Image recognition method and system in intelligent patrol of transformer substation

Abstract

The invention discloses an image recognition method and system in transformer substation intelligent patrol, and relates to the technical field of transformer substation intelligent patrol, and the method comprises the steps: obtaining the spatial data of a photovoltaic station according to an unmanned plane group, and generating a patrol path of intelligent patrol; a collection device is installed in the inspection path, image information is collected, and a first fusion image is obtained through a first fusion method; performing feature extraction on the first fusion image, and obtaining a second fusion image through a second fusion method; carrying out fault identification on the fused features through hot spot detection and shielding detection, and extracting fault features; and obtaining fault positioning information in the spatial model according to the extracted fault features. According to the method, the automation level of substation inspection and the accuracy and efficiency of fault detection are effectively improved, and a powerful guarantee is provided for safe and stable operation of the substation.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent inspection of substations, and in particular to an image recognition method and system in intelligent inspection of substations. Background Art

[0002] Current PV power plant inspection methods rely primarily on manual labor, which is inefficient, costly, and difficult to cover large plant areas. Manual inspections are not only time-consuming and labor-intensive, but also difficult to ensure accuracy and consistency due to human factors. Furthermore, manual inspections pose safety risks for installations in complex terrain or other inaccessible locations. Therefore, traditional manual inspection methods are not suitable for safe and efficient inspections of PV plants.

[0003] A solution is provided that can replace manual inspections with drone groups, can perform three-dimensional modeling of photovoltaic stations, determine the layout structure of photovoltaic stations, and can group drones based on the layout structure of photovoltaic stations. It can simultaneously enable drones to inspect photovoltaic stations from multiple angles and plan appropriate inspection routes. It can quickly and accurately obtain multi-spectral images of unmanned photovoltaic stations, perform fault identification on the multi-spectral images, and determine the type and location of the fault by combining visible light and infrared light images. Compared with manual inspections, this solution can not only improve inspection efficiency, but also discover potential faults, accurately obtain the fault location, and achieve precise positioning.

[0004] In view of this, an image recognition method and system for intelligent inspection of substations are needed. Summary of the Invention

[0005] In view of the above-mentioned existing problems, the present invention aims to solve the following problems: the traditional manual inspection method has high limitations in manpower, time and cost, and it is difficult to achieve efficient inspection of large-scale photovoltaic sites; the existing technology often relies on a single image data source (such as visible light or infrared images), resulting in incomplete and inaccurate fault detection in complex environments; due to the diverse and complex fault types of photovoltaic equipment, traditional methods are prone to missed detection or false detection when identifying specific faults; during the operation of photovoltaic sites, real-time monitoring of equipment and fault location are particularly important.

[0006] In order to solve the above technical problems, an image recognition method for intelligent inspection of substations is proposed, including:

[0007] Based on the spatial data of the photovoltaic station obtained by the drone group, an inspection path for intelligent patrol is generated; acquisition equipment is installed in the inspection path to collect image information, and a first fused image is obtained through a first fusion method; features are extracted from the first fused image, and a second fused image is obtained through a second fusion method; the fused features are used for fault identification through hot spot detection and occlusion detection to extract fault features; and fault location information is obtained in the spatial model based on the extracted fault features.

[0008] As a preferred solution of the image recognition method in the intelligent inspection of substations according to the present invention, wherein: the spatial data includes first dimension data and second dimension data;

[0009] Generate a photovoltaic station spatial model based on spatial data and then establish a three-dimensional spatial coordinate system, perform inspection zoning, and determine the inspection path;

[0010] Among them, the first dimension data is the two-dimensional base map data of the photovoltaic station;

[0011] The second dimension data is the three-dimensional scanning data of the photovoltaic station.

[0012] As a preferred solution of the image recognition method in the intelligent inspection of substations according to the present invention, wherein: the collected image information includes first image information and second image information;

[0013] The first image information is a visible spectrum image;

[0014] The second image information is an infrared spectrum image.

[0015] As a preferred solution of the image recognition method in the intelligent inspection of a substation according to the present invention, wherein: the obtaining of the first fused image by the first fusion method includes fusing the first image information and the second image information by the first fusion method to obtain the first fused image;

[0016] Wherein, the first fused image is a multispectral image.

[0017] As a preferred solution of the image recognition method in the intelligent inspection of substations described in the present invention, the feature extraction includes performing feature extraction on the first fused image through a preset extraction method to obtain physical features and temperature features.

[0018] As a preferred solution of the image recognition method in the intelligent inspection of substations according to the present invention, wherein: obtaining the second fused image includes fusing the physical feature and the temperature feature using a second fusion method to obtain the second fused image;

[0019] Wherein, the second fused image is a temperature distribution characteristic image.

[0020] As a preferred embodiment of the image recognition method for intelligent inspection of substations according to the present invention, the fault feature extraction includes performing hot spot detection and occlusion detection on the second fused image to obtain hot spot detection results and foreign object occlusion results, and determining the fault feature based on the hot spot detection results and foreign object occlusion detection results.

[0021] When the fault feature is a target fault feature, position information of the fault feature in the information of the first fused image is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

[0022] Another object of the present invention is to provide an image recognition system for intelligent inspection of substations.

[0023] As a preferred solution of the image recognition system in the intelligent inspection of substations described in the present invention, it is characterized by comprising a path inspection module, an image fusion module, and a fault identification module;

[0024] The path inspection module includes an inspection path determination unit and a data acquisition unit. The inspection path determination unit generates a photovoltaic station spatial model based on spatial data and then establishes a three-dimensional spatial coordinate system, performs inspection zoning, and determines the inspection path; the data acquisition unit installs acquisition equipment in the inspection path to collect image information;

[0025] The image fusion module includes a first image fusion unit and a second image fusion unit. The first image fusion unit fuses the first image information and the second image information using a first fusion method to obtain a first fused image. The second image fusion unit extracts features from the first fused image using a preset extraction method to obtain physical object features and temperature features. The physical object features and the temperature features are fused using a second fusion method to obtain a second fused image.

[0026] The fault identification module includes a fault feature extraction unit and a fault location unit. The fault feature extraction unit performs feature extraction on the first fused image through hot spot detection and occlusion detection to obtain physical features and temperature features. The fault location unit performs feature fusion on the physical features and the temperature features through a second fusion method to obtain a second fused image, performs hot spot detection and occlusion detection on the second fused image to obtain hot spot detection results and foreign object occlusion results, and determines the fault feature based on the hot spot detection results and foreign object occlusion detection results. When the fault feature is a target fault feature, the position information of the fault feature in the first fused image information is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

[0027] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the image recognition method in the intelligent inspection of substations are implemented.

[0028] A computer-readable storage medium stores a computer program thereon, characterized in that when the computer program is executed by a processor, the steps of the image recognition method in the intelligent inspection of substations are implemented.

[0029] Beneficial effects of the present invention: The present invention not only improves the accuracy and real-time performance of fault detection by deeply fusing the three-dimensional spatial data and multispectral images of photovoltaic stations, but also provides strong support for the positioning of equipment faults. Specifically, by patrolling the photovoltaic stations using drones, an accurate three-dimensional spatial model is first generated, and then by collecting visible light and infrared images, advanced image fusion technology is used to form a multispectral image, thereby integrating visible light and thermal information. Combining deep learning algorithms for feature extraction and fault identification can effectively cope with the challenges brought about by environmental changes, especially when multiple fault features exist at the same time, to achieve a more comprehensive fault judgment. In addition, the fault positioning capability based on the spatial model significantly improves the efficiency and accuracy of fault handling. All of this makes the present invention not only have high technical content, but also can meet the growing demand for photovoltaic inspections in practice, and provide a new solution for the intelligent construction of the photovoltaic industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 This is a general flow chart of an image recognition method for intelligent substation inspection provided by one embodiment of the present invention.

[0032] Figure 2 A diagram illustrating an apparatus for an image recognition method for intelligent substation inspection according to an embodiment of the present invention.

[0033] Figure 3 A schematic diagram of the system framework of an image recognition system for intelligent substation inspection provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0034] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0035] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an image recognition method for intelligent inspection of substations, including:

[0036] S1: Generate an inspection route for intelligent patrol based on the spatial data of the photovoltaic station obtained by the drone group.

[0037] Furthermore, the spatial data includes first dimension data and second dimension data;

[0038] Among them, the first dimension data is the two-dimensional base map data of the photovoltaic station;

[0039] The second dimension data is the three-dimensional scanning data of the photovoltaic station.

[0040] Specifically, a spatial model of the photovoltaic station is generated based on the first dimensional data and the second dimensional data by the drone group, and a three-dimensional spatial coordinate system is established based on the spatial model. The method of establishing the three-dimensional spatial coordinate system includes but is not limited to a local Cartesian coordinate system, a point cloud registration technology, etc.

[0041] Querying and determining grouping information of the drone group, and segmenting the spatial model of the photovoltaic station based on the grouping information, wherein the segmentation method includes but is not limited to a grid division method, a device cluster division method, etc.;

[0042] A plurality of inspection partitions are obtained, and an inspection path is determined based on the inspection partitions and the grouping information. The method for determining the inspection path includes but is not limited to multi-agent collaborative planning, traveling salesman problem (TSP) optimization, etc.

[0043] S2: Installing a collection device in the inspection path, collecting image information, and obtaining a first fused image using a first fusion method.

[0044] Furthermore, the drone can be equipped with acquisition equipment including visible light image acquisition equipment and infrared image acquisition equipment according to mission requirements;

[0045] Specifically, in order to improve the performance of the UAV, a single UAV includes one of a visible light image acquisition device and an infrared image acquisition device;

[0046] It should be noted that in a squad formation, the equipment difference between the visible light image acquisition equipment and the infrared image acquisition equipment should not be greater than 1;

[0047] After obtaining the visible spectrum images and infrared spectrum images collected by the visible light image acquisition device and the infrared image acquisition device, fault identification can be performed on the visible spectrum images and the infrared spectrum images to determine the fault type in the photovoltaic station and the corresponding fault characteristics.

[0048] It should be noted that the collected image information includes first image information and second image information; the first image information is a visible spectrum image; the second image information is an infrared spectrum image;

[0049] It should be noted that the visible spectrum image and the infrared spectrum image can form a first fused image, so that one image includes both visible light information and infrared information, and can simultaneously reflect the texture details and thermal information of the photovoltaic panel.

[0050] It should also be noted that the first image information and the second image information are fused by a first fusion method to obtain a first fused image; wherein the first fused image is a multispectral image.

[0051] Specifically, the first fusion method is to fuse visible spectrum images and infrared spectrum images, and improve the fault detection capability and efficiency of drones when inspecting photovoltaic stations by combining the complementary information of different spectral images, including but not limited to fusion methods based on pyramid decomposition, fusion methods based on wavelet transform, fusion based on gradient transformation (such as NSCT, DTCWT), deep learning fusion (such as CNN, GAN), principal component analysis (PCA) fusion, etc.

[0052] In an optional embodiment, the first fusion method can be a fusion method based on pyramid decomposition, which decomposes the visible light and infrared images into pyramid layers of different resolutions (high-frequency details + low-frequency background), respectively, and adopts maximum fusion (retaining significant edges) for the high-frequency layer and weighted averaging (retaining background information) for the low-frequency layer, and finally reconstructs the image; the multi-scale spatial details are retained by the Laplacian pyramid fusion method, while the advantages of the two images are fused at the same time.

[0053] In another optional embodiment, the first fusion method can also be a fusion method based on wavelet transform, which performs wavelet decomposition on the two images to obtain low-frequency (approximate) coefficients and high-frequency (detail) coefficients. The low-frequency coefficients can be fused using the average value (smooth background), and the high-frequency coefficients can be selected using the maximum absolute value (highlighting edges and hot spots). The fused image is reconstructed by inverse wavelet transform; information is separated in the frequency domain to avoid distortion caused by direct pixel superposition, which is suitable for retaining local features (such as cracks or hot spots in photovoltaic panels).

[0054] In the present invention, an optional method is as follows: a drone flies along a preset inspection route to collect first image information and second image information of a photovoltaic station, and fuses the collected first image information and second image information, i.e., a visible spectrum image and an infrared spectrum image, to obtain a first fused image, i.e., a multispectral image;

[0055] When performing fusion, the calculation is as follows:

[0056] I fusion =α·I visible +(1-α)·I infrared

[0057] Among them, I fusion is the fused multispectral image, I visible is a visible light image, I infrared is the infrared image, and α is the fusion weight, which is used to balance the contribution of visible light and infrared images.

[0058] S3: Extract features from the first fused image and obtain a second fused image using a second fusion method.

[0059] It should be noted that the first fused image is subjected to feature extraction using a preset extraction method to obtain physical object features and temperature features.

[0060] Specifically, the preset extraction method is to extract complementary physical features (such as edges, textures) and temperature features (such as hot spot gradients) from multispectral fusion images to provide robust data representation for subsequent fault classification, including but not limited to edge detection, principal component analysis, convolutional neural networks, texture analysis (such as gray-level co-occurrence matrix, GLCM), etc.

[0061] In an optional embodiment, the preset extraction method can be edge detection (such as the Canny operator), which detects the location of pixel brightness mutations by calculating image gradients (such as the Sobel operator), combines non-maximum suppression and double threshold screening, and retains significant edge contours; extracts physical boundaries or structural anomalies such as cracks in photovoltaic panels, and assists in locating damage, fractures and other faults.

[0062] In another optional embodiment, the preset extraction method can also be principal component analysis (PCA), which reduces the dimensionality of the multispectral image, retains the main information components, and eliminates redundancy; compresses the amount of fused image data, and at the same time highlights key spectral features related to the fault (such as the association between abnormal hot areas and textures).

[0063] An optional method in the present invention is as follows: for each pixel in the visible light image, a traditional visible light channel texture extraction value is calculated. The grayscale value of the center pixel is used as the threshold and compared with that of the surrounding eight neighboring pixels. If the neighboring pixel value is greater than or equal to the center pixel, it is marked as 1, otherwise it is marked as 0. An 8-bit binary code is generated and converted to a decimal value.

[0064] For the same pixel in the infrared image, the thermal difference pattern (TDP) is calculated, a dynamic temperature threshold is introduced, and the temperature difference between the neighboring pixels and the central pixel is compared. If the difference exceeds the dynamic temperature threshold, it is marked as 1, otherwise it is marked as 0, and an 8-bit code TDP is generated.

[0065] Furthermore, the acquiring of the second fused image includes fusing the physical feature and the temperature feature through a second fusion method to obtain a second fused image; wherein the second fused image is a temperature distribution feature image.

[0066] Specifically, the physical characteristics and temperature characteristics of the photovoltaic panel are fused to obtain a temperature distribution characteristic image of the photovoltaic panel;

[0067] Specifically, the second fusion method is to spatially align and enhance the features reflected by visible light through deep fusion at the feature level to generate a temperature distribution feature map with clear fault indication significance, including but not limited to weighted feature fusion, CNN-based feature fusion, attention mechanism fusion, feature cascade, etc.

[0068] In an optional embodiment, the second fusion method may be weighted feature fusion, which assigns weights to physical features and temperature features, and highlights the importance of temperature features through weight adjustment while retaining visible light details;

[0069] In another optional embodiment, the second fusion method can also be based on CNN feature fusion, constructing a dual-branch network, extracting multi-scale features through convolution layers respectively, and then superimposing feature maps, and using the local perception characteristics of the convolution kernel to fuse spatially related features.

[0070] An optional method in the present invention is as follows: Adaptive Dual-Modal HOG (ADM-HOG) is used as a feature extraction technology to innovatively solve the multispectral feature coupling problem by dynamically fusing visible light texture gradients and infrared temperature gradients;

[0071] The Sobel operator is used to calculate the horizontal and vertical gradients. Similarly, the gradient amplitude and direction of the infrared image are calculated, focusing on the area of ​​sudden temperature change. For each pixel, a weight is dynamically assigned based on the contribution of the two-modal gradient amplitude. The direction is converted into a unit vector and then weighted fusion is performed.

[0072] S4: The fused features are used for fault identification through hot spot detection and occlusion detection to extract fault features; fault location information is obtained in the spatial model based on the extracted fault features.

[0073] It should also be noted that hot spot detection and occlusion detection are performed on the second fused image to obtain hot spot detection results and foreign object occlusion results, and fault characteristics are determined based on the hot spot detection results and foreign object occlusion detection results;

[0074] Specifically, when performing foreign object occlusion detection, the grayscale value in the image can be detected. This process can judge the grayscale values ​​of adjacent areas and determine that the area with higher grayscale value is the occluded area. Hot spot detection uses the temperature curve graph formed by the infrared image to determine the area with sudden temperature change as the hot spot area.

[0075] Among them, the fault characteristics include externally induced faults and internal native faults. Externally induced faults are faults caused by uneven heating of photovoltaic panels due to external obstructions, and internal native faults refer to abnormal heating faults caused by short circuits and overloads of electronic components in photovoltaic panels.

[0076] It should also be noted that the steps of performing hot spot detection and occlusion detection on the fused image to obtain hot spot detection results and foreign object occlusion detection results, and determining fault characteristics based on the hot spot detection results and foreign object occlusion detection results include:

[0077] Performing hot spot detection on the fused image to determine a temperature-differentiated area in the fused image, and determining a temperature difference between the maximum temperature of the temperature-differentiated area and other temperature areas;

[0078] When the temperature difference is greater than a preset temperature difference, determining the temperature-differentiated area as a target hot spot area, and generating a hot spot detection result according to the target hot spot area;

[0079] and performing occlusion detection on the fused image to determine a shadow area in the fused image;

[0080] When the shadow area has a preset shadow feature, determining that the shadow area is a target shadow area, and generating a foreign object occlusion detection result according to the target shadow area;

[0081] When any one of the hot spot detection result and the foreign object obstruction detection result is a fault characteristic result, the fault characteristic is determined according to the hot spot detection result and / or the foreign object obstruction detection result.

[0082] In the specific implementation, the temperature difference area in the fused image is determined through infrared image analysis:

[0083] T a={(x,y)|T(x,y)>T t}

[0084] in, T a is the set of temperature anisolation regions, T (x,y) is the temperature value of point (x,y), T t is the preset temperature threshold;

[0085] Determine the temperature difference between the temperature-differentiation region and the highest temperature of other temperature regions, and when the temperature difference is greater than a preset temperature difference, determine the temperature-differentiation region as a target hot spot region;

[0086] Generate hot spot detection results based on the target hot spot area, and the temperature difference is the maximum temperature difference, that is, the difference between the highest temperature in the alienated area and the lowest temperature in the remaining areas:

[0087] ΔT max =max(T a )-min(T other )

[0088] Where, ΔT max is the maximum temperature difference, max( T a ) is the highest temperature in the temperature alienation area, min(T other ) is the lowest temperature in other areas. When determining the hot spot area, the formula is:

[0089]

[0090] Among them, H is the target hot spot area set, T p is the preset temperature threshold;

[0091] Similarly, when determining the occlusion area, according to the formula:

[0092]

[0093] Among them, S t is the target shadow area, S p is the preset shadow area threshold;

[0094] When any one of the hot spot detection result and the foreign object occlusion detection result is a fault characteristic result, the fault characteristic is determined according to the hot spot detection result or the foreign object occlusion detection result.

[0095] Furthermore, when the fault feature is a target fault feature, position information of the fault feature in the information of the first fused image is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

[0096] Example 2, reference Figure 2 The second embodiment of the present invention is different from the previous embodiment in that:

[0097] The substation intelligent patrol image recognition device may include a processing device 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 1002 or programs loaded from storage device 1003 into random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the substation intelligent patrol image recognition technology. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input device 1007, such as a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output device 1008, such as a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003, such as a magnetic tape or hard disk; and communication device 1009. The communication device 1009 can allow the substation intelligent patrol image recognition technology to communicate with other devices wirelessly or by wire to exchange data.

[0098] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, 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. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0099] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0100] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0101] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0102] Example 3, reference Figure 3 , which is the third embodiment of the present invention, provides an image recognition system for intelligent inspection of substations, including a path inspection module, an image fusion module, and a fault identification module;

[0103] The path inspection module includes an inspection path determination unit and a data acquisition unit. The inspection path determination unit generates a photovoltaic station spatial model based on spatial data and then establishes a three-dimensional spatial coordinate system, performs inspection zoning, and determines the inspection path; the data acquisition unit installs acquisition equipment in the inspection path to collect image information;

[0104] The image fusion module includes a first image fusion unit and a second image fusion unit. The first image fusion unit fuses the first image information and the second image information using a first fusion method to obtain a first fused image. The second image fusion unit extracts features from the first fused image using a preset extraction method to obtain physical object features and temperature features. The physical object features and the temperature features are fused using a second fusion method to obtain a second fused image.

[0105] The fault identification module includes a fault feature extraction unit and a fault location unit. The fault feature extraction unit performs feature extraction on the first fused image through hot spot detection and occlusion detection to obtain physical features and temperature features. The fault location unit performs feature fusion on the physical features and the temperature features through a second fusion method to obtain a second fused image, performs hot spot detection and occlusion detection on the second fused image to obtain hot spot detection results and foreign object occlusion results, and determines the fault feature based on the hot spot detection results and foreign object occlusion detection results. When the fault feature is a target fault feature, the position information of the fault feature in the first fused image information is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

[0106] Among them, the substation intelligent patrol image recognition system in the present invention can include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., as well as fixed terminals such as digital TVs, desktop computers, etc.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An image recognition method for intelligent inspection of substations, characterized by: include, Generate intelligent inspection routes based on the spatial data of photovoltaic stations acquired by drone groups; Installing a collection device in the inspection path to collect image information and obtain a first fused image using a first fusion method; performing feature extraction on the first fused image and obtaining a second fused image by a second fusion method; The fused features are used for fault identification through hot spot detection and occlusion detection to extract fault features; The fault location information is obtained in the spatial model based on the extracted fault features.

2. The image recognition method for intelligent inspection of substations according to claim 1, characterized in that: The spatial data includes first dimensional data and second dimensional data; Generate a photovoltaic station spatial model based on spatial data and then establish a three-dimensional spatial coordinate system, perform inspection zoning, and determine the inspection path; Among them, the first dimension data is the two-dimensional base map data of the photovoltaic station; The second dimension data is the three-dimensional scanning data of the photovoltaic station.

3. The image recognition method for intelligent inspection of substations according to claim 2, characterized in that: The collected image information includes first image information and second image information; The first image information is a visible spectrum image; The second image information is an infrared spectrum image.

4. The image recognition method for intelligent inspection of substations according to claim 3, characterized in that: Acquiring the first fused image by the first fusion method includes fusing the first image information and the second image information by the first fusion method to acquire the first fused image; Wherein, the first fused image is a multispectral image.

5. The image recognition method for intelligent inspection of substations according to claim 4, characterized in that: The feature extraction includes performing feature extraction on the first fused image using a preset extraction method to obtain physical object features and temperature features.

6. The image recognition method for intelligent inspection of substations according to claim 5, characterized in that: Acquiring the second fused image includes fusing the physical feature and the temperature feature using a second fusion method to obtain the second fused image; Wherein, the second fused image is a temperature distribution characteristic image.

7. The image recognition method for intelligent inspection of substations according to claim 6, characterized in that: Extracting the fault feature includes performing hot spot detection and occlusion detection on the second fused image to obtain a hot spot detection result and a foreign object occlusion result, and determining the fault feature based on the hot spot detection result and the foreign object occlusion detection result; When the fault feature is a target fault feature, position information of the fault feature in the information of the first fused image is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

8. A system using the image recognition method for intelligent inspection of substations according to any one of claims 1 to 7, characterized in that: Including path inspection module, image fusion module, and fault identification module; The path inspection module includes an inspection path determination unit and a data acquisition unit. The inspection path determination unit generates a photovoltaic station space model based on spatial data and then establishes a three-dimensional spatial coordinate system, performs inspection zoning, and determines the inspection path. The data acquisition unit installs acquisition equipment in the inspection path to collect image information; The image fusion module includes a first image fusion unit and a second image fusion unit. The first image fusion unit fuses the first image information and the second image information using a first fusion method to obtain a first fused image. The second image fusion unit extracts features from the first fused image using a preset extraction method to obtain physical object features and temperature features. fusing the physical feature and the temperature feature using a second fusion method to obtain a second fused image; The fault identification module includes a fault feature extraction unit and a fault location unit. The fault feature extraction unit performs feature extraction on the first fused image through hot spot detection and occlusion detection to obtain physical object features and temperature features. The fault location unit fuses the physical object features and the temperature features through a second fusion method to obtain a second fused image. The second fused image is subjected to hot spot detection and occlusion detection to obtain hot spot detection results and foreign object occlusion results, and the fault feature is determined based on the hot spot detection results and foreign object occlusion detection results. When the fault feature is a target fault feature, position information of the fault feature in the first fused image information is determined, fault location information is obtained based on the position information and the three-dimensional space coordinate system, and the fault location information is output.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the image recognition method in the intelligent inspection of substations according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image recognition method in the intelligent inspection of substations according to any one of claims 1 to 7 are implemented.

Images