Method, device, system and equipment for maintenance work order distribution of unmanned aerial vehicle and medium

Abstract

The application discloses a kind of unmanned vehicle's maintenance work order distribution method, device, system, equipment and medium, the method is: calling preset image recognition model identifies the image feature of the real-time image of unmanned vehicle, and based on image feature calculation unmanned vehicle's fault risk value;When fault risk value is greater than preset risk value, obtain the historical operation information of unmanned vehicle, based on historical operation information and fault risk value calculation fault comprehensive score value;Calling preset Attention algorithm model according to image feature in fault comprehensive score value corresponding scheme database filters target maintenance scheme and constructs maintenance task work order;According to maintenance level matching corresponding target personnel portrait, maintenance task work order is distributed to the maintenance personnel corresponding to target personnel portrait.The application can be maintained and detected by shooting image, a large number of unmanned vehicles can be batch processed to shorten processing time and improve efficiency, and the processing accuracy can be further improved by image fault detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle maintenance and repair, and particularly relates to a maintenance work order distribution method, device, system, equipment and medium for an unmanned aerial vehicle. BACKGROUND

[0002] With the construction of digital power grids, the application of unmanned aerial vehicles in the power grid industry continues to increase. A large number of unmanned aerial vehicles are used in different production and operation scenarios of the power grid on a daily basis, bringing great benefits to the construction of digital power grids. As the frequency of using unmanned aerial vehicles in the industry increases, maintenance and repair processing becomes increasingly important.

[0003] In order to improve the safety and stability of daily flight operations of unmanned aerial vehicles, the commonly used maintenance and repair processing method is as follows: a technical personnel performs preliminary inspection on the unmanned aerial vehicle that has completed the flight task on site, determines whether the unmanned aerial vehicle has a fault, and uploads the information of the preliminary inspection to a background operation and maintenance system. The operation and maintenance system constructs a maintenance task or work order according to the preliminary inspection and the use frequency of the unmanned aerial vehicle and distributes it to the background personnel for maintenance and repair processing.

[0004] However, the commonly used method has the following technical problems: the information can only be uploaded after the technical personnel performs inspection on the unmanned aerial vehicle after each operation on site, which not only reduces the efficiency of maintenance and repair processing, but also makes the inspection and maintenance levels of different technical personnel uneven, the inspection is biased, the information is prone to errors, and the subsequent maintenance and repair processing is prone to errors, resulting in a high failure rate of the unmanned aerial vehicle and a low actual use rate. SUMMARY

[0005] The present application provides a maintenance work order distribution method, device, system, equipment and medium for an unmanned aerial vehicle, which can solve the technical problems of low efficiency and low inspection accuracy of the prior art maintenance and repair processing, and poor repair processing effect.

[0006] A first aspect of the embodiments of the present application provides a maintenance work order distribution method for an unmanned aerial vehicle, the method comprising:

[0007] calling a preset image recognition model to recognize image features of real-time images of the unmanned aerial vehicle, and calculating a fault risk value of the unmanned aerial vehicle based on the image features, wherein the real-time images are images taken from different angles after the unmanned aerial vehicle completes the on-site operation and is stored on a storage platform;

[0008] when the fault risk value is greater than a preset risk value, obtaining historical operation information of the unmanned aerial vehicle, determining a maintenance score value and a depreciation score value based on the historical operation information, and calculating a fault comprehensive score value using the maintenance score value, the depreciation score value and the fault risk value;

[0009] determine a maintenance level based on the failure comprehensive score value, call a preset Attention algorithm model to screen a plurality of target maintenance schemes from a scheme database corresponding to the maintenance level according to the image features, and construct a maintenance task work order by using the plurality of target maintenance schemes;

[0010] match a corresponding target personnel profile from a preset personnel database according to the maintenance level, and distribute the maintenance task work order to a maintenance personnel corresponding to the target personnel profile.

[0011] In combination with the first aspect, in an implementation manner, the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the failure risk value of the unmanned aerial vehicle based on the image features include:

[0012] The preset image recognition model is called to identify the chroma value of the component region corresponding to each of the real-time images to obtain first image features.

[0013] Each of the real-time images is subjected to image cropping based on the first image features to obtain component images.

[0014] An edge detection algorithm is used to identify and extract the edge lines of the component outlines of each of the component images to obtain component outline lines.

[0015] The similarity of each of the component outline lines to a preset line is determined to obtain line similarities, and the failure risk value is calculated by using the plurality of line similarities.

[0016] In combination with the first aspect, in an implementation manner, the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the failure risk value of the unmanned aerial vehicle based on the image features include:

[0017] The preset image recognition model is called to identify cracks of the real-time images according to image pixels to obtain second image features.

[0018] The area and the number of the damaged regions corresponding to each of the real-time images are determined based on the second image features to obtain an area parameter and a number parameter.

[0019] The failure risk value is calculated by using the area parameter and the number parameter.

[0020] In combination with the first aspect, in an implementation manner,

[0021] The calling of the preset Attention algorithm model to screen a plurality of target maintenance schemes from the scheme database corresponding to the maintenance level according to the image features includes:

[0022] The Attention algorithm model is called to calculate the matching degrees of each historical maintenance scheme in a scheme database corresponding to the image feature and the maintenance level, and a plurality of scheme matching degrees are obtained.

[0023] A scheme score value of each historical maintenance scheme is obtained, and a plurality of historical maintenance schemes are sorted according to the scheme score value and the scheme matching degree, and a scheme list is obtained.

[0024] A plurality of target maintenance schemes are extracted from the scheme list based on a preset quantity value.

[0025] In combination with the first aspect, in an implementation manner,

[0026] The corresponding target personnel portrait is matched from a preset personnel database according to the maintenance level, and the matching comprises:

[0027] The maintenance task order is added to a preset task list, the maintenance task order in the preset task list is adjusted according to the maintenance level, and a maintenance time node after sorting is determined;

[0028] When the maintenance time node is reached, a text matching degree of the scheme text of the maintenance task order and the text of each personnel portrait contained in the preset personnel database is determined;

[0029] The target personnel portrait is determined according to the size of the text matching degree.

[0030] In combination with the first aspect, in an implementation manner, after the step that the fault risk value is greater than the preset risk value, the method further comprises:

[0031] The idle duration of the unmanned aerial vehicle parked on the storage platform is counted;

[0032] The push list of the unmanned aerial vehicle is updated according to the size of the idle duration;

[0033] The push list is visually displayed on a preset screen for reference by a technician.

[0034] The second aspect of the embodiment of the application provides a maintenance order distribution device of an unmanned aerial vehicle, and the device comprises:

[0035] The recognition module is configured to call a preset image recognition model to recognize an image feature of a real-time image of the unmanned aerial vehicle, and calculate a fault risk value of the unmanned aerial vehicle based on the image feature, wherein the real-time image is an image captured from different angles after the unmanned aerial vehicle completes field work and is stored on a storage platform.

[0036] The computing module is configured to, when the fault risk value is greater than a preset risk value, acquire historical operation information of the unmanned aerial vehicle, determine a maintenance score value and a depreciation score value based on the historical operation information, and calculate a fault comprehensive score value by using the maintenance score value, the depreciation score value, and the fault risk value.

[0037] The constructing module is configured to determine a maintenance level based on the fault comprehensive score value, call a preset Attention algorithm model to filter a plurality of target maintenance schemes from a scheme database corresponding to the maintenance level according to the image features, and construct a maintenance task work order by using the plurality of target maintenance schemes.

[0038] The assigning module is configured to match a corresponding target personnel profile from a preset personnel database according to the maintenance level, and assign the maintenance task work order to a maintenance personnel corresponding to the target personnel profile.

[0039] In an implementation manner of the second aspect, the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the fault risk value of the unmanned aerial vehicle based on the image features include:

[0040] The calling of the preset image recognition model to identify the chroma values of the component regions corresponding to each of the real-time images to obtain first image features;

[0041] The image cropping of each of the real-time images based on the first image features to obtain component images;

[0042] The identification and extraction of the edge lines of the component outlines of each of the component images by using an edge detection algorithm to obtain component outline lines;

[0043] The determination of the similarity of each of the component outline lines to a preset line to obtain line similarities, and the calculation of the fault risk value by using the plurality of line similarities.

[0044] In an implementation manner of the second aspect, the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the fault risk value of the unmanned aerial vehicle based on the image features include:

[0045] The calling of the preset image recognition model to identify cracks of the real-time images according to image pixels to obtain second image features;

[0046] The determination of the area and the number of the damaged regions corresponding to each of the real-time images based on the second image features to obtain an area parameter and a number parameter;

[0047] The calculation of the fault risk value by using the area parameter and the number parameter.

[0048] In an implementation manner of the second aspect,

[0049] The calling preset Attention algorithm model filters a plurality of target maintenance schemes from the scheme database corresponding to the maintenance level according to the image features, and the method comprises the steps of:

[0050] The Attention algorithm model is called to calculate the matching degree of the image features and each historical maintenance scheme in the scheme database corresponding to the maintenance level, and a plurality of scheme matching degrees are obtained.

[0051] The scheme score value of each historical maintenance scheme is obtained, the plurality of historical maintenance schemes are sorted according to the scheme score value and the scheme matching degree, and a scheme list is obtained.

[0052] A plurality of target maintenance schemes are extracted from the scheme list based on a preset number value.

[0053] In combination with the second aspect, in an implementation mode,

[0054] The corresponding target personnel portrait is matched from the preset personnel database according to the maintenance level, and the method comprises the steps of:

[0055] The maintenance task order is added to a preset task list, the sorting of the maintenance task order in the preset task list is adjusted according to the maintenance level, and a maintenance time node after sorting is determined.

[0056] When the maintenance time node is reached, the text matching degree of the scheme text of the maintenance task order and the text of each personnel portrait contained in the preset personnel database is determined.

[0057] The target personnel portrait is determined according to the size of the text matching degree.

[0058] In combination with the second aspect, in an implementation mode, the device further comprises:

[0059] The statistical module is configured to, after the step of the fault risk value being greater than the preset risk value, count the idle duration of the unmanned aerial vehicle parked on the storage platform.

[0060] The update module is configured to update the push list of the unmanned aerial vehicle according to the size of the idle duration.

[0061] The display module is configured to visually display the push list on a preset screen for reference by a technician.

[0062] The third aspect of the embodiment of the application provides a maintenance order distribution system of an unmanned aerial vehicle, and the system comprises a storage platform and a management platform, and the management platform is suitable for the maintenance order distribution method of the unmanned aerial vehicle.

[0063] The storage platform is provided with a plurality of cameras, the storage platform is used for placing a UAV, the plurality of cameras are distributed and arranged at different positions of the storage platform to shoot real-time images of the UAV from different angles, and the management platform is connected with the plurality of cameras.

[0064] Compared with the prior art, the embodiment of the application provides a UAV maintenance work order distribution method, device, system, equipment and medium, which has the beneficial effects that the image recognition model can be called to identify the image features of the real-time image of the UAV, and the fault risk value of the UAV can be calculated based on the image features; when the fault risk value is greater than the preset risk value, the historical operation information of the UAV is obtained, the maintenance score value and the depreciation score value are determined based on the historical operation information, and the fault comprehensive score value is calculated by using the maintenance score value, the depreciation score value and the fault risk value; the repair level is determined based on the fault comprehensive score value, the Attention algorithm model is called to filter a plurality of target maintenance schemes from the scheme database corresponding to the repair level according to the image features, and the maintenance task work order is constructed by using the plurality of target maintenance schemes; the target personnel portrait corresponding to the maintenance personnel is matched from the preset personnel database according to the repair level, and the maintenance task work order is distributed to the target personnel portrait corresponding to the maintenance personnel. The application can determine the fault risk of the UAV by shooting the image of the UAV after the UAV completes the operation, and then detect the maintenance and repair according to the fault risk. Through image detection, a large number of UAVs can be batch processed to shorten the processing time and improve the efficiency. Through image fault detection, omission can be avoided, and the processing accuracy is further improved. On this basis, the corresponding maintenance scheme is matched according to the image features of the UAV to perform maintenance, which can improve the maintenance effect and prolong the service life of the UAV. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 is a flow diagram of a UAV maintenance work order distribution method provided by an embodiment of the application;

[0066] Figure 2 is a structural diagram of a UAV maintenance work order distribution device provided by an embodiment of the application;

[0067] Figure 3 is a structural diagram of a UAV maintenance work order distribution system provided by an embodiment of the application. DETAILED DESCRIPTION

[0068] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0069] With the construction of digital power grids, the application of unmanned aerial vehicles in the power grid industry continues to increase, and a large number of unmanned aerial vehicles are used in different production and operation scenarios of the power grid in daily application, bringing great benefits to the construction of digital power grids. As the frequency of using unmanned aerial vehicles in the industry gradually increases, maintenance processing (repair and maintenance processing) becomes increasingly important.

[0070] In order to improve the safety and stability of daily flight operations of unmanned aerial vehicles, the commonly used maintenance processing method at present is: a technical personnel performs preliminary inspection on the unmanned aerial vehicle that has completed the flight task on site, determines whether the unmanned aerial vehicle has a fault, and uploads the information of the preliminary inspection to a background operation and maintenance system, and the operation and maintenance system constructs a maintenance task or work order according to the preliminary inspection and the use frequency of the unmanned aerial vehicle and distributes it to the background personnel for maintenance and repair processing.

[0071] However, the commonly used method has the following technical problems: the information can be uploaded only after the technical personnel performs inspection on the unmanned aerial vehicle after each on-site operation, which not only has low maintenance processing efficiency, but also has uneven inspection and maintenance levels of different technical personnel, and the inspection has deviation, so that the information is prone to error, resulting in error in subsequent maintenance processing, and the fault rate of the unmanned aerial vehicle is high and the actual use rate is low.

[0072] In order to solve the above problems, the maintenance work order distribution method, device, system, equipment and medium for unmanned aerial vehicles provided by the embodiments of the present application will be described and explained in detail below through the following specific embodiments.

[0073] In order to solve the technical problems of low monitoring and processing efficiency and low inspection accuracy of the cable distribution box in the prior art, with reference to Figure 1 , a flowchart of a maintenance work order distribution method for unmanned aerial vehicles provided by an embodiment of the present application is shown.

[0074] The maintenance work order distribution method for unmanned aerial vehicles is applicable to the management background or management platform of the storage platform of the unmanned aerial vehicles.

[0075] The storage platform can be a background for placing unmanned aerial vehicles. Technical personnel can apply for extracting unmanned aerial vehicles and obtaining unmanned aerial vehicles from the storage platform every time they perform high-altitude maintenance operations of the power grid. After completing the operation, the unmanned aerial vehicles can be placed on the storage platform, thereby facilitating unified management.

[0076] In an embodiment, the management background or management platform can be a data platform for managing various data of the unmanned aerial vehicle.

[0077] The maintenance order allocation method of the unmanned aerial vehicle, as an example, can include:

[0078] S11, calling a preset image recognition model to identify the image features of the real-time image of the unmanned aerial vehicle, and calculating the fault risk value of the unmanned aerial vehicle based on the image features, wherein the real-time image is an image taken from different angles after the unmanned aerial vehicle completes the field operation and is stored on the storage platform.

[0079] In an embodiment, a plurality of cameras can be arranged on the storage platform or the storage platform, and the unmanned aerial vehicle can be monitored through the cameras.

[0080] In an operation mode, the management background can be connected with different cameras, and the images of the unmanned aerial vehicle are taken through the cameras to obtain real-time images.

[0081] Specifically, when the technician finishes using the unmanned aerial vehicle, the unmanned aerial vehicle is placed on the storage platform or the storage platform. At this time, the camera can be controlled to take the image of the unmanned aerial vehicle to obtain the real-time image.

[0082] If the unmanned aerial vehicle is taken from a single angle, the detection may be biased. Therefore, the unmanned aerial vehicle can be taken from different angles to obtain real-time images from different angles. For example, a real-time image is taken from a top view, a real-time image is taken from a left side view, a real-time image is taken from a right side view, a real-time image is taken from an overhead view, and a real-time image is taken from a front view.

[0083] When taking pictures, the same time node can be used for taking pictures, so as to obtain several real-time images from different angles.

[0084] After obtaining the real-time image, a preset image recognition model can be called to identify the image features of the real-time image, and the image features can be chroma values or textures. Whether the unmanned aerial vehicle has a fault or damage is determined according to the chroma values or textures of the image, and then the fault risk value of the unmanned aerial vehicle is calculated according to the identified image features, so as to determine whether the unmanned aerial vehicle has a fault risk according to the fault risk value.

[0085] When the technician uses the unmanned aerial vehicle for high-altitude operation, the unmanned aerial vehicle is generally used to monitor the cables in the suburbs or on the hills. However, during the flight, the unmanned aerial vehicle may collide with the branches on the hills or the cables, and then the propeller, the arm or the various devices (cameras or sensors, etc.) carried by the unmanned aerial vehicle may be deformed.

[0086] In an optional embodiment, to determine whether the various components of the drone have deformed, and to determine the failure risk based on the degree of deformation, the step of calling a preset image recognition model to identify image features of the real-time image and calculating the failure risk value of the drone based on the image features may include the following sub-steps:

[0087] S21. Call the preset image recognition model to identify the chromaticity value of the machine area corresponding to each real-time image, and obtain the first image feature.

[0088] S22. Based on the first image features, each of the real-time images is cropped to obtain a machine part image.

[0089] S23. Use an edge detection algorithm to identify and extract the edge lines of the machine part contour in each machine part image to obtain the machine part contour lines.

[0090] S24. Determine the similarity between each of the machine part contour lines and the preset lines to obtain the line similarity, and use several of the line similarities to calculate the fault risk value.

[0091] Since different components have different colors, different components in the image can be identified by their chromaticity values. Specifically, a preset image recognition model is called to identify the chromaticity values ​​of the component regions corresponding to each real-time image, thereby obtaining the first image feature.

[0092] In one operating mode, the preset image recognition model can be an RGB model, HSV model, HSL model, or Lab model. Taking the HSV or HSL model as an example, it can identify the chromaticity values ​​of the image region where the component is located in each real-time image, obtaining the first image feature. This model can achieve color separation and adjustment by independently adjusting color attributes. Users can pre-input the chromaticity values ​​of different components or parts to train the model. When invoked, the model can be applied to color detection and segmentation in image processing, extracting objects of specific colors by setting a hue threshold.

[0093] Next, the model can be invoked to crop each real-time image based on the chromaticity values ​​of the first image features to obtain the machine component image. For example, after identifying the chromaticity values ​​of the propeller, the propeller region can be cropped from the real-time image to obtain the machine component image. Similarly, the arm region can be cropped from the real-time image to obtain the machine component image.

[0094] In practice, each real-time image corresponds to a different angle, and images of different machine parts can be cropped from each real-time image to obtain machine part images. For example, images of different machine parts can be cropped from a top-down view of a real-time image; images of different machine parts can be cropped from a front-view view of a real-time image to obtain machine part images.

[0095] Next, edge detection algorithms (such as the LoG (Log) algorithm or the Canny algorithm) can be used to identify the outline of each part in the image, and then the lines of the edge of the outline can be extracted to obtain the outline lines of the part.

[0096] Once the outline of the machine part is obtained, the similarity between the outline of the machine part and its corresponding preset lines can be calculated to obtain the line similarity.

[0097] The preset lines can be the outline of the same component. To determine whether each operation causes deformation of the component, each drone can acquire multiple images before going out to work. Then, each image is identified, cropped, and its outline is extracted according to the above operation method to obtain the corresponding lines, which are used as preset lines.

[0098] For example, before determining that technicians need to use drone A for field operations, images taken by drone A from a top-down angle, from a left-side angle, from a right-side angle, from a bottom-up angle, and from a front-view angle can be obtained.

[0099] Then, the propeller in the image taken from a top-down angle is identified, cropped, and its contour extracted to obtain the preset lines. Correspondingly, after the drone completes its operation and is placed on the platform, a real-time image can also be taken from a top-down angle. This real-time image is then identified, cropped, and its contour extracted to obtain the propeller's component contour lines. Next, the similarity between the propeller's component contour lines and the preset lines of the propeller can be calculated to obtain the line similarity score.

[0100] By calculating the line similarity of the contours of images taken from the same device at the same angle, it can be determined whether the drone has been deformed after the operation. Then, based on the line similarity, a fault risk assessment can be conducted to determine whether the drone poses a risk.

[0101] In one embodiment, after calculating multiple line similarities, the fault risk value can be calculated using several line similarities.

[0102] Since there are multiple line similarities, each device has multiple line similarities. For example, if five real-time images are captured, and each real-time image contains a propeller device, there are a total of five line similarities related to the propeller. Similarly, if five real-time images are captured, and four real-time images contain an arm device, there are a total of four line similarities related to the arm.

[0103] When calculating the fault risk value, the average value of the line similarity of the same device can be calculated, and finally, multiple average values ​​are used to calculate the fault risk value.

[0104] In one operating mode, the fault risk value can be calculated as follows:

[0105] R = A1S1 + A2S2 + … + A N S N .

[0106] Where R is the fault risk value; A1, A2 and A N The calculation parameters can be user-preset risk weight values, with different weight values ​​for different components or parts, and their range can be [-1, 1]; S i S1 represents the average line similarity, i = 1 - N, where N is the number of parts or devices for image recognition. For example, S1 is the average line similarity of the arm, S2 is the average line similarity of the propeller, and S3 is the average line similarity of the fuselage probe.

[0107] In another optional embodiment, when a drone is operating at high altitude outdoors, it may be scraped by various objects during flight, causing cracks or damage to different components. In order to conduct a risk assessment for cracks and damage, as an example, the step of calling a preset image recognition model to identify the image features of the real-time image and calculating the failure risk value of the drone based on the image features may include the following sub-steps:

[0108] S31. Call the preset image recognition model to identify the cracks in the real-time image based on the image pixels, and obtain the second image features.

[0109] S32. Based on the second image features, determine the area and number of damaged areas corresponding to each real-time image to obtain area parameters and number parameters.

[0110] S33. Calculate the fault risk value using the area parameter and the quantity parameter.

[0111] In one embodiment, a preset image recognition model can be invoked to identify cracks in the real-time image based on image pixels, thereby obtaining a second image feature.

[0112] In specific operations, a preset image recognition model can be called to calculate the gradient magnitude of each pixel in the real-time image and establish a gradient histogram. Specifically, the Sobel operator can be used to calculate the gradient in the x-direction and the gradient in the y-direction of each pixel in the surface image to obtain the corresponding gradient magnitude, and then a gradient histogram can be established based on the gradient magnitude of each pixel in the surface image.

[0113] Gradient histogram equalization can effectively solve the problems of low image contrast and blurred details. After processing, the contrast between cracked areas and other areas is enhanced, making crack edges clearer and facilitating the extraction of crack severity data. However, traditional gradient histogram equalization can lead to enhancement and loss of detail. Gradient loss can cause loss of image detail, so the surface image needs to be corrected before equalization.

[0114] In this embodiment of the invention, the gradient histogram is divided into two parts by selecting the optimal segmentation threshold, thereby correcting the surface image. This is because when the gradient level range is too large or the frequency distribution is uneven during gradient histogram equalization, the equalized image will be over-enhanced.

[0115] Next, the crack probability corresponding to each gradient level is obtained based on the magnitude of each gradient level in the gradient histogram, resulting in a crack probability function. Further, the crack probability corresponding to each gradient level is obtained based on the magnitude of each gradient level in the gradient histogram. Specifically: the maximum number of pixels and the maximum gradient level corresponding to the gradient level with the highest frequency in the gradient histogram are obtained. For example, if the gradient levels of the gradient histogram are 0-256, then the maximum gradient level corresponding to the gradient histogram is 256. The ratio of the maximum number of pixels to the maximum gradient level is the crack probability weight. The product of this crack probability weight and each gradient level is the crack probability of that corresponding gradient level. It should be noted that each gradient level corresponds to a crack probability, thus yielding a corresponding crack probability function. Since a larger gradient magnitude increases the likelihood of a crack, i.e., a larger gradient level increases the likelihood of a crack, a larger gradient level corresponds to a larger crack probability; the gradient level and crack probability are directly proportional. Therefore, as the gradient level increases, the crack probability function gradually increases.

[0116] The crack probability function and the gradient histogram curve are mapped to the same coordinate axis to obtain the intersection point between them. Since the gradient histogram shows a gradually decreasing trend in the latter half, while the crack probability function shows a gradually increasing trend, there must be an intersection point between them. The intersection point and the corresponding initial gradient level are then obtained. The purpose of obtaining the intersection point is to obtain an initial gradient level, which is located in the middle of the gradient histogram. The gradient level corresponding to this intersection point may be a boundary point, but it is only a rough estimate. Obtaining the initial gradient level reduces the massive computational cost of traversal. That is, this initial gradient level may be the segmentation threshold between edge and non-edge regions, but it has significant inaccuracies, so it is further corrected. Specifically, the pixels belonging to the initial gradient level are used as the center points of the windows, and the grayscale similarity of multiple windows is obtained. In this embodiment, the window size is 2*2.

[0117] Using the pixels belonging to the initial gradient level as the center point of the window, the gray-level similarity of multiple windows is obtained. With the initial gradient level as the center, the gradient levels are moved left and right to obtain the gray-level similarity corresponding to each gradient level. The gradient level corresponding to the maximum gray-level similarity is used as the optimal segmentation threshold. Then, based on the optimal segmentation threshold, the gradient histogram is divided into two intervals, and the average frequency and maximum frequency within each interval are obtained. A correction threshold is then obtained for each interval based on the average frequency and maximum frequency, and the correction threshold for each interval is updated by comparing the correction threshold with the maximum frequency. Finally, the updated correction threshold is used to correct the gradient histogram, and the corrected gradient histogram is equalized to obtain the corresponding crack image in the real-time image. The crack image is then extracted from the real-time image and used as a second image feature to determine whether the crack poses a risk to the drone.

[0118] Next, the area value of each crack image can be determined, and the number of crack images identified can be counted.

[0119] Since there may be multiple cracks, we can count the number of crack images contained in each real-time image to obtain data parameters; then determine the image area value of the crack image contained in each real-time image, and finally calculate the sum of the area values ​​of multiple images to obtain area parameters.

[0120] Finally, the failure risk value can be calculated based on the area and quantity parameters.

[0121] Since there are multiple real-time images, each real-time image corresponds to an area parameter and a quantity parameter. In order to integrate the parameters corresponding to multiple real-time images, the area parameter and quantity parameter corresponding to each real-time image can be used to calculate the image parameters, and then the fault risk value can be calculated using multiple image parameters.

[0122] In one embodiment, the fault risk value can be calculated as follows:

[0123] R = L1 + L2 + L3 + ... + L N .

[0124] Where R is the fault risk value, L1 is the image parameter corresponding to the first real-time image, L2 is the image parameter corresponding to the second real-time image, L3 is the image parameter corresponding to the third real-time image, and L... N Here are the image parameters corresponding to the Nth real-time image, where N is the number of real-time images.

[0125] The image parameters can be calculated as follows:

[0126] L i =X qi *Ei.

[0127] L i Let be the image parameters corresponding to the i-th real-time image, and let X be a calculation constant, which can be a weight value preset by the user, and its value range can be [0-1]; q i E represents the quantity parameter corresponding to the i-th real-time image; i Let be the area parameter corresponding to the i-th real-time image.

[0128] S12. When the fault risk value is greater than the preset risk value, the historical operation information of the UAV is obtained, and the maintenance score value and depreciation score value are determined based on the historical operation information. The comprehensive fault score value is calculated using the maintenance score value, the depreciation score value and the fault risk value.

[0129] After calculating the fault risk value, it can be determined whether the fault risk value is greater than the preset risk value. If the fault risk value is greater than the preset risk value, it indicates that the drone has a high fault risk and requires repair and maintenance. Conversely, if the fault risk value is less than the preset risk value, it indicates that the drone has a low fault risk and does not require repair and maintenance at this time.

[0130] Once it is determined that the drone has a high risk of failure and requires repair and maintenance, the drone's historical operation information can be obtained. The failure risk value and historical operation information can be used to create a maintenance task order, which can then be assigned to different maintenance personnel to notify them to perform repair and maintenance on the drone.

[0131] Historical operation information can include details of high-altitude operations performed by drones, such as the operation scenario, location, route, and information recorded by technicians. It can also include prior operation and maintenance information. Based on this information and fault risk values, work orders can be generated, allowing maintenance personnel to perform appropriate repairs according to the work order details.

[0132] In one embodiment, maintenance score and depreciation score can be determined based on historical operation information, and a comprehensive fault score can be calculated using the maintenance score, depreciation score and fault risk value.

[0133] In one embodiment, the comprehensive fault score P can be calculated as follows:

[0134] P=G1R-G2Q-G3D.

[0135] In the above formula, P is the comprehensive fault score, Q is the maintenance score, D is the depreciation score, R is the fault risk value, and G1, G2 and G3 are calculation constants, which can be user-preset values ​​and can be adjusted according to actual needs. The values ​​of G1, G2 and G3 are greater than 0 and less than or equal to 1.

[0136] In one embodiment, the maintenance score value Q can be calculated as follows:

[0137] .

[0138] In the above formula, Q is the maintenance score, K is the single repair score, n is the number of repairs, and C is the overall score percentage.

[0139] In one embodiment, the overall score percentage value C can be calculated as follows:

[0140] C = (-1 / 6T + 10) * 100%.

[0141] In the above formula, T represents the expected number of days.

[0142] K, the score for a single repair, can be calculated as follows:

[0143] .

[0144] K1 is the rating for the first repair, K2 is the rating for the second repair, and K3 is the rating for the third repair. n Let be the score for the nth repair, where n is the number of historical repairs performed on the drone.

[0145] In one embodiment, the depreciation score D can be calculated as follows:

[0146] D=BJ.

[0147] In the above formula, D is the depreciation score, B is the original value of the drone, and J is the depreciation value of the drone. The depreciation value of the drone is calculated using the straight-line method, with a depreciation period of five years.

[0148] After calculating the overall fault score, it can be determined whether the overall fault score is greater than the preset score. If the overall fault score is greater than the preset score, it indicates that the fault is serious and the drone repair is worthwhile, making repair more urgent. Conversely, if the overall fault score is less than the preset score, it indicates that the fault is not serious or the drone repair is less worthwhile, making repair less urgent.

[0149] S13. Based on the comprehensive fault score, determine the maintenance level, call the preset Attention algorithm model to select several target maintenance solutions from the solution database corresponding to the maintenance level according to the image features, and construct a maintenance task work order using several target maintenance solutions.

[0150] In one embodiment, when the overall fault score is greater than the preset score, it indicates that the UAV fault is serious and the UAV has high repair value, and it is necessary to quickly find some maintenance solutions for maintenance.

[0151] In one embodiment, after calculating the comprehensive fault score, the maintenance level can be determined based on the comprehensive fault score. This maintenance level can be the priority of UAV maintenance, and work orders can be allocated according to the maintenance priority.

[0152] Specifically, the maintenance levels corresponding to the comprehensive fault score are shown in the table below:

[0153]

[0154] It should be noted that 60 points can be used as the scoring threshold for the comprehensive fault score. Fault scores below 60 points will not be repaired or maintained. Alternatively, 60 points can also be a preset score.

[0155] In one embodiment, different maintenance levels have different solution databases, and each solution database contains multiple different historical maintenance solutions. A preset Attention algorithm model can be invoked to select several solutions from the multiple different historical maintenance solutions contained in the solution database corresponding to the maintenance level based on image features. These selected solutions are then used as target maintenance solutions. Finally, maintenance task work orders are constructed using these target maintenance solutions for technicians to view or refer to.

[0156] In one embodiment, if maintenance personnel inspect the drone after receiving the work order and then devise a maintenance plan based on the inspection results, the entire process is time-consuming and inefficient. To improve processing efficiency, as an example, the step of calling a preset Attention algorithm model to select several target maintenance plans from the plan database corresponding to the maintenance level based on the image features may include the following sub-steps:

[0157] S131. Call the Attention algorithm model to calculate the matching degree between the image features and each historical maintenance plan in the plan database corresponding to the maintenance level, and obtain multiple plan matching degrees.

[0158] S132. Obtain the scheme score value for each historical maintenance scheme, and sort the multiple historical maintenance schemes according to the scheme score value and the scheme matching degree to obtain a scheme list.

[0159] S133. Extract several target maintenance schemes from the scheme list based on a preset quantity value.

[0160] In one embodiment, the Attention algorithm model can include two linear layers. The first linear layer performs a non-linear transformation on the image features to obtain transformed features. Specifically, the first and second image features can be used as model inputs, and the transformed features are obtained after passing through the first linear layer. Then, the transformed features can be matched with each historical maintenance plan to obtain the plan matching degree. In practice, the matching degree can be calculated using cosine similarity.

[0161] After calculating the matching degree between image features and each historical maintenance plan, a plan score value can be obtained for each historical maintenance plan. This plan score value is the score value added after the previous technician used the maintenance plan.

[0162] The system calculates the sum of the scheme score and the scheme matching degree, sorts multiple historical maintenance schemes according to the magnitude of the sum, and obtains a scheme list. Finally, several target maintenance schemes can be extracted from the scheme list based on a preset quantity value.

[0163] For example, if the preset quantity is 10, then the top 10 historical maintenance plans with the largest sum can be selected from the plan list as the target maintenance plan.

[0164] In practice, a database of solutions corresponding to different maintenance levels can be established. Then, the operational text is retrieved from the drone's historical operational information, and the fault type is determined based on image features (crack images or chromaticity values). For example, crack images correspond to damage types, and chromaticity values ​​correspond to deformation types. Several target maintenance solutions are then selected from the solution database based on the operational text and fault type. Specifically, solution selection can be performed through text matching; for example, the operational text and fault type are matched against the text of maintenance solutions in the solution database to select several target maintenance solutions.

[0165] Finally, maintenance task work orders can be constructed using information such as several target maintenance plans, fault types, maintenance score values, depreciation score values, and fault risk values.

[0166] S14. Match the corresponding target personnel profile from the preset personnel database according to the maintenance level, and assign the maintenance task work order to the maintenance personnel corresponding to the target personnel profile.

[0167] In one embodiment, there are multiple maintenance workers. In order to improve the efficiency of work order processing and match suitable workers for maintenance, the corresponding target personnel profile can be matched from a preset personnel database according to the maintenance level of the maintenance task work order, and the maintenance task work order can be assigned to the maintenance personnel corresponding to the target personnel profile.

[0168] In one embodiment, each worker has their own individual characteristics; for example, worker a is good at repairing circuits, worker b is good at repairing hardware, and worker c is good at handling robotic arms, etc.

[0169] Therefore, individual profiles can be constructed based on the characteristics of different staff members, resulting in personnel profiles. These profiles are then stored in a personnel database. In practice, tags can be created for each staff member's characteristics or traits, and these tags are used to construct personnel profiles.

[0170] Then, the maintenance work orders can be sorted according to their repair level. The sorting results can then be matched and filtered to find suitable personnel to perform maintenance, thereby improving the response speed and processing efficiency of the work orders.

[0171] In one alternative embodiment, there are a large number of drones but a small number of maintenance and repair personnel, making it difficult to perform one-on-one maintenance and repair. Therefore, maintenance and repair need to be carried out step by step.

[0172] If maintenance is performed sequentially, the processing efficiency is low. To improve efficiency, as an example, the step of matching the corresponding target personnel profile from a preset personnel database based on the maintenance level may include the following sub-steps:

[0173] S141. Add the maintenance task work order to the preset task list, adjust the sorting of the maintenance task work order in the preset task list according to the maintenance level, and determine the maintenance time node after sorting.

[0174] S142. When the maintenance time node is reached, determine the matching degree between the solution text of the maintenance task work order and the text of each personnel profile contained in the preset personnel database.

[0175] S143. Determine the target person profile based on the degree of text matching.

[0176] In one embodiment, the maintenance task work order's repair level and solution text can be obtained. The repair level corresponds to the overall fault score, as detailed in the table above. A higher overall fault score corresponds to a higher level. The solution text can be the keyword text of each repair solution within the maintenance task work order.

[0177] Each maintenance operation may require the handling of multiple drones, resulting in multiple maintenance work orders. These work orders can be prioritized based on their maintenance level. It's worth noting that multiple maintenance work orders can be created and awaiting technician execution. These work orders can be compiled into a pre-defined task list. The currently created work order is added to this list, and then the work orders within the list are sorted according to their maintenance level. Work orders with the same maintenance level are further sorted based on their creation order, and the final sorted order is determined. This sorted information is then transmitted to management personnel, who can determine the appropriate maintenance timeframe for each drone based on the order of their creation.

[0178] When the maintenance timeline is reached, the text of the maintenance task order can be compared with the tags of each personnel profile in the pre-set personnel database to calculate the text matching degree. Since the pre-set personnel database contains multiple personnel profiles, there are also multiple calculated text matching degrees. Finally, the target personnel profile can be determined based on the magnitude of the text matching degree.

[0179] Specifically, the person profile with the highest text match score can be selected as the target person profile.

[0180] After identifying the target personnel profile, maintenance work orders can be assigned to the smart terminals of the maintenance personnel corresponding to the target personnel profile, allowing the maintenance personnel to view the maintenance work orders on their smart terminals and perform repairs and maintenance according to the contents of the maintenance work orders.

[0181] In an optional embodiment, the drones stored on the platform may be needed by other technicians. To prevent technicians from selecting drones requiring maintenance within the platform, after the step where the fault risk value is greater than a preset risk value, the method may further include the following sub-steps:

[0182] S41. Calculate the idle time of drones parked on the statistical storage platform.

[0183] S42. Update the drone's push list based on the idle time.

[0184] S43. The push list is displayed visually on a preset screen for technical personnel to refer to.

[0185] In one embodiment, the idle time of drones parked on the storage platform can be counted. Specifically, the idle time of drones without maintenance work orders can be counted.

[0186] Then, the drone push list is updated based on the length of idle time. For example, idle drones are arranged from longest to shortest idle time, and information about each idle drone is obtained. The drone information is then arranged according to the sorted order to obtain the task push list.

[0187] The system displays a list of drones on a pre-defined screen for technical staff to reference. By creating and displaying this list, technicians can prioritize idle drones, thereby reducing drone idleness, increasing the utilization rate of idle drones, and rationally allocating drone resources.

[0188] In this embodiment, the present invention provides a method for allocating maintenance work orders for unmanned aerial vehicles (UAVs). Its advantages are as follows: The present invention can call a preset image recognition model to identify image features of real-time images of the UAV and calculate the UAV's fault risk value based on these image features; when the fault risk value is greater than a preset risk value, the historical operation information of the UAV is obtained, and maintenance score and depreciation score are determined based on the historical operation information, and a comprehensive fault score is calculated using the maintenance score, depreciation score, and fault risk value; the maintenance level is determined based on the comprehensive fault score, and a preset Attention algorithm model is called to select several target maintenance schemes from the scheme database corresponding to the maintenance level based on image features, and a maintenance task work order is constructed using these target maintenance schemes; the corresponding target personnel profile is matched from a preset personnel database according to the maintenance level, and the maintenance task work order is allocated to the maintenance personnel corresponding to the target personnel profile. This invention can determine the failure risk of a drone by taking images of the drone after it has completed its operation, and then carry out maintenance and inspection based on the failure risk. Image detection can process a large number of drones in batches, thereby shortening the processing time and improving efficiency. Fault detection through images can avoid omissions and further improve the processing accuracy. On this basis, by combining the drone image features with the corresponding maintenance plan for repair, the maintenance effect can be improved and the service life of the drone can be extended.

[0189] This invention also provides a maintenance work order allocation device for unmanned aerial vehicles (UAVs), see [link to relevant documentation]. Figure 2 The diagram shows a structural schematic of a maintenance work order allocation device for a drone according to an embodiment of the present invention.

[0190] As an example, the maintenance work order allocation device for the drone may include:

[0191] The recognition module 201 is used to call a preset image recognition model to recognize the image features of the drone's real-time images and calculate the drone's fault risk value based on the image features. The real-time images are images taken from different angles after the drone completes its on-site operations and is stored on the storage platform.

[0192] The calculation module 202 is used to obtain the historical operation information of the UAV when the fault risk value is greater than the preset risk value, determine the maintenance score value and the depreciation score value based on the historical operation information, and calculate the comprehensive fault score value using the maintenance score value, the depreciation score value and the fault risk value.

[0193] The construction module 203 is used to determine the maintenance level based on the comprehensive fault score, call the preset Attention algorithm model to select several target maintenance solutions from the solution database corresponding to the maintenance level according to the image features, and construct a maintenance task work order using several target maintenance solutions;

[0194] The allocation module 204 is used to match the corresponding target personnel profile from the preset personnel database according to the maintenance level, and allocate the maintenance task work order to the maintenance personnel corresponding to the target personnel profile.

[0195] Optionally, the step of calling a preset image recognition model to identify image features of the real-time image and calculating the failure risk value of the UAV based on the image features includes:

[0196] A preset image recognition model is invoked to identify the chromaticity value of the component region corresponding to each real-time image, thereby obtaining the first image feature;

[0197] Based on the first image features, each real-time image is cropped to obtain a machine component image;

[0198] The edge lines of the machine part contour in each machine part image are identified and extracted using an edge detection algorithm to obtain the machine part contour lines;

[0199] The similarity between each of the machine part contour lines and the preset lines is determined to obtain the line similarity, and the fault risk value is calculated using several of the line similarities.

[0200] Optionally, the step of calling a preset image recognition model to identify image features of the real-time image and calculating the failure risk value of the UAV based on the image features includes:

[0201] A preset image recognition model is invoked to identify cracks in the real-time image based on image pixels, thereby obtaining second image features;

[0202] Based on the second image features, the area and number of damaged areas corresponding to each real-time image are determined respectively, and the area parameters and number parameters are obtained;

[0203] The fault risk value is calculated using the area parameter and the quantity parameter.

[0204] Optionally,

[0205] The invocation of the preset Attention algorithm model filters several target maintenance solutions from the solution database corresponding to the maintenance level based on the image features, including:

[0206] The Attention algorithm model is called to calculate the matching degree between the image features and each historical maintenance plan in the plan database corresponding to the maintenance level, and multiple plan matching degrees are obtained.

[0207] Obtain the scheme score for each historical maintenance scheme, and sort the multiple historical maintenance schemes according to the scheme score and the scheme matching degree to obtain a scheme list;

[0208] Several target maintenance plans are extracted from the plan list based on a preset quantity value.

[0209] Optionally,

[0210] The step of matching the corresponding target personnel profile from a preset personnel database based on the maintenance level includes:

[0211] Add the maintenance task work order to the preset task list, adjust the sorting of the maintenance task work order in the preset task list according to the maintenance level, and determine the maintenance time node after sorting.

[0212] When the maintenance time node is reached, the matching degree between the solution text of the maintenance task work order and the text of each personnel profile contained in the preset personnel database is determined;

[0213] The target person's profile is determined based on the degree of text matching.

[0214] Optionally, the device further includes:

[0215] The statistics module is used to calculate the idle time of the drones parked on the storage platform after the step in which the fault risk value is greater than the preset risk value.

[0216] The update module is used to update the drone's push list based on the idle time.

[0217] The display module is used to visually display the push list on a preset screen for technical personnel to refer to.

[0218] This invention also provides a maintenance work order allocation system for unmanned aerial vehicles (UAVs), see [link to relevant documentation]. Figure 3 The diagram shows a structural schematic of a maintenance work order allocation system for a drone provided in an embodiment of the present invention.

[0219] As an example, the drone maintenance work order allocation system may include: a storage platform and a management platform, wherein the management platform is applicable to the drone maintenance work order allocation method as described in the above embodiments;

[0220] The storage platform is equipped with multiple cameras and is used to place drones. The multiple cameras are distributed and set at different positions on the storage platform to capture real-time images of the drones from different angles. The management platform is connected to the multiple cameras.

[0221] Those skilled in the art will understand that, for ease of description and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0222] Furthermore, this application also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the drone maintenance work order allocation method as described in the above embodiments.

[0223] Furthermore, this application embodiment also provides a computer-readable storage medium storing a computer-executable program, which is used to cause a computer to execute the maintenance work order allocation method for UAVs as described in the above embodiments.

[0224] In the description of the embodiments of the present invention, it should be noted that the terms "above," "below," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. When an element such as a layer, region, or substrate is referred to as being "above" or "on top of" another element, it may be directly on the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly on" or "above" another element, there is no intermediate element. It should also be understood that when an element is referred to as being "below" or "under" another element, it may be directly below or under the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly below" or "under" another element, there is no intermediate element. Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0225] Those skilled in the art will understand that embodiments of this application may also include computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may 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.

[0226] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), devices, 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.

[0227] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0228] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0229] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for allocating maintenance work orders for unmanned aerial vehicles (UAVs), characterized in that, The method comprises: calling a preset image recognition model to identify image features of real-time images of the unmanned aerial vehicle, and calculating a failure risk value of the unmanned aerial vehicle based on the image features, wherein the real-time images are images taken from different angles after the unmanned aerial vehicle completes field work and is stored on a storage platform; when the failure risk value is greater than a preset risk value, obtaining historical work information of the unmanned aerial vehicle, determining a maintenance score value and a depreciation score value based on the historical work information, and calculating a failure comprehensive score value using the maintenance score value, the depreciation score value and the failure risk value; determining a maintenance level based on the failure comprehensive score value, calling a preset Attention algorithm model to filter a plurality of target maintenance schemes from a scheme database corresponding to the maintenance level according to the image features, and constructing a maintenance task work order using the plurality of target maintenance schemes; matching a corresponding target personnel profile from a preset personnel database according to the maintenance level, and distributing the maintenance task work order to a maintenance personnel corresponding to the target personnel profile; the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the failure risk value of the unmanned aerial vehicle based on the image features comprises: calling a preset image recognition model to identify the chroma value of the part area corresponding to each of the real-time images to obtain a first image feature; performing image cropping on each of the real-time images based on the first image feature to obtain a part image; identifying and extracting the edge line of the part contour of each of the part images using an edge detection algorithm to obtain a part contour line; determining the similarity of each of the part contour lines to a preset line to obtain a line similarity, and calculating a failure risk value using a plurality of the line similarities; or; the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the failure risk value of the unmanned aerial vehicle based on the image features comprises: calling a preset image recognition model to identify cracks in the real-time images according to image pixels to obtain a second image feature; determining the area and the number of the damaged areas corresponding to each of the real-time images based on the second image feature to obtain an area parameter and a number parameter; calculating a failure risk value using the area parameter and the number parameter. 2.The method of claim 1, wherein, the calling of the preset Attention algorithm model to filter a plurality of target maintenance schemes from the scheme database corresponding to the maintenance level according to the image features comprises: calling an Attention algorithm model to calculate the matching degree of the image features and each historical maintenance scheme in the scheme database corresponding to the maintenance level to obtain a plurality of scheme matching degrees; obtaining a scheme score value of each historical maintenance scheme, sorting a plurality of historical maintenance schemes according to the scheme score value and the scheme matching degree to obtain a scheme list; extracting a plurality of target maintenance schemes from the scheme list based on a preset number value. 3.The method of claim 2, wherein, the matching of the corresponding target personnel profile from the preset personnel database according to the maintenance level comprises: add the maintenance task order to a preset task list, adjust an order of the maintenance task order in the preset task list according to the maintenance level, and determine a maintenance time node after the adjustment; when the maintenance time node is reached, determine a text matching degree between the scheme text of the maintenance task order and each personnel portrait contained in a preset personnel database; determine a target personnel portrait according to the size of the text matching degree. 4.The method of claim 1, wherein, After the step of determining whether the fault risk value is greater than the preset risk value, the method further comprises: counting an idle duration of the unmanned aerial vehicle parked on the storage platform; updating a push list of the unmanned aerial vehicle according to the size of the idle duration; displaying the push list on a preset screen for reference by the technician.

5. A maintenance work order distribution device for unmanned aerial vehicles (UAVs), characterized in that, The device comprises: an identification module configured to call a preset image recognition model to identify image features of real-time images of the unmanned aerial vehicle, and to calculate a fault risk value of the unmanned aerial vehicle based on the image features, wherein the real-time images are images captured from different angles after the unmanned aerial vehicle completes a field operation and is stored on the storage platform; a calculation module configured to, when the fault risk value is greater than a preset risk value, obtain historical operation information of the unmanned aerial vehicle, determine a maintenance score value and a depreciation score value based on the historical operation information, and calculate a fault comprehensive score value using the maintenance score value, the depreciation score value, and the fault risk value; a construction module configured to determine a maintenance level based on the fault comprehensive score value, call a preset Attention algorithm model to filter a plurality of target maintenance schemes from a scheme database corresponding to the maintenance level according to the image features, and construct a maintenance task order using the plurality of target maintenance schemes; an allocation module configured to match a target personnel portrait corresponding to the maintenance level from a preset personnel database, and allocate the maintenance task order to a maintenance technician corresponding to the target personnel portrait; the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the fault risk value of the unmanned aerial vehicle based on the image features comprises: calling the preset image recognition model to identify chroma values of component regions corresponding to each of the real-time images to obtain first image features; performing image cropping on each of the real-time images based on the first image features to obtain component images; identifying and extracting edge lines of component outlines of each of the component images using an edge detection algorithm to obtain component outline lines; determining a similarity degree between each of the component outline lines and a preset line to obtain line similarity degrees, and calculating the fault risk value using the line similarity degrees; or; the calling of the preset image recognition model to identify the image features of the real-time images and the calculation of the fault risk value of the unmanned aerial vehicle based on the image features comprises: calling the preset image recognition model to identify cracks in the real-time images according to image pixels to obtain second image features; determining an area and a number of damaged regions corresponding to each of the real-time images based on the second image features to obtain area parameters and number parameters; calculating the fault risk value using the area parameters and the number parameters.

6. A maintenance work order assignment system for unmanned aerial vehicles, the system comprising: The system comprises a storage platform and a management platform, and the management platform is suitable for the maintenance work order distribution method of the unmanned aerial vehicle according to any one of claims 1-4. The storage platform is provided with a plurality of cameras, the storage platform is used for placing the unmanned aerial vehicle, the plurality of cameras are distributed at different positions of the storage platform to shoot real-time images of the unmanned aerial vehicle from different angles, and the management platform is connected with the plurality of cameras.

7. An electronic device comprising: The memory, the processor and the computer program stored in the memory and executable on the processor are characterized in that the processor executes the computer program to realize the maintenance work order distribution method of the unmanned aerial vehicle according to any one of claims 1-4.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer executable program, and the computer executable program is used for enabling the computer to execute the maintenance work order distribution method of the unmanned aerial vehicle according to any one of claims 1-4.

Images