Electric power inspection method based on AI identification and AR and AR glasses

Abstract

The invention relates to an electric power inspection method based on AI identification and AR and AR glasses, and the method comprises the following steps: obtaining historical inspection image data of a target electric power inspection scene, and carrying out the preprocessing of the historical inspection image data to form a training sample set; establishing a neural network model based on a self-attention mechanism, inputting the training sample set into the neural network model for training, and when the maximum number of iterations of training is reached or the recognition precision of the neural network model reaches a preset value, ending the training to obtain a trained electric power inspection scene defect recognition model; configuring the trained electric power inspection scene defect identification model on wearable AR equipment, acquiring image data of a current scene through the wearable AR equipment in an inspection process of a target electric power circuit, identifying whether an electric power inspection scene in the current scene has defects or not, and if yes, sending the defect identification model to the wearable AR equipment; and generating an alarm identifier, and displaying the alarm identifier and the defect occurrence position in a display picture of the wearable AR equipment.

Description

Technical Field

[0001] The present invention relates to an electric power inspection method based on AI recognition and AR and AR glasses, belonging to the technical field of electric power inspection. Background Art

[0002] At present, power line inspection basically relies on manual inspection, which is difficult to find all defects in the scene. If it is not discovered in time, it may easily cause the power equipment to fail to operate normally, or even cause failures or even fires. Therefore, an intelligent power inspection solution is urgently needed. Summary of the invention

[0003] In order to solve the problems existing in the above-mentioned prior art, the present invention proposes an electric power inspection method based on AI recognition and AR and AR glasses.

[0004] The technical solution of the present invention is as follows:

[0005] In one aspect, the present invention provides a power inspection method based on AI recognition and AR, comprising the following steps:

[0006] Acquire historical inspection image data of the target power inspection scene, including defect-free images and defective images, and preprocess the historical inspection image data to form a training sample set;

[0007] Establish a neural network model based on the self-attention mechanism, initialize the neural network model using transfer learning, and input the training sample set into the initialized neural network model for training;

[0008] During the training process, the neural network model is optimized by using feedback optimization. When the maximum number of training iterations is reached or the recognition accuracy of the neural network model reaches a preset value, the training is terminated to obtain a trained power inspection scene defect recognition model.

[0009] The trained power inspection scene defect recognition model is configured on the wearable AR device. During the inspection of the target power line, the image data of the current scene is obtained through the wearable AR device and input into the trained power inspection scene defect recognition model to identify whether there are defects in the power inspection scene in the current scene. When defects exist, an alarm mark is generated and the alarm mark is bound to the location where the defect occurs to generate an AR label model, which is displayed on the display screen of the wearable AR device.

[0010] As a preferred implementation, the step of preprocessing the image data includes image normalization processing, specifically:

[0011] The collected image data is converted from RGB image to grayscale image, and the following formula is used:

[0012]

[0013] V (x,y) =MAX;

[0014] Among them, L (x,y) is the gray value of the (x, y) coordinate after conversion to a grayscale image, δ is the correction coefficient, α represents the standard spatial distance, and H s is the hue value of sampling point s, S s is the saturation value of sampling point s, V s is the brightness value of sampling point s, H (x,y) is the hue value of the (x, y) coordinate in the original image, S (x,y) is the saturation value of the (x, y) coordinate in the original image, V (x,y) is the brightness value of the (x, y) coordinate in the original image; R, G, B are the red value, green value and blue value of the corresponding pixel respectively, mod6 is the modulus operation; θ = MAX-MIN; MAX is the maximum value among R, G, B, and MIN is the minimum value among R, G, B;

[0015] The converted grayscale image is normalized as follows:

[0016]

[0017] Among them, L (x,y) ′ represents the grayscale value of the normalized grayscale image at the (x, y) coordinate, L max Represents the maximum grayscale value of all coordinates in the original grayscale image, L min Represents the minimum grayscale value among all coordinates in the original grayscale image.

[0018] As a preferred implementation, the step of preprocessing the image data further includes a sample multiplication step, specifically:

[0019] Determine the defect types of all power inspection scenarios, and establish a corresponding number of GAN generation networks according to the number of defect types in the power inspection scenarios, wherein the GAN generation network includes a generator and a discriminator, wherein the generator generates a new defect image according to random noise, and the discriminator is used to identify the similarity between the new defect image and the original defect image;

[0020] Each GAN generation network is used to generate defect images of the corresponding defect type, so that the difference in the number of defect images of each defect type is less than a preset range, and the difference between the number of defect-free images and the total number of defect images is less than a preset range.

[0021] As a preferred implementation, the self-attention mechanism cuts the input image data into n image blocks, and obtains the average grayscale value L1, L2, L3, ... L of each image block. n , and convert the average gray value of each image block into a feature vector through a fully connected layer;

[0022] The self-attention mechanism calculates the attention weight of each image block by using query vector, key vector and value vector, where the query vector corresponds to the average grayscale value of the image block, and both the key vector and the value vector use feature vectors.

[0023] On the other hand, the present invention also provides a smart AR glasses for power inspection based on AI recognition, including an AR glasses body and a processor;

[0024] The processor is used to execute the power inspection method based on AI recognition and AR as described in any embodiment of the present invention;

[0025] The AR glasses also include a head, a first temple, a second temple, a speaker, a microphone, a fixed card block, a protective lens and a main camera. The heads are installed on both sides of the AR glasses body, the first temple is rotatably installed on the head on one side, and the second temple is rotatably installed on the head on the side away from the first temple. The speaker and the microphone are both installed on the first temple, the fixed card block is installed on the AR glasses body, the protective lens is installed on the fixed card block, and the main camera is installed on one side of the AR glasses body.

[0026] As a preferred embodiment, main lenses are installed on both sides of the AR glasses body, and the main camera is installed between the two main lenses.

[0027] As a preferred embodiment, a network communication module is installed on the side of the AR glasses body away from the main camera, and the network communication module is installed between the two main body lenses.

[0028] As a preferred embodiment, infrared cameras are installed on both sides of the AR glasses body close to the main camera, and the infrared cameras are symmetrically installed on the AR glasses body with respect to the main camera.

[0029] As a preferred embodiment, a card block clamp is installed on the fixed card block, and the fixed card block is installed on the AR glasses body through the card block clamp, and the protective lens is installed on the side close to the main camera.

[0030] As a preferred embodiment, a charging port is installed on the first temple, and an adjustment button is installed on the first temple; ear hooks are installed on both the first temple and the second temple, and a nose pad is installed on the side of the AR glasses body away from the main camera.

[0031] The present invention has the following beneficial effects:

[0032] 1. The power inspection method based on AI recognition and AR provided by the present invention greatly shortens the time of defect detection and improves work efficiency by training a defect recognition model based on a neural network. Compared with traditional manual inspections, the intelligent recognition system can detect and report defects in real time or near real time, reduce the burden on operation and maintenance personnel, and reduce labor costs.

[0033] 2. In the power inspection method based on AI recognition and AR provided by the present invention, in the image preprocessing step, the image is converted into a grayscale image by combining the RGB model and the HSV model, and the grayscale value is normalized at the same time to reduce the noise impact caused by shooting lighting, shadows and other factors.

[0034] 3. The power inspection method based on AI recognition and AR provided by the present invention reduces the impact caused by the imbalance in the number of image samples of different defect types through sample proliferation, thereby improving the recognition accuracy of the neural network.

[0035] 4. The smart AR glasses for power inspection based on AI recognition provided by the present invention are worn on the head by the wearer, and the image in front is collected by the main camera, projected onto the main lens using optical waveguide technology, and the image is transmitted to an external computer through a network communication module, so that the background personnel can observe and collect on-site images conveniently. When the ambient light is poor, the infrared camera can use infrared technology to assist in image collection, so that the user can work normally and collect images in dark light.

[0036] 5. When using the smart AR glasses for power inspection based on AI recognition provided by the present invention, the clamping claws on the fixed clamping block are inserted into the top of the AR glasses body, so that the protective lens can be clamped in front of the lens and the camera, which can avoid direct contact with dust or sparks of electrical components during work, thereby protecting the lenses and cameras with higher maintenance costs and increasing their service life.

[0037] 6. The smart AR glasses for power inspection based on AI recognition provided by the present invention can connect to the network through the speaker and microphone inside the first glasses leg in conjunction with the network communication module, so that on-site staff and back-end technicians can have real-time conversations and provide technical guidance through images captured by the main camera. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of a method flow chart of Embodiment 1 of the present invention;

[0039] Figure 2 This is a schematic diagram of the three-dimensional structure of the AR glasses according to the second embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of the structure of the fixed card block and the protective lens according to the second embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of the structure of the main lens and main camera of the second embodiment of the present invention;

[0042] Figure 5 This is a schematic diagram of the structure of the network communication module and nose pad etc. of the second embodiment of the present invention;

[0043] Figure 6 It is a structural diagram of structures such as a speaker and a microphone according to the second embodiment of the present invention.

[0044] The accompanying drawings in the figure are marked as follows:

[0045] 1. AR glasses body; 2. Fixed card block; 3. Protective lens; 4. Card block clamp; 5. First glasses leg; 6. Ear hook leg; 7. Speaker; 8. Charging port; 9. Adjustment button; 10. Microphone; 11. Nose pad; 12. Second glasses leg; 13. Main lens; 14. Network communication module; 15. Main camera; 16. Infrared camera; 17. Head. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0048] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0049] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0050] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0051] Embodiment 1:

[0052] See also Figure 1 , this embodiment provides a power inspection method based on AI recognition and AR, comprising the following steps:

[0053] S100: According to the needs, several cameras, drones, inspection robots and manual collection of historical image data of the target power inspection scene can be installed to ensure that the collected data covers all key equipment and areas in the machine room, including but not limited to transformers, switch cabinets, cable trays, air conditioning systems, etc. The historical image data is classified into defect-free images and defective images, and the image data is preprocessed to form a training sample set; the preprocessing includes confirming the defect type of the collected defective image and marking its relevant location information;

[0054] S200, establishing a neural network model based on a self-attention mechanism, and initializing the neural network model using transfer learning. The neural network model may adopt a convolutional neural network (CNN), a YOLO series (such as YOLOv5), a Faster R-CNN model, etc.; inputting the training sample set into the initialized neural network model for training;

[0055] S300. During the training process, the neural network model is optimized by using feedback optimization. Cross-validation, early stopping and other techniques can also be used to prevent overfitting. The loss function of the neural network model can be cross entropy loss, mean square error loss, etc. The formula of the cross entropy loss function is:

[0056]

[0057] Where N is the number of input training samples, C is the number of defect categories, and y ic is the true label (0 or 1) of whether the training sample i belongs to category c, and pic is the probability that the model predicts that the training sample i belongs to category c;

[0058] The formula of the mean square error loss function is:

[0059]

[0060] Where N is the number of samples, yi is the true value of training sample i, is the value predicted by the model for sample i;

[0061] The model is trained through back-propagation, the gradient is calculated based on the loss value, and the model parameters are updated through the back-propagation algorithm.

[0062] The parameter update formula of the gradient descent algorithm is:

[0063]

[0064] Among them, θt is the parameter value at the current moment, θt+1 is the updated parameter value, η is the learning rate, is the gradient of the loss function with respect to the parameter θ.

[0065] When the maximum number of training iterations is reached or the recognition accuracy of the neural network model reaches a preset value, the training is terminated to obtain a trained machine room defect recognition model;

[0066] S400. Configure the trained power inspection scene defect recognition model on the wearable AR device. During the inspection of the target power line, obtain the image data of the current scene through the wearable AR device, input it into the trained power inspection scene defect recognition model, identify whether there are defects in the power inspection scene in the current scene, and when there are defects, generate an alarm mark and bind the alarm mark to the location where the defect occurs to generate an AR label model to display it on the display screen of the wearable AR device. The staff can perform relevant operations according to the alarm mark prompted by the display screen. At the same time, once a defect is detected, the wearable AR device should immediately trigger the alarm mechanism to notify relevant personnel to handle it, and the alarm information should include detailed information on the defect type, location and severity, so that personnel can respond quickly.

[0067] As a preferred implementation of this embodiment, the step of preprocessing the image data includes image normalization processing, specifically:

[0068] The collected image data is converted from RGB image to grayscale image, and the following formula is used:

[0069]

[0070] V (x,y) =MAX;

[0071] Among them, L (x,y) is the gray value of the (x, y) coordinate after conversion to a grayscale image, δ is the correction coefficient, α represents the standard spatial distance, and H s is the hue value of sampling point s, S s is the saturation value of sampling point s, V s is the brightness value of sampling point s, H (x,y) is the hue value of the (x, y) coordinate in the original image, S (x,y) is the saturation value of the (x, y) coordinate in the original image, V (x,y)is the brightness value of the (x, y) coordinate in the original image; R, G, B are the red value, green value and blue value of the corresponding pixel respectively, mod6 is the modulus operation; θ = MAX-MIN; MAX is the maximum value among R, G, B, and MIN is the minimum value among R, G, B;

[0072] The converted grayscale image is normalized as follows:

[0073]

[0074] Among them, L (x,y) ′ represents the grayscale value of the normalized grayscale image at the (x, y) coordinate, L max Represents the maximum grayscale value of all coordinates in the original grayscale image, L min Represents the minimum grayscale value among all coordinates in the original grayscale image.

[0075] As a preferred implementation of this embodiment, the step of preprocessing the image data further includes a sample multiplication step, specifically:

[0076] Determine the defect types of all power inspection scenarios, and establish a corresponding number of GAN generation networks according to the number of defect types in the power inspection scenarios, wherein the GAN generation network includes a generator and a discriminator, wherein the generator generates a new defect image according to random noise, and the discriminator is used to identify the similarity between the new defect image and the original defect image;

[0077] Each GAN generation network is used to generate defect images of the corresponding defect type, so that the difference in the number of defect images of each defect type is less than a preset range, and the difference between the number of defect-free images and the total number of defect images is less than a preset range.

[0078] As a preferred implementation of this embodiment, the self-attention mechanism cuts the input image data into n image blocks, and obtains the average grayscale value L1, L2, L3, ... L of each image block. n , and convert the average gray value of each image block into a feature vector through a fully connected layer;

[0079] The self-attention mechanism calculates the attention weight of each image block by using query vector, key vector and value vector, where the query vector corresponds to the average grayscale value of the image block, and both the key vector and the value vector use feature vectors.

[0080] Embodiment 2:

[0081] See also Figure 2-6 , This embodiment provides a smart AR glasses for power inspection based on AI recognition, including an AR glasses body 1 and a processor;

[0082] The processor is used to execute the power inspection method based on AI recognition and AR as described in the first embodiment above;

[0083] In order to facilitate the use of AR glasses to assist in power inspection, the AR glasses also include a post 17, a first temple 5, a second temple 12 and a main camera 15. The post 17 is installed on both sides of the AR glasses body 1, the first temple 5 is rotatably installed on the post 17 on one side, and the second temple 12 is rotatably installed on the post 17 on the side away from the first temple 5. The protective lens 3 is installed on the fixed block 2, and the main camera 15 is installed on one side of the AR glasses body 1. The device is worn through the first temple 5 and the second temple 12, and the inspection image is collected through the main camera 15.

[0084] In order to realize the function of using AR glasses to assist power inspection, main lenses 13 are installed on both sides of the AR glasses body 1, a main camera 15 is installed between the two main lenses 13, a network communication module 14 is installed on the side of the AR glasses body 1 away from the main camera 15, and the network communication module 14 is installed between the two main lenses 13. Infrared cameras 16 are installed on both sides of the AR glasses body 1 close to the main camera 15, and the infrared camera 16 is symmetrically installed on the AR glasses body 1 about the main camera 15. A charging interface 8 is installed on the first glasses leg 5, an adjustment button 9 is installed on the first glasses leg 5, and ear hanging legs 6 are installed on the first glasses leg 5 and the second glasses leg 12. A nose pad 11 is installed on the side of the AR glasses body 1 away from the main camera 15. When in use, the angles of the first glasses leg 5 and the second glasses leg 12 can be rotated by the pile head 17. The first glasses leg 5 and the pile head 17 are rotated respectively to be perpendicular to the surface of the AR glasses body 1, and the device can be worn on the head. When worn, the AR glasses body 1 is located in front of the user, and the nose pad 11 is located on the back of the nose on both sides of the user. When the first glasses leg 5 and the second glasses leg 12 are located at the temporal part on both sides of the user, and the ear hook legs 6 are located on the ears on both sides of the user. The AR glasses are prior art, and the necessary working components such as a central processing unit are installed inside, which will not be repeated here. During inspection, the main camera 15 can capture the image in front, and the image can be transmitted to an external computer or other device through the network communication module 14, which can facilitate real-time observation by background personnel, and can use the computer to record the picture captured by the main camera 15. The network communication module 14 relies on wireless communication protocols to transmit images, such as Wi-Fi and 5G signal connections. This is prior art and will not be repeated here. When the on-site environment is dark, the infrared camera 16 replaces the main camera 15. The infrared camera 16 uses infrared technology to provide the operator with better visual assistance at night or in low-light environments, which is convenient for operators to work in dark places. Since the equipment is provided with a network communication module 14, the network can be used to cooperate with the camera to call AI to identify the collected images, which can assist and assist power personnel to complete daily inspections and recording of electricity.

[0085] See Figure 2 and Figure 3 In order to prevent the main lens 13 and the camera from being polluted and damaged, the AR glasses further include a fixed card block 2 and a protective lens 3. The fixed card block 2 is mounted on the AR glasses main body 1. A card block clamp 4 is mounted on the fixed card block 2. The fixed card block 2 is mounted on the AR glasses main body 1 through the card block clamp 4. The protective lens 3 is mounted on the side close to the main camera 15. Before entering the site for inspection, the card block clamp 4 on the fixed card block 2 is clamped into the AR glasses main body 1. The material of the card block clamp 4 has a certain elasticity, and the gap between the two card block clamps 4 is slightly smaller than the thickness of the AR glasses main body 1, so that when the card block clamp After the claw 4 is inserted, the clamping claw 4 will clamp the AR glasses body 1 to fix it. At this time, the protective lens 3 is located in front of the AR glasses body 1. The protective lens 3 can block the main lens 13, the main camera 15 and the infrared cameras 16 on both sides. The protective lens 3 is made of transparent and light-transmitting material and will not affect the line of sight and camera shooting. The protective lens 3 can prevent dust and splashing objects from directly contacting the main lens 13 or the camera. After the protective lens 3 is worn out after long-term use, the clamping claw 4 can be removed from the AR glasses body 1 and replaced with a new protective lens 3. The remaining features are the same as those in Example 1.

[0086] See Figure 6 In order to facilitate communication between front-end and back-end staff during inspection, the AR glasses also include a speaker 7 and a microphone 10, both of which are installed on the first glasses leg 5. Through the cooperation of the speaker 7 and the microphone 10 and relying on the network communication module 14, the back-end technicians and the on-site inspection personnel can have real-time conversations. The speaker 7 can play the voices of the back-end personnel, and the microphone 10 can collect the voices of the on-site inspection personnel. The remaining features are the same as those in Example 1.

[0087] The working principle is as follows: when in use, the angles of the first eyeglass leg 5 and the second eyeglass leg 12 can be rotated through the stud 17, and the first eyeglass leg 5 and the stud 17 are rotated respectively to be perpendicular to the surface of the AR glasses body 1, so that the device can be worn on the head. When worn, the AR glasses body 1 is located in front of the user, the nose pads 11 are located on the back of the nose on both sides of the user, and the first eyeglass leg 5 and the second eyeglass leg 12 are located on the temporal part on both sides of the user, and the ear hook legs 6 are located on both sides of the ears of the user. In order to prevent the main lens 13 and the camera from being contaminated and damaged, before entering the site for inspection, the clamping claw 4 on the fixed clamping block 2 is clamped into the AR glasses body 1, and the clamping block is clamped The material of the claw 4 has a certain elasticity, and the gap between the two clamping claws 4 is slightly smaller than the thickness of the AR glasses body 1, so that when the clamping claw 4 is inserted, the clamping claw 4 will clamp the AR glasses body 1 to fix it. At this time, the protective lens 3 is located in front of the AR glasses body 1. The protective lens 3 is made of transparent and light-transmitting material and will not affect the line of sight and camera shooting. The protective lens 3 can block the main lens 13, the main camera 15 and the infrared cameras 16 on both sides. The protective lens 3 can prevent dust and splashes from directly contacting the main lens 13 or the camera. After the protective lens 3 is worn out after long-term use, the clamping claw 4 is removed from the AR glasses body 1. The AR glasses are prior art, and the necessary working components such as the central processing unit are installed inside, which will not be described in detail here. During the inspection, the main camera 15 can collect the image in front, and transmit the image to an external computer or other device through the network communication module 14, which can facilitate the background personnel to observe in real time, and can use the computer to record the image collected by the main camera 15. The network communication module 14 relies on wireless communication protocols to transmit images, such as Wi-Fi and 5G signal connections. This is a prior art and will not be described in detail here. When the on-site environment is dark, the infrared camera 16 replaces The main camera 15 is working, and the infrared camera 16 uses infrared technology to provide the operator with better visual assistance at night or in low-light environments, which is convenient for operators to work in dark places. Through the cooperation of the speaker 7 and the microphone 10 and relying on the network communication module 14, the back-end technicians and on-site inspection personnel can have real-time conversations. The speaker 7 can play the voices of the back-end personnel, and the microphone 10 can collect the voices of the on-site inspection personnel. Since the equipment is equipped with a network communication module 14, the network can be used to cooperate with the camera to call AI to identify the collected images, which can assist and assist the power personnel to complete the daily inspection and recording of electricity.

[0088] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0089] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0091] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0092] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A power inspection method based on AI recognition and AR, characterized in that: The following steps are involved: Acquire historical inspection image data of the target power inspection scene, including defect-free images and defective images, and preprocess the historical inspection image data to form a training sample set; Establish a neural network model based on the self-attention mechanism, initialize the neural network model using transfer learning, and input the training sample set into the initialized neural network model for training; During the training process, the neural network model is optimized by using feedback optimization. When the maximum number of training iterations is reached or the recognition accuracy of the neural network model reaches a preset value, the training is terminated to obtain a trained power inspection scene defect recognition model. The trained power inspection scene defect recognition model is configured on the wearable AR device. During the inspection of the target power line, the image data of the current scene is obtained through the wearable AR device and input into the trained power inspection scene defect recognition model to identify whether there are defects in the power inspection scene in the current scene. When defects exist, an alarm mark is generated and the alarm mark is bound to the location where the defect occurs to generate an AR label model, which is displayed on the display screen of the wearable AR device.

2. According to claim 1, a power inspection method based on AI recognition and AR is characterized in that: The step of preprocessing the image data includes image normalization processing, specifically: The collected image data is converted from RGB image to grayscale image, and the following formula is used: V (x,y) =MAX; Among them, L (x,y) is the gray value of the (x, y) coordinate after conversion to a grayscale image, δ is the correction coefficient, α represents the standard spatial distance, and H s is the hue value of sampling point s, S s is the saturation value of sampling point s, V s is the brightness value of sampling point s, H (x,y) is the hue value of the (x, y) coordinate in the original image, S (x,y) is the saturation value of the (x, y) coordinate in the original image, V (x,y) is the brightness value of the (x, y) coordinate in the original image; R, G, B are the red value, green value and blue value of the corresponding pixel respectively, mod6 is the modulus operation; θ = MAX-MIN; MAX is the maximum value among R, G, B, and MIN is the minimum value among R, G, B; The converted grayscale image is normalized as follows: Among them, L (x,y) ′ represents the grayscale value of the normalized grayscale image at the (x, y) coordinate, L max Represents the maximum grayscale value of all coordinates in the original grayscale image, L min Represents the minimum grayscale value among all coordinates in the original grayscale image.

3. The power inspection method based on AI recognition and AR according to claim 1 is characterized in that: The step of preprocessing the image data also includes a sample multiplication step, specifically: Determine the defect types of all power inspection scenarios, and establish a corresponding number of GAN generation networks according to the number of defect types in the power inspection scenarios, wherein the GAN generation network includes a generator and a discriminator, wherein the generator generates a new defect image according to random noise, and the discriminator is used to identify the similarity between the new defect image and the original defect image; Each GAN generation network is used to generate defect images of the corresponding defect type, so that the difference in the number of defect images of each defect type is less than a preset range, and the difference between the number of defect-free images and the total number of defect images is less than a preset range.

4. The power inspection method based on AI recognition and AR according to claim 1 is characterized in that: The self-attention mechanism cuts the input image data into n image blocks and obtains the average grayscale value L1, L2, L3, ...L of each image block. n , and convert the average gray value of each image block into a feature vector through a fully connected layer; The self-attention mechanism calculates the attention weight of each image block by using query vector, key vector and value vector, where the query vector corresponds to the average grayscale value of the image block, and both the key vector and the value vector use feature vectors.

5. A smart AR glasses for power inspection based on AI recognition, comprising an AR glasses body (1) and a processor; characterized in that: The processor is used to execute the power inspection method based on AI recognition and AR as described in any one of claims 1 to 4; The AR glasses further comprise a head (17), a first temple (5), a second temple (12), a speaker (7), a microphone (10), a fixed card block (2), a protective lens (3) and a main camera (15); the head (17) is mounted on both sides of the AR glasses body (1); the first temple (5) is rotatably mounted on the head (17) on one side; the second temple (12) is rotatably mounted on the head (17) on the side away from the first temple (5); the speaker (7) and the microphone (10) are both mounted on the first temple (5); the fixed card block (2) is mounted on the AR glasses body (1); the protective lens (3) is mounted on the fixed card block (2); and the main camera (15) is mounted on one side of the AR glasses body (1).

6. The smart AR glasses for power inspection based on AI recognition according to claim 5 are characterized by: Main lenses (13) are installed on both sides of the AR glasses main body (1), and the main camera (15) is installed between the two main lenses (13).

7. The smart AR glasses for power inspection based on AI recognition according to claim 6 are characterized by: A network communication module (14) is installed on the side of the AR glasses body (1) away from the main camera (15), and the network communication module (14) is installed between the two main body lenses (13).

8. The smart AR glasses for power inspection based on AI recognition according to claim 5 are characterized by: Infrared cameras (16) are installed on both sides of the AR glasses body (1) close to the main camera (15), and the infrared cameras (16) are installed on the AR glasses body (1) symmetrically with respect to the main camera (15).

9. The smart AR glasses for power inspection based on AI recognition according to claim 5, characterized in that: A card block clamp (4) is installed on the fixed card block (2), and the fixed card block (2) is installed on the AR glasses body (1) through the card block clamp (4), and the protective lens (3) is installed on the side close to the main camera (15).

10. The smart AR glasses for power inspection based on AI recognition according to claim 5, characterized in that: The first spectacles leg (5) is provided with a charging interface (8), and the first spectacles leg (5) is provided with an adjustment button (9); The first glasses leg (5) and the second glasses leg (12) are both provided with ear hook legs (6), and a nose pad (11) is provided on the side of the AR glasses body (1) away from the main camera (15).

Images