Electricity target identification method and device, monitoring system, and readable storage medium

Abstract

This invention provides a power target recognition method and apparatus, a monitoring system, and a readable storage medium. The method includes: acquiring a first inspection image from a power inspection device; wherein the first inspection image is an image captured by the power inspection device during inspection, containing the target to be identified and environmental features; extracting environmental features from the first inspection image as a first image; inputting the first image into a pre-trained deep learning model to determine preset weight coefficients of a first target recognition model; the deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model; and inputting the first inspection image into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image. This invention can minimize the influence of the shooting environment, thereby improving the accuracy of power target recognition.

Description

Technical Field

[0001] This invention belongs to the field of target recognition technology, and more specifically, relates to a power target recognition method and device, a monitoring system, and a readable storage medium. Background Technology

[0002] With the improvement of people's living standards and the continuous expansion of the power grid, the reliable operation of power lines and equipment has become a key concern. Therefore, the condition monitoring of power equipment / lines is extremely important. Current technologies typically use intelligent power inspection devices (such as inspection robots) to capture images of power equipment / lines during automated inspections. Based on this, a target recognition model is used to automatically identify power targets in the images captured by the power inspection device, and then the condition of the identified power targets is monitored.

[0003] However, in applying the existing technology, the inventors of this application discovered that the existing target recognition models are fixed after training and cannot change with the shooting environment. Since the shooting environment often changes, the existing target recognition models often result in inaccurate recognition, missed recognition, and false recognition when used to identify power targets. In other words, the existing power target recognition accuracy is low. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for identifying power targets, a monitoring system, and a readable storage medium, so as to solve the problem of low identification accuracy of power targets in the prior art.

[0005] A first aspect of the present invention provides a method for identifying power targets, comprising:

[0006] The first inspection image is obtained from the power inspection device; wherein the first inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during inspection.

[0007] An image containing environmental features extracted from the first inspection image is used as the first image;

[0008] The first image is input into a pre-trained deep learning model to determine the weight coefficients of a preset first target recognition model; wherein, the deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model;

[0009] The first inspection image is input into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

[0010] In one possible implementation, the power target recognition method further includes: training the deep learning model; the training of the deep learning model includes:

[0011] The second inspection image is obtained from the power inspection device under different shooting environments; wherein, the second inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during historical inspections;

[0012] Multiple second target recognition models are trained based on the second inspection images; wherein, the network structure of the second target recognition model is the same as that of the first target recognition model, and each shooting environment corresponds to one second target recognition model;

[0013] The image from which environmental features are extracted from the second inspection image is used as the second image;

[0014] The deep learning model is trained based on the second image under each shooting environment and the weight coefficients of the second target recognition model corresponding to that shooting environment.

[0015] In one possible implementation, acquiring second inspection images from the power inspection device under different shooting environments includes:

[0016] The power inspection device acquires second inspection images from different shooting angles under different shooting environments;

[0017] Each shooting environment corresponds to multiple second inspection images from different shooting angles.

[0018] In one possible implementation, training multiple second target recognition models based on the second inspection image includes:

[0019] For the target shooting environment, a second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment; wherein, the target shooting environment is any shooting environment among the different shooting environments;

[0020] For each non-target shooting environment, the second target recognition model corresponding to the target shooting environment is updated based on the second inspection image in the non-target shooting environment to obtain the second target recognition model corresponding to the non-target shooting environment.

[0021] In one possible implementation, the second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment, including:

[0022] The targets in the second inspection images under the target shooting environment are labeled to obtain the labeling information of each second inspection image under the target shooting environment;

[0023] A second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment and the annotation information.

[0024] In one possible implementation, the step of extracting environmental features from the first inspection image as the first image includes:

[0025] The first inspection image is processed based on a pre-trained target detection model to determine the region where the target is located in the first image;

[0026] In the first inspection image, a portion of the area excluding the area where the target is located is selected as the first image.

[0027] In one possible implementation, both the deep learning model and the first target recognition model are neural network models.

[0028] A second aspect of the present invention provides a power target identification device, comprising:

[0029] The data acquisition module is used to acquire a first inspection image from the power inspection device; wherein, the first inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during inspection;

[0030] The image extraction module is used to extract an image of environmental features from the first inspection image as the first image;

[0031] The model determination module is used to input the first image into a pre-trained deep learning model and determine the weight coefficients of a preset first target recognition model; wherein, the deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model;

[0032] The target recognition module is used to input the first inspection image into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

[0033] A third aspect of the present invention provides a monitoring system, the monitoring system including a monitoring terminal, the monitoring terminal including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described power target identification method.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the power target identification method described above.

[0035] The beneficial effects of the power target identification method and device, monitoring system, and readable storage medium provided in the embodiments of the present invention are as follows:

[0036] To avoid the impact of changes in the shooting environment on target recognition, this embodiment of the invention pre-trains a deep learning model. This deep learning model essentially describes the mapping relationship between the shooting environment (environmental features) and the weight coefficients of the first target recognition model. Based on this, a first image containing the shooting environment (environmental features) can be extracted from the first inspection image. Then, inputting this first image into the aforementioned deep learning model determines the first target recognition model matching the shooting environment. Finally, inputting the first inspection image into the first target recognition model matching its shooting environment yields the target recognition result. Analysis of this embodiment shows that the weight coefficients of the first target recognition model adaptively change with different shooting environments, thus minimizing the influence of the shooting environment and improving the accuracy of target recognition. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating a power target identification method provided in an embodiment of the present invention;

[0039] Figure 2 This is a structural block diagram of a power target identification device provided in an embodiment of the present invention;

[0040] Figure 3 This is a schematic block diagram of a monitoring terminal provided in an embodiment of the present invention. Detailed Implementation

[0041] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0042] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0043] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a power target recognition method according to an embodiment of the present invention. The power target recognition method is used to identify power targets in a power system in an inspection image.

[0044] The aforementioned identification of power targets can refer to identifying power targets in inspection images, identifying the state of power targets, or identifying the location / height of power targets, etc. This embodiment of the invention does not limit this.

[0045] The electrical targets include, but are not limited to, power lines and / or electrical equipment. For example, an electrical target can be a cable, a circuit breaker, an insulator, or an air switch, etc.

[0046] Based on this, the power target identification method described in the embodiments of the present invention includes:

[0047] S101: Obtain the first inspection image from the power inspection device. The first inspection image is an image captured by the power inspection device during inspection, containing the target to be identified and environmental features.

[0048] In this embodiment, the first inspection image is the image to be identified as the target. After the power inspection device captures the inspection image of its inspection area, it can be transmitted to the execution entity of the method described in this embodiment of the invention via a communication device, whereby the execution entity processes the inspection image of the power inspection device.

[0049] S102: Extract the image with environmental features from the first inspection image and use it as the first image.

[0050] In this embodiment, the first inspection image includes regions containing the target to be identified and regions without the target. Based on this, the image of the region without the target can be extracted from the first inspection image to obtain the first image. The first image is essentially an image of the shooting environment corresponding to the first inspection image, and the features contained in the first image are essentially environmental features of the shooting environment corresponding to the first inspection image. Therefore, based on the first image, shooting environment factors can be taken into account in the target identification of the first inspection image.

[0051] S103: Input the first image into a pre-trained deep learning model to determine the weight coefficients of the preset first target recognition model. The deep learning model is used to establish the mapping relationship between environmental features and the weight coefficients of the first target recognition model.

[0052] In this embodiment, the deep learning model is essentially used to describe the mapping relationship between the shooting environment (environmental features) and the weight coefficients of the first target recognition model. Specifically, it is used to output the corresponding weight coefficients based on the input first image, which are the weight coefficients of the first target recognition model.

[0053] S104: Input the first inspection image into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

[0054] In this embodiment, the first target recognition model is essentially an untrained model, where the training process is the process of determining the various weight coefficients of the model. Based on this, the first target recognition model with determined weight coefficients is essentially a trained model. The trained first target recognition model is used to output the corresponding target recognition result based on the input inspection image. Therefore, the first inspection image can be input into the first target recognition model to obtain the target recognition result of the first inspection image.

[0055] As described above, to avoid the impact of changes in the shooting environment on target recognition, this embodiment of the invention pre-trains a deep learning model. This deep learning model essentially describes the mapping relationship between the shooting environment (environmental features) and the weight coefficients of the first target recognition model. Based on this, a first image containing the shooting environment (environmental features) can be extracted from the first inspection image. Then, inputting this first image into the aforementioned deep learning model determines the first target recognition model matching the shooting environment. Finally, inputting the first inspection image into the first target recognition model matching its shooting environment yields the target recognition result. Analysis of this embodiment shows that the weight coefficients of the first target recognition model adaptively change with different shooting environments, thus minimizing the influence of the shooting environment and improving the accuracy of target recognition.

[0056] In one possible implementation, the power target recognition method further includes training a deep learning model. Training the deep learning model includes:

[0057] Second inspection images are acquired from the power line inspection device under different shooting environments. These second inspection images are images taken by the power line inspection device during historical inspections, containing both the target to be identified and environmental features.

[0058] Multiple second target recognition models are trained based on the second inspection images. The network structure of the second target recognition model is the same as that of the first target recognition model, and one second target recognition model corresponds to each shooting environment.

[0059] The image from which environmental features are extracted from the second inspection image is used as the second image.

[0060] A deep learning model is trained based on the second image in each shooting environment and the weight coefficients of the second target recognition model corresponding to that shooting environment.

[0061] In this embodiment, the second inspection image is essentially a historical inspection image captured by the power inspection device, i.e., an inspection image used as a training sample. Based on this, a second target recognition model for each shooting environment can be trained using the second inspection images under each shooting environment. That is, the weight coefficients of the second target recognition model under each shooting environment are determined based on the second inspection images under each shooting environment.

[0062] In this embodiment, a second image that does not contain the target to be identified can be extracted from the second inspection image under each shooting environment. The second image is essentially an image of the shooting environment corresponding to the second inspection image, and the features contained in the second image are essentially environmental features of the shooting environment corresponding to the second inspection image.

[0063] Based on this, the aforementioned deep learning model can be trained using the second image under each shooting environment and the weight coefficients under each shooting environment.

[0064] In this embodiment, the second target recognition model is essentially the same as the first target recognition model; the difference is that the second target recognition model is applied during the data training phase. Having the same network structure means that their data structures are completely identical. For example, when the first target recognition model is implemented using a neural network structure, having the same network structure means that the number and location of neurons are the same.

[0065] In one possible implementation, acquiring second inspection images from the power line inspection device under different shooting environments includes:

[0066] Second inspection images are obtained from different shooting angles under different shooting environments using a power line inspection device.

[0067] Each shooting environment corresponds to multiple second inspection images from different shooting angles.

[0068] In this embodiment, in order to train the second target recognition model more accurately, and thus train the deep learning model more accurately, second inspection images from different shooting angles can be acquired in each shooting environment to improve the diversity of samples and improve the training accuracy of various models.

[0069] In one possible implementation, multiple second target recognition models are trained based on the second inspection image, including:

[0070] For the target shooting environment, a second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment. Here, the target shooting environment can be any shooting environment under different shooting environments.

[0071] For each non-target shooting environment, the second target recognition model corresponding to the target shooting environment is updated based on the second inspection image in the non-target shooting environment to obtain the second target recognition model corresponding to the non-target shooting environment.

[0072] In this embodiment, the target shooting environment is any shooting environment selected from the aforementioned different shooting environments, and the non-target shooting environment refers to the shooting environment other than the target shooting environment among the aforementioned different shooting environments.

[0073] In this embodiment, a shooting environment (i.e., the target shooting environment) can be selected first, and a second target recognition model under that shooting environment can be trained. Based on this, considering the large computational load of training multiple second target recognition models, this embodiment provides a method to improve training speed. That is, after training one second target recognition model, when training other second target recognition models, it is not necessary to start training from scratch; instead, the weight coefficients of the already trained second target recognition model can be updated directly to quickly obtain other second target recognition models.

[0074] In this embodiment, it should be noted that the model training process is usually as follows: at the beginning of model training, the weight coefficients of the model are randomly initialized. On this basis, the weight coefficients are continuously updated using training samples (in this example, the second inspection image) until the model meets the predetermined accuracy requirements or the number of iterations of model training reaches the preset number of iterations. At this time, the model training is completed, and the current weight coefficients are the weight coefficients of the trained model.

[0075] Based on this, the updating of the trained model described in this embodiment of the invention refers to the following: at the start of model training, the weight coefficients of the model are not randomly initialized. Instead, the weight coefficients of the already trained model are directly used as the initial weight coefficients for this model training (that is, the weight coefficients of the second target recognition model under the target shooting environment are used as the initial weight coefficients of the second target recognition model under the non-target shooting environment). Since there are certain commonalities between the models under different shooting environments in this embodiment, the model update method can accelerate the convergence progress of the model during subsequent model training, thereby achieving rapid model training. Therefore, this invention can first train the second target recognition model under the target shooting environment. Subsequently, when training the second target recognition model under other shooting environments (i.e., non-target shooting environments), the second target recognition model under the target shooting environment can be directly updated through training samples, thereby achieving rapid training of a large number of second target recognition models and reducing the amount of computation. It should be noted that the network structure of the second target recognition model is consistent under different shooting environments.

[0076] In one possible implementation, a second target recognition model corresponding to the target shooting environment is trained based on a second inspection image of the target shooting environment, including:

[0077] The targets in the second inspection images under the target shooting environment are labeled to obtain the labeling information of each second inspection image under the target shooting environment.

[0078] A second target recognition model corresponding to the target shooting environment is trained based on the second inspection image and annotation information under the target shooting environment.

[0079] It is known that during model training, the training samples typically need to include the model's input data and theoretical output data. In this embodiment, the model's input data is the second inspection image under the target shooting environment. Based on this, the targets in the second inspection image can be labeled to obtain the model's theoretical output data, i.e., the labeling information. Subsequently, the second target recognition model under the target shooting environment can be trained based on the second inspection image and its labeling information.

[0080] In one possible implementation, the image from which environmental features are extracted from the first inspection image is used as the first image, including:

[0081] The first inspection image is processed based on a pre-trained target detection model to determine the region where the target is located in the first image.

[0082] In the first inspection image, a portion of the area excluding the target area is randomly selected as the first image.

[0083] In this embodiment, the bounding box of the target in the first inspection image can be determined using an existing target detection model. The area occupied by the bounding box in the first inspection image is the area where the target is located in the first image. Based on this, the area in the first inspection image other than the bounding box is the area where the first image is located. Based on this, a portion of the image in the area where the first image is located can be randomly selected as the first image to describe the shooting environment.

[0084] In one possible implementation, both the deep learning model and the first target recognition model are neural network models.

[0085] In this embodiment, both the deep learning model and the first target recognition model can be implemented based on a neural network structure. Preferably, the deep learning model and the first target recognition model can use different types of neural network models (for example, the deep learning model can be implemented using a fully connected neural network (FCN), and the first target recognition model can be implemented using a convolutional neural network (CNN)) to better suit their functions, make better use of the advantages of different neural network structures, and thus improve the accuracy of target recognition.

[0086] Corresponding to the power target identification method in the above embodiments, Figure 2 This is a structural block diagram of a power target identification device according to an embodiment of the present invention. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown. (See references) Figure 2 The power target recognition device 20 includes: a data acquisition module 21, an image extraction module 22, a model determination module 23, and a target recognition module 24.

[0087] The data acquisition module 21 is used to acquire a first inspection image from the power inspection device. The first inspection image is an image captured by the power inspection device during an inspection, containing the target to be identified and environmental features.

[0088] Image extraction module 22 is used to extract an image of environmental features from the first inspection image as the first image.

[0089] The model determination module 23 is used to input the first image into a pre-trained deep learning model and determine the weight coefficients of the preset first target recognition model. The deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model.

[0090] The target recognition module 24 is used to input the first inspection image into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

[0091] In one possible implementation, the model determination module 23 is further used to: train a deep learning model. Training the deep learning model includes:

[0092] Second inspection images are acquired from the power line inspection device under different shooting environments. These second inspection images are images taken by the power line inspection device during historical inspections, containing both the target to be identified and environmental features.

[0093] Multiple second target recognition models are trained based on the second inspection images. The network structure of the second target recognition model is the same as that of the first target recognition model, and one second target recognition model corresponds to each shooting environment.

[0094] The image from which environmental features are extracted from the second inspection image is used as the second image.

[0095] A deep learning model is trained based on the second image in each shooting environment and the weight coefficients of the second target recognition model corresponding to that shooting environment.

[0096] In one possible implementation, the model determination module 23 is specifically used for:

[0097] Second inspection images are obtained from different shooting angles under different shooting environments using a power line inspection device.

[0098] Each shooting environment corresponds to multiple second inspection images from different shooting angles.

[0099] In one possible implementation, the model determination module 23 is specifically used for:

[0100] For the target shooting environment, a second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment. Here, the target shooting environment can be any shooting environment under different shooting environments.

[0101] For each non-target shooting environment, the second target recognition model corresponding to the target shooting environment is updated based on the second inspection image in the non-target shooting environment to obtain the second target recognition model corresponding to the non-target shooting environment.

[0102] In one possible implementation, the model determination module 23 is specifically used for:

[0103] The targets in the second inspection images under the target shooting environment are labeled to obtain the labeling information of each second inspection image under the target shooting environment.

[0104] A second target recognition model corresponding to the target shooting environment is trained based on the second inspection image and annotation information under the target shooting environment.

[0105] In one possible implementation, the image extraction module 22 is specifically used for:

[0106] The first inspection image is processed based on a pre-trained target detection model to determine the region where the target is located in the first image.

[0107] In the first inspection image, a portion of the area excluding the target area is randomly selected as the first image.

[0108] In one possible implementation, both the deep learning model and the first target recognition model are neural network models.

[0109] This invention also provides a monitoring system, which includes a monitoring terminal, see [link to relevant documentation]. Figure 3 , Figure 3 This is a schematic block diagram of a monitoring terminal provided in an embodiment of the present invention. Figure 3 The terminal 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules / units in the above-described device embodiments, such as... Figure 2 The functions of modules 21 to 24 are shown.

[0110] It should be understood that, in this embodiment of the invention, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0111] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0112] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0113] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of the present invention can execute the implementation methods described in the first and second embodiments of the power target identification method provided in the embodiments of the present invention, or they can execute the implementation methods of the terminal described in the embodiments of the present invention, which will not be repeated here.

[0114] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. The computer program can also instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0115] The computer-readable storage medium can be an internal storage unit of the terminal in any of the foregoing embodiments, such as the terminal's hard disk or memory. The computer-readable storage medium can also be an external storage device of the terminal, such as a plug-in hard disk, SmartMediaCard (SMC), SecureDigital (SD) card, FlashCard, etc., equipped on the terminal. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the terminal. The computer-readable storage medium is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0116] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0117] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the terminals and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed terminals and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0120] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying power targets, characterized in that, include: The first inspection image is obtained from the power inspection device; wherein the first inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during inspection. An image from which environmental features are extracted from the first inspection image is used as the first image; The first image is input into a pre-trained deep learning model to determine the weight coefficients of a preset first target recognition model; wherein, the deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model; The first inspection image is input into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

2. The power target identification method as described in claim 1, characterized in that, The power target identification method further includes: training the deep learning model; Training the deep learning model includes: The second inspection image is obtained from the power inspection device under different shooting environments; wherein, the second inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during historical inspections; Multiple second target recognition models are trained based on the second inspection images; wherein, the network structure of the second target recognition model is the same as that of the first target recognition model, and each shooting environment corresponds to one second target recognition model; The image from which environmental features are extracted from the second inspection image is used as the second image; The deep learning model is trained based on the second image under each shooting environment and the weight coefficients of the second target recognition model corresponding to that shooting environment.

3. The power target identification method as described in claim 2, characterized in that, The acquisition of second inspection images under different shooting environments from the power inspection device includes: The power inspection device acquires second inspection images from different shooting angles under different shooting environments; Each shooting environment corresponds to multiple second inspection images from different shooting angles.

4. The power target identification method as described in claim 2, characterized in that, The training of multiple second target recognition models based on the second inspection image includes: For the target shooting environment, a second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment; wherein, the target shooting environment is any shooting environment among the different shooting environments; For each non-target shooting environment, the second target recognition model corresponding to the target shooting environment is updated based on the second inspection image in the non-target shooting environment to obtain the second target recognition model corresponding to the non-target shooting environment.

5. The power target identification method as described in claim 4, characterized in that, The second target recognition model, trained based on the second inspection image under the target shooting environment, includes: The targets in the second inspection images under the target shooting environment are labeled to obtain the labeling information of each second inspection image under the target shooting environment; A second target recognition model corresponding to the target shooting environment is trained based on the second inspection image under the target shooting environment and the annotation information.

6. The power target identification method as described in claim 1, characterized in that, The first image, which is the image from which environmental features are extracted from the first inspection image, includes: The first inspection image is processed based on a pre-trained target detection model to determine the region where the target is located in the first image; In the first inspection image, a portion of the area excluding the area where the target is located is selected as the first image.

7. The power target identification method according to any one of claims 1 to 6, characterized in that, Both the deep learning model and the first target recognition model are neural network models.

8. A power target identification device, characterized in that, include: The data acquisition module is used to acquire a first inspection image from the power inspection device; wherein, the first inspection image is an image containing the target to be identified and environmental features captured by the power inspection device during inspection; The image extraction module is used to extract an image of environmental features from the first inspection image as the first image; The model determination module is used to input the first image into a pre-trained deep learning model and determine the weight coefficients of a preset first target recognition model; wherein, the deep learning model is used to establish a mapping relationship between environmental features and the weight coefficients of the first target recognition model; The target recognition module is used to input the first inspection image into the first target recognition model with determined weight coefficients to obtain the target recognition result of the first inspection image.

9. A monitoring system, characterized in that, include: Monitoring terminal; The monitoring terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images