to the control standards to obtain the identification result. Based on the identification results, abnormal targets are obtained, and the associated targets of the abnormal targets are recorded as key targets; The development rate of the abnormal target is obtained, and the operational status of the key target is predicted based on the development rate and the correlation coefficient to obtain the predicted status. Based on the identification result of the key target and the predicted state, a detection deviation result is obtained. Based on the detection deviation result, the key target is re-inspected to obtain a re-inspection result.

2. The detection method according to claim 1, characterized in that, The step of determining the correlation standard based on historical maintenance data, and obtaining the correlation target based on the detection target and the correlation standard, specifically includes: The historical maintenance data is obtained from the data management platform, and the location information, co-occurrence frequency and transmission time of each detection target are obtained based on the historical maintenance data; The association criteria between each of the detection targets are determined based on the location information, the co-occurrence frequency, and the transmission time. The detection targets that meet the association criteria are denoted as the association targets.

3. The detection method according to claim 2, characterized in that, The step of determining the association criteria between each of the detected targets based on the location information, the co-occurrence frequency, and the transmission time specifically includes: The detection targets whose co-occurrence frequency is greater than the frequency threshold or whose transmission time is less than the time threshold are recorded as the target group to be associated. Obtain the position information between the detected targets within the target group to be associated, and obtain the positional distance and positional relationship between the detected targets based on the position information; The target group to be associated is filtered by the location distance and the location relationship to obtain the associated targets.

4. The detection method according to claim 3, characterized in that, The step of determining the correlation coefficient between the associated targets based on the service life of the tower crane specifically includes: Determine the service life range of the tower crane, and divide the service life range into multiple service life intervals; Based on the frequency of occurrence of the associated targets within each of the said working years intervals, calculate the co-occurrence frequency and the transmission time of the associated targets within each of the said working years intervals; The correlation coefficient between the associated targets is determined based on the co-occurrence frequency and the transmission time.

5. The detection method according to claim 3, characterized in that, The process of determining the control standards for the detection target based on the historical maintenance data, acquiring image data of the detection target, and identifying the image data according to the control standards to obtain the identification result specifically includes: Based on the historical maintenance data, the risk characteristics of each of the detection targets are obtained, and based on the risk characteristics, the control standards for each of the detection targets are determined. The risk probability is obtained by identifying the image data according to the control standards. When the risk probability is greater than the first risk threshold, the abnormal target is obtained; When the risk probability is less than a first risk threshold and greater than a second risk threshold, the image data is identified according to a preset method to obtain the identification result.

6. The detection method according to claim 4, characterized in that, The process of obtaining the development rate of the abnormal target and predicting the operational status of the key target based on the development rate and the correlation coefficient to obtain the predicted status specifically includes: The image data of the abnormal target at different time points are acquired, and the development rate of the abnormal target is determined based on the image data at different time points; The predicted state is obtained based on the development rate and the correlation coefficient; Based on the predicted state and state threshold, it is determined whether there is any abnormal risk to the key target.

7. The detection method according to claim 5, characterized in that, The process of obtaining a detection deviation result based on the identification result and the predicted state of the key target, and then re-inspecting the key target based on the detection deviation result to obtain a re-inspection result, specifically includes: When the key target has an abnormal risk, the identification result of the key target is compared with the predicted state to obtain a similarity, and the similarity is recorded as the detection deviation result; When the detection deviation result is greater than the deviation threshold, the key target is re-inspected, and the re-inspection result is generated.

8. The detection method according to claim 7, characterized in that, The historical maintenance data is updated based on the re-inspection results and the predicted status.

9. The detection method according to claim 8, characterized in that, The step of updating the historical maintenance data based on the re-inspection results and the predicted status specifically includes: When the re-inspection result differs from the predicted state, the type of difference is determined, and the historical maintenance data is updated accordingly based on the type of difference. When the re-inspection result is not different from the predicted state, the re-inspection result is updated to the historical maintenance data.

10. A tower crane maintenance and inspection system based on image recognition, used to implement the inspection method according to any one of claims 1 to 9, characterized in that, The detection system includes: An image acquisition module, wherein the image acquisition module is used to acquire the image data; An image recognition module is used to recognize the image data according to the control standard and obtain the recognition result; A data judgment module is used to obtain the predicted state and the abnormal target; A data management module is used to store and update the historical maintenance data.

Images