Unmanned aerial vehicle inspection picture defect identification method, device and equipment and storage medium

Abstract

The application relates to an unmanned aerial vehicle inspection picture defect identification method, device and equipment and a storage medium, wherein the method comprises the following steps: acquiring line image information; generating component model information according to the line image information, comparing the component model information with standard model information, obtaining a comparison result between the component model information and the standard model information and setting the comparison result as component fault information; generating fine model information according to the line image information, comparing the fine model information with the standard model information; obtaining the comparison result between the component model information and the standard model information and setting the comparison result as fine fault information; and sending prompt information related to the component fault information and the fine fault information to an intelligent terminal of a worker. The application has the technical effect that the possibility that concentrated distribution faults cannot be accurately identified due to too high resolution of a photographed photo during unmanned aerial vehicle shooting is reduced, and the accuracy of unmanned aerial vehicle identification fault information is improved.

Description

Technical Field

[0001] This application relates to the technical field of inspection defect identification, and in particular to methods, devices, equipment and storage media for defect identification of UAV inspection images. Background Technology

[0002] In order to understand the operating status of the lines and promptly detect equipment defects and threats to line safety, it is necessary to conduct inspections of the transmission lines. Since the transmission lines are widely distributed and operate in the open for a long time, they are often affected by the surrounding environment.

[0003] Because power distribution lines have more and more complex equipment than power transmission lines, drones are now commonly used to inspect power transmission lines in order to reduce labor costs and improve inspection efficiency. The drones take pictures of the power transmission lines and compare the pictures with preset standard pictures to determine whether the power transmission lines are operating normally.

[0004] During the aforementioned drone inspection process, the inventors believe that at least the following problems exist: when the drone takes pictures, the distribution of line defects may be relatively concentrated, and the high resolution of the pictures makes it impossible to accurately identify line defects. Summary of the Invention

[0005] To address the issue of viewers only seeing edited videos, resulting in a poor viewing experience, this application provides a method and system for identifying defects in drone inspection images.

[0006] In a first aspect, this application provides a method for defect identification in UAV inspection images, which adopts the following technical solution: the method includes: acquiring line image information;

[0007] Component model information is generated based on line image information, wherein the component model information is the model information corresponding to the line image information;

[0008] The component model information is compared with the corresponding standard model information in the preset model information library, which stores different standard model information.

[0009] Obtain the comparison results between the component model information and the standard model information, and set the comparison results as component fault information;

[0010] Detailed model information is generated based on the line image information, wherein the detailed model information is the model information of the small parts corresponding to the component model information;

[0011] The detailed model information is compared with the corresponding standard model information in the preset model information library;

[0012] Obtain the comparison results between the component model information and the standard model information, and set the comparison results as minor fault information;

[0013] Send alerts related to component failure and minor fault information to staff's smart terminals.

[0014] The above technical solution first acquires line image information, then generates corresponding component model information based on the line image information. By comparing the component model information with standard model information, it is determined whether the component model information is abnormal. This reduces the possibility that unclear image information may prevent the drone from accurately identifying component fault information, thus improving the accuracy of the drone in identifying line faults. After identifying the component model, it generates small model information based on the line image. By comparing the small model with the standard model, it accurately identifies fault information appearing on the small model. This reduces the possibility that high-resolution photos taken by the drone may prevent the accurate identification of concentrated faults, thus improving the accuracy of the drone in identifying fault information.

[0015] In one specific implementation scheme, generating component model information based on line image information includes:

[0016] Obtain the line resolution corresponding to the line image information;

[0017] Obtain the corresponding adjusted image information according to the preset adjustment resolution;

[0018] Generate adjustment component model information based on the adjustment image information;

[0019] Generate model information of line components based on line image information;

[0020] Component model information is generated based on the adjustment component model information and the circuit component model information.

[0021] Through the above technical solution, different adjustment image information is generated according to the preset adjustment resolution, and different component model information is generated based on the adjustment image information of different resolutions. This allows the system to obtain component model information at different resolutions, and enables the system to generate the clearest model information corresponding to different resolutions, thereby improving the clarity of the component model.

[0022] In one specific implementation, sending alerts related to component failure information and minor fault information to the worker's smart terminal includes:

[0023] A corresponding fault confidence score is generated based on the fault information. The fault confidence score is the degree of credibility of the corresponding fault information. The fault information includes component fault information and minor fault information.

[0024] The fault confidence level is compared with the preset standard confidence level;

[0025] If the fault confidence level is less than the preset standard confidence level, then the sending of fault-related prompts to the staff's smart terminals will be cancelled.

[0026] Otherwise, send alerts related to component failure and minor fault information to the staff's smart terminal.

[0027] The above technical solution generates corresponding fault information through model comparison, and then generates a fault confidence level based on the fault information. By comparing the fault confidence level with the standard confidence level, the corresponding fault prompt information is triggered only when the fault confidence level is not less than the preset standard confidence level. This reduces the possibility of generating the corresponding fault information incorrectly, thereby improving the reliability of the fault information triggering.

[0028] In one specific implementation, sending prompts related to component failure information and minor fault information to the worker's smart terminal includes:

[0029] Summarize component fault information and minor fault information, and set the summarized component fault information and minor fault information as fault information;

[0030] Obtain the corresponding fault overlap based on different fault information;

[0031] The fault overlap is compared with the preset standard fault value;

[0032] If the fault overlap is greater than the preset standard fault value, then the corresponding fault information is obtained;

[0033] The fault information is merged until the overlap between different fault information is no greater than the preset standard fault value.

[0034] Otherwise, send alerts related to component failure and minor fault information to the staff's smart terminal.

[0035] Through the above technical solution, after the fault information has been tested for confidence, the degree of fault overlap corresponding to different fault information can be detected, which reduces the possibility of repeatedly recording fault information in different image information, so that different fault information corresponds to different fault problems, thereby improving the accuracy of the fault problem corresponding to each fault information.

[0036] In one specific implementation scheme, generating model information based on line image information includes:

[0037] Obtain line image information;

[0038] Obtain the component model information and detailed model information corresponding to the line image information respectively;

[0039] The number of component model information corresponding to the line image information is counted, and the number of component model information is set as the number of component models generated.

[0040] The number of fine model information corresponding to the line image information is counted, and the number of fine model information is set as the number of fine model generation;

[0041] Calculate the difference between the number of component models generated and the number of detailed models generated, and set the difference as the model generation difference;

[0042] If the model generation difference is outside the preset standard model difference range, then supplement the model with the corresponding number of models until the model generation difference after the supplementation is within the range of the standard model difference.

[0043] If the model generation difference is within the preset standard model difference range, then model information is generated based on the line image information.

[0044] By employing the above technical solution, when generating model information based on line image information, the number of generated component model information and fine model information is counted separately, ensuring that the number of corresponding models generated is in a balanced state. This reduces the possibility that different training amounts may lead to different accuracy levels in the generated component models and fine models when the training system generates corresponding models based on line images, thereby improving the stability of the system in generating model information based on line images.

[0045] In one specific implementation scheme, after generating component model information based on the adjustment component model information and the circuit component model information, the method further includes:

[0046] The adjustment component model information is compared with the line component model information;

[0047] If the adjustment component model information is different from the line component model information, then obtain the corresponding line resolution and adjustment resolution;

[0048] Obtain the standard resolution range corresponding to the differences in model information from the preset standard resolution library. The standard resolution library stores different parts of the model information and the standard resolution range of the corresponding parts.

[0049] Compare the line resolution and the adjusted resolution with the standard resolution range respectively;

[0050] If the line resolution is within the standard resolution range, then generate a model that differs from the standard resolution.

[0051] If the adjusted resolution is within the standard resolution range, then a model is generated based on the adjusted image information corresponding to the adjusted resolution to differentiate the features.

[0052] Through the above technical solution, when the model information of the adjustment component differs from that of the circuit component, the corresponding circuit resolution and adjustment resolution are obtained. By comparing the circuit resolution, adjustment resolution, and standard resolution respectively, the resolution that is closer to the standard resolution is selected. Based on the image information corresponding to the closer resolution, model information for the differences is generated. This allows the system to automatically select the corresponding image information to generate the corresponding model information based on the degree of resolution matching, thereby improving the accuracy of model information generation.

[0053] In one specific implementation, the method further includes:

[0054] If both the adjustable resolution and the line resolution are within the preset standard resolution range, then the adjustment matching degree and the line matching degree are calculated respectively. The adjustment matching degree is the degree of matching between the adjustable resolution and the standard resolution, and the line matching degree is the degree of matching between the line resolution and the standard resolution.

[0055] The model is generated by selecting the image information corresponding to the larger value of the adjustment matching degree and the line matching degree to identify the differences.

[0056] If both the adjusted resolution and the line resolution are outside the preset standard resolution range, the image information that is closer to the standard resolution range between the adjusted resolution and the line resolution will be used to generate the model that differs from the one that is closer to the standard resolution range.

[0057] Through the above technical solution, when both the line resolution and the adjustment resolution are within the standard resolution range, a resolution that is more in line with the standard resolution range is selected to generate model information that differs from the standard resolution range; when both the line resolution and the adjustment resolution are outside the standard resolution range, a resolution that is closer to the standard resolution range is selected to generate model information that differs from the standard resolution range, thereby further improving the accuracy of the generated model information.

[0058] Secondly, this application provides a device for identifying defects in images from unmanned aerial vehicle (UAV) inspections, employing the following technical solution: The device includes:

[0059] The line image acquisition module is used to acquire line image information;

[0060] The component model generation module is used to generate component model information based on the line image information, wherein the component model information is the model information corresponding to the line image information;

[0061] The component model comparison module is used to compare the component model information with the corresponding standard model information in the preset model information library. The model information library stores different standard model information.

[0062] The component fault acquisition module is used to acquire the comparison results between component model information and standard model information and set the comparison results as component fault information.

[0063] The fine model generation module is used to generate fine model information based on the line image information. The fine model information is the model information of the fine parts corresponding to the component model information.

[0064] The small model comparison module is used to compare the small model information with the corresponding standard model information in the preset model information database;

[0065] The minor fault acquisition module is used to acquire the comparison results between component model information and standard model information and set the comparison results as minor fault information.

[0066] The fault information sending module is used to send prompts related to component fault information and minor fault information to the staff's smart terminal.

[0067] Thirdly, this application provides a computer device that adopts the following technical solution: it includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as any of the above-described UAV inspection image defect recognition methods.

[0068] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: storing a computer program that can be loaded by a processor and executed by any of the above-mentioned UAV inspection image defect recognition methods.

[0069] In summary, this application includes at least one of the following beneficial technical effects:

[0070] 1. By first acquiring line image information, generating corresponding component model information based on the line image information, and comparing the component model information with standard model information to determine whether the component model information is abnormal, the possibility of the drone being unable to accurately identify component fault information due to unclear image information is reduced, thereby improving the accuracy of the drone in identifying line faults. After identifying the component model, small model information is generated based on the line image recognition, and the fault information appearing on the small model is accurately identified by comparing the small model with the standard model. This reduces the possibility of concentrated faults not being accurately identified due to high resolution photos taken by the drone, thereby improving the accuracy of the drone in identifying fault information.

[0071] 2. Generate different adjustment image information according to the preset adjustment resolution, and generate different component model information based on the adjustment image information of different resolutions. This allows the system to obtain component model information at different resolutions, enabling the system to generate the clearest model information corresponding to each resolution, thereby improving the clarity of the component model. Attached Figure Description

[0072] Figure 1 This is a flowchart of the method for identifying defects in drone inspection images in the embodiments of this application.

[0073] Figure 2 This is a structural block diagram of the drone inspection image defect recognition device in the embodiments of this application.

[0074] Attached reference numerals: 201, Line image acquisition module; 202, Component model generation module; 203, Component model comparison module; 204, Component fault acquisition module; 205, Small model generation module; 206, Small model comparison module; 207, Small fault acquisition module; 208, Fault information sending module. Detailed Implementation

[0075] The following is in conjunction with the appendix Figure 1-2 This application will be described in further detail.

[0076] This application discloses a method for defect recognition in UAV inspection images. The method is based on a UAV inspection recognition system. The UAV is equipped with a camera device that captures images of the inspection route from a preset angle and sends the image information to the recognition system. The recognition system analyzes and compares the image information according to preset standards and finally summarizes the fault information on the inspection route into a single image. The control system then sends the summarized image information to the staff's smart terminal.

[0077] like Figure 1 As shown, the method includes the following steps:

[0078] S10, acquire line image information.

[0079] The route image information is obtained by the drone flying along a preset route while simultaneously controlling the camera device to take pictures of the preset route and sending the captured image information to the recognition system.

[0080] S11, Generate component model information based on line image information.

[0081] Among them, the recognition system is trained using deep learning methods, so that the recognition system can generate corresponding component model information based on the line image information.

[0082] S12, compare the component model information with the preset standard model information;

[0083] The system stores various standard model information in the preset model information library. The system queries the standard model corresponding to the generated component model in the model information library and determines whether the inspection route is normal by comparing the component model with the standard model.

[0084] S13, obtain component fault information.

[0085] The comparison results between the component model information and the standard model information are obtained and the comparison results are set as component fault information. The comparison results can indicate that the component model is normal or that the corresponding component in the component model is damaged, such as loose strands or broken strands in the wire.

[0086] S14, Obtain detailed model information.

[0087] After comparing the component models, the component model information is detected. The preset component detail library stores the detailed model information corresponding to different components. If the component model information contains a detailed model, the component model is subjected to the identification of small target defects. By changing the resolution of the line image information, the corresponding small model information is generated again based on the line image information.

[0088] S15, obtain minor fault information.

[0089] This involves obtaining the comparison results between the information of the small model and the information of the standard model, and setting the comparison results as small fault information. The comparison results can be either that the small model is normal or that there are fault information such as missing pins or missing nuts in the small model.

[0090] S16, send a prompt message related to the fault information to the staff's smart terminal.

[0091] After the system completes the comparison and obtains component fault information and minor fault information, it first summarizes the component fault information and minor fault information and generates corresponding line image information. By marking and annotating the corresponding fault information content in the corresponding position of the line image information, the line image information with standard and annotation is finally sent to the staff's smart terminal.

[0092] In one embodiment, considering that the images generated by the drone after shooting are usually of the same resolution, and the corresponding line image at the current resolution may be blurry, it is necessary to shoot images of different resolutions during the shooting process; the specific shooting operation can be performed as follows:

[0093] The process involves acquiring the line resolution corresponding to the line image information, obtaining the corresponding adjustment image information according to a preset adjustment resolution, generating adjustment component model information based on the adjustment image information, generating line component model information based on the line image information, and generating component model information based on the adjustment component model information and the line component model information. This allows the drone to capture images at different resolutions when shooting the same location, selecting the most suitable resolution to generate the corresponding model information, thereby improving the accuracy of model information generation. For example, if the existing line resolution is X, the corresponding line model X1 is generated based on line resolution X. If the adjustment resolutions are Y, U, and I, then the adjustment model information generated based on adjustment resolution Y is Y1, the adjustment model information generated based on adjustment resolution U is U1, and the adjustment model information generated based on adjustment resolution I is I1. Finally, component model information is generated using X1, Y1, U1, and I1.

[0094] In one embodiment, considering that the credibility of different fault information generation methods varies, it is necessary to evaluate the credibility of the fault information. The specific evaluation steps can be performed as follows:

[0095] The system generates corresponding fault confidence levels based on fault information, which represents the degree to which the fault information is plausible. Fault information includes component fault information and minor fault information. The system compares the fault confidence levels with preset standard confidence levels. If the fault confidence level is lower than the preset standard confidence level, the system cancels sending fault-related prompts to the staff's smart terminals; otherwise, it sends prompts related to component fault information and minor fault information to the staff's smart terminals. This allows the system to automatically filter fault information that is unlikely to occur, thereby improving the reliability of the fault information. For example, if the component fault information generated through comparison is foreign object on the tower, tower corrosion, and suspension clamp misalignment, with corresponding confidence levels of 0.5, 0.7, and 0.3 respectively, and the corresponding configured standard confidence levels are 0.4, 0.5, and 0.6, then the component fault information for foreign object on the tower and tower corrosion is retained; the component fault information for suspension clamp misalignment is not retained because its confidence level is lower than the standard confidence level.

[0096] In one embodiment, considering that different obstacle information may be repeatedly identified, it is necessary to identify the overlap of obstacle information. The specific identification operation can be performed as follows:

[0097] The system summarizes component fault information and minor fault information, and sets the summarized component fault information and minor fault information as fault information. It then obtains the corresponding fault overlap degree based on different fault information; compares the fault overlap degree with a preset standard fault value; if the fault overlap degree is greater than the preset standard fault value, the corresponding fault information is obtained; the system obtains the confidence level of the fault information with a fault overlap degree exceeding the preset value, and deletes fault information with lower confidence levels until the fault overlap between different fault information is no greater than the preset standard fault value; otherwise, it sends a prompt message related to the component fault information and minor fault information to the staff's smart terminal. This allows the system to automatically filter repeatedly identified fault information, thereby improving the accuracy of fault information. For example, after summarizing the identified fault information, two fault information related to foreign objects on poles are identified, with corresponding confidence levels of 0.55 and 0.5. After detection, the overlap degree of the foreign objects on the poles is found to be 0.55, which is higher than the preset standard fault value of 0.5. Therefore, the fault information with a confidence level of 0.5 is deleted.

[0098] In one embodiment, considering that the system generates corresponding model information from line image information through a large amount of data for training, it is necessary to perform a statistical operation on the number of training samples generated for different model information during the training process. The specific statistical operation can be performed as follows:

[0099] The system acquires line image information; acquires component model information and detailed model information corresponding to the line image information respectively; counts the number of component model information corresponding to the line image information and sets the number of component model information as the component model generation quantity; counts the number of detailed model information corresponding to the line image information and sets the number of detailed model information as the detailed model generation quantity; calculates the difference between the number of component model generation and the number of detailed model generation and sets the difference as the model generation difference; if the model generation difference is outside the preset standard model difference range, the corresponding number of models is supplemented until the model generation difference after supplementation is within the range of the standard model difference; if the model generation difference is within the preset standard model difference range, model information is generated based on the line image information; this allows the system to automatically adjust and balance the training amount of the model during the process of generating model information based on line image information through deep learning training, so that the system has the same ability to generate different model information based on line image information, thereby improving the accuracy of the model information.

[0100] In one embodiment, considering that the model information generated from image information of different resolutions is also different, it is necessary to optimize the generated model information. The specific optimization operation can be performed as follows:

[0101] The system compares the adjustment component model information with the line component model information. If the adjustment component model information differs from the line component model information, the corresponding line resolution and adjustment resolution are obtained. The system retrieves the standard resolution range corresponding to the differences in model information from a preset standard resolution library, which stores different parts of the model information and their corresponding standard resolution ranges. The line resolution and adjustment resolution are then compared with the standard resolution ranges. If the line resolution is within the standard resolution range, a model of the differences is generated according to the line resolution. If the adjustment resolution is within the standard resolution range, a model of the differences is generated according to the adjustment image information corresponding to the adjustment resolution. This allows the system to automatically select the image information corresponding to a resolution that better matches the standard resolution range to generate the corresponding model information based on the relationship between the line resolution, adjustment resolution, and standard resolution range, thereby improving the accuracy of the model information.

[0102] In one embodiment, considering the possibility that the adjustment resolution and line resolution may be close, it is necessary to further restrict the model information in special cases. The specific restriction operation can be performed as follows:

[0103] If both the adjustable resolution and the line resolution are within the preset standard resolution range, the adjustment matching degree and the line matching degree are calculated separately. The adjustment matching degree is the degree of matching between the adjusted resolution and the standard resolution, and the line matching degree is the degree of matching between the line resolution and the standard resolution. The image information corresponding to the larger value of the adjustment matching degree and the line matching degree is selected to generate a model showing the differences. If both the adjustable resolution and the line resolution are outside the preset standard resolution range, the image information that is closer to the standard resolution range is selected to generate a model showing the differences. This allows the system to select a more suitable resolution to generate the corresponding model information, so that even when both the adjustable resolution and the line resolution are special, the system can still generate relatively accurate model information.

[0104] Based on the above method, this application also discloses a device for identifying defects in images from drone inspections.

[0105] like Figure 2 As shown, the device includes the following modules:

[0106] Line image acquisition module 201 is used to acquire line image information;

[0107] The component model generation module 202 is used to generate component model information based on the line image information, wherein the component model information is the model information corresponding to the line image information.

[0108] The component model comparison module 203 is used to compare the component model information with the corresponding standard model information in the preset model information library, which stores different standard model information.

[0109] The component fault acquisition module 204 is used to acquire the comparison result between component model information and standard model information and set the comparison result as component fault information;

[0110] The fine model generation module 205 is used to generate fine model information based on the line image information, wherein the fine model information is the model information of the fine parts corresponding to the component model information;

[0111] The small model comparison module 206 is used to compare the small model information with the corresponding standard model information in the preset model information database;

[0112] The minor fault acquisition module 207 is used to acquire the comparison result between the component model information and the standard model information and set the comparison result as minor fault information.

[0113] The fault information sending module 208 is used to send prompts related to component fault information and minor fault information to the staff's smart terminal.

[0114] In one embodiment, the component model generation module 202 is further configured to generate component model information based on line image information, including: obtaining the line resolution corresponding to the line image information; obtaining the corresponding adjustment image information according to a preset adjustment resolution; generating adjustment component model information based on the adjustment image information; generating line component model information based on the line image information; and generating component model information based on the adjustment component model information and the line component model information.

[0115] In one embodiment, the component model generation module 202 is further configured to send prompts related to component fault information and minor fault information to the worker's smart terminal, including: generating a corresponding fault confidence level based on the fault information, wherein the fault confidence level is the degree of credibility of the corresponding fault information, and the fault information includes component fault information and minor fault information; comparing the fault confidence level with a preset standard confidence level; if the fault confidence level is less than the preset standard confidence level, canceling the sending of prompts related to the fault information to the worker's smart terminal; otherwise, sending prompts related to component fault information and minor fault information to the worker's smart terminal.

[0116] In one embodiment, the component model generation module 202 is further configured to send prompts related to component fault information and minor fault information to the worker's smart terminal, including: summarizing the component fault information and minor fault information, and setting the summarized component fault information and minor fault information as fault information; obtaining the corresponding fault overlap degree according to different fault information; comparing the fault overlap degree with a preset standard fault value; if the fault overlap degree is greater than the preset standard fault value, obtaining the corresponding fault information; merging the fault information until the fault overlap degree between different fault information is not greater than the preset standard fault value; otherwise, sending prompts related to component fault information and minor fault information to the worker's smart terminal.

[0117] In one embodiment, the component model generation module 202 is further configured to generate model information based on line image information, including: acquiring line image information; acquiring component model information and detailed model information corresponding to the line image information respectively; counting the number of component model information corresponding to the line image information and setting the number of component model information as the component model generation quantity; counting the number of detailed model information corresponding to the line image information and setting the number of detailed model information as the detailed model generation quantity; calculating the difference between the component model generation quantity and the detailed model generation quantity and setting the difference as the model generation difference; if the model generation difference is outside the preset standard model difference range, then supplementing the corresponding number of models until the model generation difference after supplementation is within the range of the standard model difference; if the model generation difference is within the preset standard model difference range, then generating model information based on the line image information.

[0118] In one embodiment, the component model generation module 202 is further configured to, after generating component model information based on the adjustment component model information and the line component model information, further include: comparing the adjustment component model information with the line component model information; if the adjustment component model information and the line component model information are different, obtaining the corresponding line resolution and adjustment resolution; obtaining the standard resolution range corresponding to the differences in model information from a preset standard resolution library, the standard resolution library storing different parts of model information and the standard resolution range of the corresponding parts; comparing the line resolution and adjustment resolution with the standard resolution range respectively; if the line resolution is within the standard resolution range, generating the model of the differences according to the line resolution; if the adjustment resolution is within the standard resolution range, generating the model of the differences according to the adjustment image information corresponding to the adjustment resolution.

[0119] In one embodiment, the component model generation module 202 is further configured to: if both the adjustment resolution and the line resolution are within a preset standard resolution range, calculate the adjustment matching degree and the line matching degree respectively, wherein the adjustment matching degree is the degree of matching between the adjustment resolution and the standard resolution, and the line matching degree is the degree of matching between the line resolution and the standard resolution; select the image information corresponding to the larger value of the adjustment matching degree and the line matching degree to generate a model showing the differences; if both the adjustment resolution and the line resolution are outside the preset standard resolution range, select the image information of the adjustment resolution and the line resolution that is closer to the standard resolution range to generate a model showing the differences.

[0120] This application also discloses a computer device.

[0121] Specifically, the computer device includes a memory and a processor, with the memory storing a computer program that can be loaded by the processor and executed to perform the aforementioned method for identifying defects in UAV inspection images.

[0122] This application also discloses a computer-readable storage medium.

[0123] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the aforementioned UAV inspection image defect identification method. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for defect identification in drone inspection images, characterized in that, The method comprises: Obtaining line image information; According to the deep learning method, the recognition system is trained, and the component model information is generated according to the line image information, and the component model information is the model information corresponding to the line image information; Compare the component model information with the corresponding standard model information in the preset model information library, and the model information library stores different standard model information; Obtain the comparison result between the component model information and the standard model information and set the comparison result as the component fault information; Detect the component model information, if the component model information contains detailed model, identify the small target defect of the component model, generate small model information according to the line image information by changing the resolution of the line image information, and the small model information is the model information of the small part corresponding to the component model information; Compare the small model information with the corresponding standard model information in the preset model information library; Obtain the comparison result between the component model information and the standard model information and set the comparison result as the component fault information; Send the prompt information related to the component fault information and the small fault information to the intelligent terminal of the worker; The generation of component model information according to line image information comprises: Obtain the line resolution corresponding to the line image information; Obtain the corresponding adjustment image information according to the preset adjustment resolution; Generate adjustment component model information according to the adjustment image information; Generate line component model information according to line image information; Generate component model information according to adjustment component model information and line component model information; The sending of prompt information related to component fault information and small fault information to the intelligent terminal of the worker comprises: Generate the corresponding fault confidence according to the fault information, the fault confidence is the confidence degree of the occurrence of the corresponding fault information, and the fault information includes component fault information and small fault information; Compare the fault confidence with the preset standard confidence; If the fault confidence is less than the preset standard confidence, cancel the sending of prompt information related to fault information to the intelligent terminal of the worker; Otherwise, send the prompt information related to the component fault information and the small fault information to the intelligent terminal of the worker; The sending of prompt information related to component fault information and small fault information to the intelligent terminal of the worker comprises: Summarize the component fault information and the small fault information, and set the summarized component fault information and the small fault information as fault information; Obtain the corresponding fault overlap according to different fault information; Compare the fault overlap with the preset standard fault value; If the fault overlap is greater than the preset standard fault value, obtain the corresponding fault information; Merge the fault information until the fault overlap between different fault information is not greater than the preset standard fault value; Otherwise, send the prompt information related to the component fault information and the small fault information to the intelligent terminal of the worker; The generation of model information according to line image information comprises: Obtain the line image information; Obtain the component model information and the small model information corresponding to the line image information respectively; counting the number of component model information corresponding to the line image information and setting the number of component model information as a component model generation number; counting the number of fine model information corresponding to the line image information and setting the number of fine model information as a fine model generation number; calculating the difference between the component model generation number and the fine model generation number and setting the difference as a model generation difference; if the model generation difference is outside the preset standard model difference range, supplementing the corresponding model number until the model generation difference after the supplement is within the standard model difference range; if the model generation difference is within the preset standard model difference range, generating model information according to the line image information.

2. The method of claim 1, wherein, After the component model information is generated according to the adjusted component model information and the line component model information, the method further comprises: comparing the adjusted component model information with the line component model information; if the adjusted component model information is different from the line component model information, obtaining the corresponding line resolution and the adjusted resolution; obtaining the standard resolution range corresponding to the difference in the model information in the preset standard resolution library, the standard resolution library storing different parts of the model information and the standard resolution range corresponding to the parts; comparing the line resolution and the adjusted resolution with the standard resolution range, respectively; if the line resolution is within the standard resolution range, generating the model of the difference according to the line resolution; if the adjusted resolution is within the standard resolution range, generating the model of the difference according to the adjusted image information corresponding to the adjusted resolution.

3. The method of claim 2, wherein, The method further comprises: if the adjusted resolution and the line resolution are both within the preset standard resolution range, calculating the adjusted matching degree and the line matching degree, respectively, the adjusted matching degree being the matching degree between the adjusted resolution and the standard resolution, and the line matching degree being the matching degree between the line resolution and the standard resolution; selecting the picture information corresponding to the greater value of the adjusted matching degree and the line matching degree to generate the model of the difference; if the adjusted resolution and the line resolution are both outside the preset standard resolution range, selecting the picture information closer to the standard resolution range between the adjusted resolution and the line resolution to generate the model of the difference.

4. An unmanned aerial vehicle inspection picture defect identification device applied to the unmanned aerial vehicle inspection picture defect identification method of any one of claims 1-3, characterized in that, The device comprises: a line image acquisition module (201) configured to acquire line image information; a component model generation module (202) configured to generate component model information according to the line image information, the component model information being model information corresponding to the line image information; a component model comparison module (203) configured to compare the component model information with corresponding standard model information in a preset model information library, the model information library storing different standard model information; a component fault acquisition module (204) configured to obtain the comparison result between the component model information and the standard model information and set the comparison result as component fault information; a fine model generation module (205) configured to generate fine model information according to the line image information, the fine model information being model information of fine parts corresponding to the component model information; The fine model comparison module (206) is configured to compare the fine model information with corresponding standard model information in a preset model information library; The fine fault acquisition module (207) is configured to acquire a comparison result between the component model information and the standard model information and set the comparison result as fine fault information; The fault information sending module (208) is configured to send prompt information related to the component fault information and the fine fault information to a smart terminal of a worker.

5. A computer device, comprising: A memory and a processor are included, and the memory has stored thereon a computer program capable of being loaded and executed by the processor to perform any one of the methods in claims 1-3.

6. A computer-readable storage medium, characterized in that, A memory has stored thereon a computer program capable of being loaded and executed by the processor to perform any one of the methods in claims 1-3.

Images