Method and device for registering heterogeneous images of power equipment

Abstract

The application provides a registration method and device for heterogeneous images of power equipment, which comprises the following steps: calculating AGDD response normalized measure values of edge pixel points in at least two power equipment images based on an anisotropic Gaussian kernel function; for each power equipment image, calculating an angle point adaptive threshold value according to an anisotropic factor of the power equipment image; screening angle points according to the angle point adaptive threshold value of the power equipment image and the AGDD response normalized measure values of each edge pixel point; taking any angle point as a first vertex, taking minimum values on both sides of the angle point as other two vertices, constructing a Gaussian feature triangle, and determining a main direction of the angle point according to the Gaussian feature triangle; and finally constructing a PIIFD feature descriptor according to positions of each angle point in the power equipment image and the main direction of the angle point, and performing power equipment image matching. The above method can improve the accuracy of image registration.

Description

Technical Field

[0001] This invention relates to the field of image registration technology, and in particular to a method and apparatus for registering heterogeneous images of power equipment. Background Technology

[0002] The construction of digital twins for power equipment is a crucial step in promoting the digital transformation of the power grid and achieving full lifecycle management of equipment. Building high-fidelity 3D models is essential for accurate perception and prediction of equipment status. Multi-source images are vital data for constructing 3D digital twins of power equipment and realizing equipment status diagnosis. Through the registration and fusion of multi-source images, different feature information can be combined to compensate for the perception deficiencies of different sensors. Among these, high-precision infrared-visible light image registration and fusion technology, by combining thermal state information and texture information such as mechanical structure, helps to achieve precise fault location and is widely used in fault diagnosis of power equipment.

[0003] Feature extraction and orientation of power equipment images are crucial for constructing image matching descriptors, directly impacting the accuracy and recall of infrared-visible image registration. In recent years, various image feature extraction and orientation schemes have been proposed both domestically and internationally. For example, the Scale-Invariant Feature Transform (SIFT) algorithm relies on local region pixel selection of the principal feature direction, reducing the impact of scale differences in power equipment images on feature matching. The Speeded-Up Robust Features (SURF) algorithm has lower computational complexity and faster feature calculation speed compared to SIFT in building image pyramids and calculating principal feature directions. Based on the calculation of the principal direction of the squared gradient and the Partial Intensity Invariant Feature Descriptor (PIIFD), the 16-facet orientation histogram of the SIFT algorithm is transformed into an 8-facet orientation histogram, enhancing the robustness of the feature descriptor against gradient inversion. The CAO-C2F algorithm uses the principal direction angular difference of contour curvature to rotate and align power equipment images, achieving good registration of infrared and visible images.

[0004] In the task of infrared-visible image registration of the fine structure of power equipment, existing registration methods suffer from low feature point extraction accuracy and fuzzy selection of the main feature direction, resulting in low image registration accuracy. Summary of the Invention

[0005] This invention provides a method and apparatus for registering heterogeneous images of power equipment to solve the problem of low registration accuracy of heterogeneous images.

[0006] In a first aspect, embodiments of the present invention provide a registration method for heterogeneous images of power equipment, comprising:

[0007] Based on the anisotropic Gaussian kernel function, calculate the AGDD response normalization measure of each edge pixel in at least two power equipment images;

[0008] For each power equipment image, an adaptive threshold for corner points is calculated based on the anisotropy factor of the power equipment image.

[0009] Corner points are selected from the edge pixels of the power equipment image based on the corner adaptive threshold and the AGDD response normalization measure of each edge pixel in the power equipment image.

[0010] For any corner point in the image of the power equipment, take the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices to construct a Gaussian feature triangle. Determine the main direction of the corner point based on the Gaussian feature triangle.

[0011] Based on the position and main direction of each corner point in the power equipment image, PIIFD feature descriptors are constructed, and power equipment image matching is performed based on the PIIFD feature descriptors of any two power equipment images.

[0012] Secondly, embodiments of the present invention provide a registration device for heterogeneous images of power equipment, comprising:

[0013] The AGDD response normalization metric calculation module is used to calculate the AGDD response normalization metric of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function.

[0014] The threshold calculation module is used to calculate the adaptive threshold for corner points based on the anisotropy factor of each power equipment image.

[0015] The corner filtering module is used to filter corners from the edge pixels of the power equipment image based on the corner adaptive threshold and the AGDD response normalization measure value of each edge pixel in the power equipment image.

[0016] The main direction determination module is used to construct a Gaussian feature triangle for any corner point in the image of the power equipment, taking the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices, and determine the main direction of the corner point based on the Gaussian feature triangle.

[0017] The heterogeneous image matching module is used to construct PIFD feature descriptors based on the position and main direction of each corner point in the power equipment image, and to perform power equipment image matching based on the PIFD feature descriptors of any two power equipment images.

[0018] This invention provides a registration method and apparatus for heterogeneous images of power equipment. The method first calculates the AGDD response normalized measure values ​​of each edge pixel in at least two power equipment images based on anisotropic Gaussian kernel functions, wherein any two power equipment images are from different sources. For each power equipment image, an adaptive corner threshold is calculated based on the anisotropy factor of the power equipment image. Based on the adaptive corner threshold and the AGDD response normalized measure values ​​of each edge pixel in the power equipment image, corner points are selected from the edge pixels of the power equipment image. For any corner point in the power equipment image, a Gaussian feature triangle is constructed using the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalized measure values ​​are minimized as the other two vertices. The principal direction of the corner point is determined based on the Gaussian feature triangle. Finally, a PIIFD feature descriptor is constructed based on the position and principal direction of each corner point in the power equipment image, and power equipment image matching is performed based on the PIIFD feature descriptors of any two power equipment images. The above method can filter corners based on the adaptive threshold of corners in each power equipment image, thereby reducing the false extraction rate and missed extraction rate of corners in the fine structure of power equipment and improving the accuracy of image registration. Attached Figure Description

[0019] 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.

[0020] Figure 1 This is a flowchart illustrating the implementation of the registration method for heterogeneous images of power equipment provided in this embodiment of the invention.

[0021] Figure 2 This is a schematic diagram of the method for selecting Gaussian characteristic triangles provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the structure of the registration device for heterogeneous images of power equipment provided in an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of the terminal provided in an embodiment of the present invention. Detailed Implementation

[0024] 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.

[0025] 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.

[0026] See Figure 1 The document illustrates a flowchart of the registration method for heterogeneous images of power equipment provided in an embodiment of the present invention, which is described in detail below:

[0027] S101: Based on the anisotropic Gaussian kernel function, calculate the AGDD response normalization measure of each edge pixel in at least two power equipment images; wherein any two power equipment images are images from different sources.

[0028] In this embodiment, images of power equipment from different sources may include infrared images and visible light images of the power equipment.

[0029] Specifically, the set of power equipment images is K, where ki(x)∈K. Here, ki(x) represents the i-th power equipment image in the set, x=[nx,ny]T. A filtering window is generated using anisotropic Gaussian kernels (ANGKs). The anisotropic Gaussian kernel function is defined as:

[0030]

[0031] In equation (1), ρ represents the anisotropy factor; σ represents the Gaussian scale; θ represents the Gaussian kernel direction, and θ = (l-1)²π / L, l = 1, 2, ..., L; L represents the total number of directional derivative operators; R(θ) represents the rotation matrix with direction θ; R(-θ) represents the rotation matrix with direction -θ; g [σ,ρ,θ] (x) represents the anisotropic Gaussian kernel function, and H represents the anisotropic factor coefficient matrix.

[0032] Since image data exists as a two-dimensional discrete signal, discretizing the anisotropic Gaussian kernel function and its directional derivative yields:

[0033]

[0034] Where n represents the image k iThe number of pixels, n = [n x ,n y ] T , n∈Z.

[0035] The anisotropic Gaussian directional derivative can be obtained by differentiating formula (2) with respect to π / 2 + θ:

[0036]

[0037] Input power equipment image k i The anisotropic Gaussian directional derivative response of (x) is:

[0038]

[0039] Where, k i (n-n1) represents the value of the corresponding pixel n in the pixel matrix of the original power equipment image obtained by successively subtracting the corresponding position in the AGDD response matrix from the center pixel position n1 of the filter window.

[0040] In one possible implementation, the specific implementation process of S101 includes:

[0041] S201: Based on the anisotropic Gaussian kernel function, calculate the anisotropic Gaussian directional derivative response of each edge pixel in at least two power equipment images;

[0042] S202: Standardize the anisotropic Gaussian directional derivative response of each edge pixel in the image of the power equipment to obtain the standardized AGDD response measure of each edge pixel in the image of the power equipment.

[0043] In one possible implementation, the specific implementation process of S201 includes:

[0044] According to the formula Calculate the anisotropic Gaussian directional derivative response of each edge pixel in the target power equipment image; the target power equipment image is any power equipment image.

[0045] Where λ represents the anisotropic Gaussian directional derivative; ρ represents the anisotropy factor; σ represents the Gaussian scale; θ represents the Gaussian kernel direction, and θ = (l-1)²π / L, l = 1, 2, ..., L; L represents the total number of directional derivative operators; R(θ) represents the rotation matrix with direction θ; n1 represents the center pixel position of the filter window; n represents the image k. i The number of pixels, n = [n x ,n y ] T ; λ represents the derivative response of edge pixel n in the image of the target power equipment along the l-th anisotropic Gaussian sampling direction;[σ，ρ，θ] (n) represents the anisotropic Gaussian directional derivative of edge pixel n in the image of the target power equipment; where,

[0046] In one possible implementation, the specific implementation process of S202 includes:

[0047] Based on formula Calculate the normalized AGDD response measure of each edge pixel in the image of the power equipment;

[0048] in, Let ξ(n) represent the derivative response of edge pixel n in the l-th anisotropic Gaussian sampling direction in the image of the target power equipment, let ξ(n) represent the normalized AGDD response measure of edge pixel n, and let L represent the total number of directional derivative operators.

[0049] S102: For each power equipment image, calculate the corner adaptive threshold based on the anisotropy factor of the power equipment image.

[0050] In one possible implementation, the specific implementation process of S102 includes:

[0051] According to the formula Calculate the adaptive threshold for corner points in the image of the power equipment;

[0052] Where ρ represents the anisotropy factor of the power equipment image; F represents the corner adaptive threshold of the power equipment image.

[0053] Specifically, common corner types can be represented by superimposing multiple simple corner models, so the general corner model can be represented as: Where C(r,a) represents a simple corner model, r represents the polar radius corresponding to the corner, α represents the polar angle, and T p This represents the grayscale information of the p-th region surrounding the pixel; P represents the number of regions of the pixel, and a p This represents the polar angle of the p-th grayscale region.

[0054] The AGDD response of the universal corner point is then expressed as:

[0055]

[0056] In equation (5), Let T represent the corner model, λ represent the anisotropic Gaussian directional derivative, and T represent the corner model. p-1 This represents the grayscale information of the (p-1)th region surrounding the pixel, where ρ represents the anisotropy factor, σ represents the Gaussian scale, and θ represents the Gaussian kernel direction; λ [σ,ρ,θ](-r,-α) represents the anisotropic Gaussian directional derivative of edge pixel n in the polar coordinates of the target power equipment image; This represents the AGDD response for a common corner point.

[0057] Compared to the AGDD response of image edges, the corner model has a larger maximum and minimum response values ​​and exhibits extreme values ​​in multiple directions. When a1-a2=π, this corner model is the edge model, and its AGDD response is:

[0058]

[0059] The AGDD response has two extreme points at θ = a1 and θ = π + a1, with magnitudes of [value missing].

[0060]

[0061] Since edge nodes only have two regions, in the AGDD response of the image edge, T1 represents one region around the edge node and T2 represents another region around the edge node.

[0062] Specifically, pixels in an image where grayscale changes drastically in all directions are considered corner points. Common corner points exhibit large AGDD response amplitudes in almost all directions, while edge pixels only show large AGDD response amplitudes in the gradient direction. To extract a complete, effective, and accurate set of corresponding corner points from infrared and visible light images, this application traverses every pixel on the edge contour of the power equipment image, calculates the anisotropic Gaussian response (AGDD) of each edge pixel, standardizes it, and sums the results to obtain an adaptive corner point threshold. This overcomes the limitations of empirical thresholds and reduces the rate of missed or incorrect corner point detection.

[0063] Specifically, the formula for calculating the adaptive corner threshold of an electronic device image is as follows:

[0064]

[0065] in, This represents the AGDD response of edge pixels.

[0066] S103: Based on the corner adaptive threshold of the power equipment image and the AGDD response normalization measure of each edge pixel in the power equipment image, corner points are selected from the edge pixels of the power equipment image.

[0067] In one possible implementation, the specific implementation process of S103 includes:

[0068] S301: For each edge pixel in the image of the power equipment, subtract the corner adaptive threshold of the power equipment image from the AGDD response normalization measure value of the edge pixel to obtain the corner gating value;

[0069] S302: If the corner threshold value of the edge pixel is greater than zero, then the edge pixel is determined to be a corner.

[0070] In one possible implementation, after S302, the method provided in this embodiment further includes:

[0071] The non-maximum suppression method is used to filter out the final corner points from each corner point of the power equipment image.

[0072] In this embodiment, flow control is performed on edge pixels using the ReLU activation function. When the normalized AGDD response measure of an edge pixel is less than the corner adaptive threshold F, the neuron is not activated, the dot product of the normalized AGDD response measure and the corner gate value is 0, the flow gate is closed, and the pixel is identified as an edge pixel. When the normalized AGDD response measure of an edge pixel is greater than the corner adaptive threshold F, the neuron is activated, the corner verification passes, and a candidate corner set is generated. Non-maximum suppression is applied to the candidate corner set to select the final corner sets from two heterogeneous power equipment images.

[0073] S104: For any corner point in the image of the power equipment, take the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices, construct a Gaussian feature triangle, and determine the main direction of the corner point based on the Gaussian feature triangle.

[0074] S105: Construct PIIFD feature descriptors based on the positions of each corner point and its main direction in the power equipment image, and perform power equipment image matching based on the PIIFD feature descriptors of any two power equipment images.

[0075] In one possible implementation, prior to S101, the method provided in this embodiment further includes:

[0076] The Canny detector is used to extract edge pixels from the images of each power device.

[0077] In one possible implementation, the specific implementation process of S104 includes:

[0078] The side formed by the other two vertices is taken as the base of the Gaussian characteristic triangle, and the median of the Gaussian triangle is taken as the main direction of the corner point.

[0079] In this embodiment, after determining the final set of corner points, a feature direction needs to be selected for each corner point to achieve feature matching. Specifically, the AGDD responses of the corner point sets and edge pixels in the infrared and visible light images have been obtained. The main direction selection is illustrated below. Figure 2 As shown, for any corner point O, the minimum value ξ of the normalized measure of the AGDD response pointing from corner point O to both sides is defined. min1 ξ min2 The feature vectors of the corresponding pixels are used as edges to construct Gaussian feature triangles. The direction from the corner point O to the center point E is taken as the principal direction of the corner point, where the center point E is the intersection of the median and the base of the Gaussian feature triangle.

[0080] In this embodiment, the specific implementation process of S105 includes:

[0081] Based on the corner positions and corresponding principal directions of the infrared image of the power equipment, a PIIFD feature descriptor is constructed. Based on the corner positions and corresponding principal directions of the visible light image of the power equipment, a PIIFD feature descriptor is constructed. Based on the PIIFD feature descriptors of the infrared image and the visible light image, feature matching is performed to complete the heterogeneous image registration.

[0082] As can be seen from the above embodiments, firstly, the method provided in this embodiment analyzes the AGDD response characteristics of image edge pixels to adaptively set the corner point adaptive threshold, thereby reducing the false extraction rate and missed extraction rate of corner points in the fine structure of power equipment. Secondly, the algorithm proposed in this embodiment utilizes the edge shape characteristics and the relevant information between edge pixels and surrounding pixels to provide a main direction with high similarity, avoiding the problem of differences in spectrum and resolution between heterogeneous images, thereby improving the accuracy of main direction selection. In addition, the embodiments of this application can effectively extract and orient feature corner points, thereby improving the infrared-visible image registration accuracy of the fine structure of power equipment.

[0083] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0084] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0085] Figure 3 A schematic diagram of the registration device for heterogeneous images of power equipment provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below:

[0086] like Figure 3 As shown, the registration device 100 for heterogeneous images of power equipment includes:

[0087] The AGDD response normalization metric calculation module 110 is used to calculate the AGDD response normalization metric of each edge pixel in at least two power equipment images based on anisotropic Gaussian kernel functions; wherein any two power equipment images are images from different sources;

[0088] Threshold calculation module 120 is used to calculate the corner adaptive threshold based on the anisotropy factor of each power equipment image for each power equipment image.

[0089] The corner point filtering module 130 is used to filter corner points from the edge pixels of the power equipment image based on the corner point adaptive threshold and the AGDD response normalization measure value of each edge pixel in the power equipment image.

[0090] The main direction determination module 140 is used to construct a Gaussian feature triangle for any corner point in the image of the power equipment, taking the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices, and determine the main direction of the corner point based on the Gaussian feature triangle.

[0091] The heterogeneous image matching module 150 is used to construct PIFD feature descriptors based on the position of each corner point and its main direction in the power equipment image, and to perform power equipment image matching based on the PIFD feature descriptors of any two power equipment images.

[0092] In one possible implementation, the AGDD response normalization measure calculation module 110 includes:

[0093] The derivative response calculation unit is used to calculate the anisotropic Gaussian directional derivative response of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function.

[0094] The normalization unit is used to normalize the anisotropic Gaussian directional derivative response of each edge pixel in the image of the power equipment, so as to obtain the normalized AGDD response measure value of each edge pixel in the image of the power equipment.

[0095] In one possible implementation, the derivative response calculation unit includes:

[0096] According to the formula Calculate the anisotropic Gaussian directional derivative response of each edge pixel in the target power equipment image; the target power equipment image is any power equipment image.

[0097] Where λ represents the anisotropic Gaussian directional derivative; ρ represents the anisotropy factor; σ represents the Gaussian scale; θ represents the Gaussian kernel direction, and θ = (l-1)²π / L, l = 1, 2, ..., L; L represents the total number of directional derivative operators; R(θ) represents the rotation matrix with direction θ; n1 represents the center pixel position of the filter window; n represents the image k. i The number of pixels, n = [n x ,n y ] T ; λ represents the derivative response of edge pixel n in the image of the target power equipment along the l-th anisotropic Gaussian sampling direction; [σ，ρ，θ] (n) represents the anisotropic Gaussian directional derivative of edge pixel n in the image of the target power equipment;

[0098]

[0099] In one possible implementation, the standardization unit includes:

[0100] Based on formula Calculate the normalized AGDD response measure of each edge pixel in the image of the power equipment;

[0101] in, Let ξ(n) represent the derivative response of edge pixel n in the l-th anisotropic Gaussian sampling direction in the image of the target power equipment, let ξ(n) represent the normalized AGDD response measure of edge pixel n, and let L represent the total number of directional derivative operators.

[0102] In one possible implementation, the threshold calculation module 120 includes:

[0103] According to the formula Calculate the adaptive threshold for corner points in the image of the power equipment;

[0104] Where ρ represents the anisotropy factor of the power equipment image; F represents the corner adaptive threshold of the power equipment image.

[0105] In one possible implementation, the corner filtering module 130 includes:

[0106] The corner gating value calculation unit is used to subtract the corner adaptive threshold of the power equipment image from the AGDD response normalization measure value of each edge pixel in the power equipment image to obtain the corner gating value.

[0107] The corner detection unit is used to determine that the edge pixel is a corner if the corner threshold value of the edge pixel is greater than zero.

[0108] In one possible implementation, the corner filtering module 130 further includes a non-maximum suppression unit, used for:

[0109] The non-maximum suppression method is used to filter out the final corner points from each corner point of the power equipment image.

[0110] In one possible implementation, the registration device 100 for heterogeneous images of power equipment further includes an edge pixel extraction module, used for:

[0111] The Canny detector is used to extract edge pixels from the images of each power device.

[0112] In one possible implementation, the main direction determination module 140 includes:

[0113] The side formed by the other two vertices is taken as the base of the Gaussian characteristic triangle, and the median of the Gaussian triangle is taken as the main direction of the corner point.

[0114] Figure 4 This is a schematic diagram of a terminal provided in an embodiment of the present invention. Figure 4 As shown, the terminal 4 in this embodiment includes a processor 40 and a memory 41. The memory 41 stores a computer program 42, and the processor 40 calls and runs the computer program 42 stored in the memory 41 to execute the steps in the above embodiments of the registration method for heterogeneous images of power equipment, for example... Figure 2 Steps S101 to S105 are shown. Alternatively, the processor 40 is used to call and run the computer program 42 stored in the memory 41 to implement the functions of each module / unit in the above-described device embodiments, for example... Figure 4 The functions of modules 110 to 150 are shown.

[0115] For example, the computer program 42 can be divided into one or more modules / units, which are stored in the memory 41 and executed by the processor 40 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 42 in the terminal 4. For example, the computer program 42 can be divided into... Figure 4 Modules 110 to 150 are shown.

[0116] The terminal 4 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The terminal 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that... Figure 4This is merely an example of terminal 4 and does not constitute a limitation on terminal 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal may also include input / output devices, network access devices, buses, etc.

[0117] The processor 40 may be a Central Processing Unit (CPU), or 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. A general-purpose processor may be a microprocessor or any conventional processor.

[0118] The memory 41 can be an internal storage unit of the terminal 4, such as a hard disk or memory of the terminal 4. The memory 41 can also be an external storage device of the terminal 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal 4. Furthermore, the memory 41 can include both internal storage units and external storage devices of the terminal 4. The memory 41 is used to store the computer program and other programs and data required by the terminal. The memory 41 can also be used to temporarily store data that has been output or will be output.

[0119] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0120] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0121] 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, or a combination of computer software and electronic hardware. 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.

[0122] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or 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. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0123] 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 this embodiment according to actual needs.

[0124] 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.

[0125] If the integrated module / unit is implemented as 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, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. 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 above embodiments of the registration method for heterogeneous images of various power devices. 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 the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately added to or subtracted from the content as required by the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium may not include electrical carrier signals and telecommunication signals.

[0126] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for registering heterogeneous images of power equipment, characterized in that, include: Based on the anisotropic Gaussian kernel function, calculate the AGDD response normalization measure of each edge pixel in at least two power equipment images; For each power equipment image, an adaptive threshold for corner points is calculated based on the anisotropy factor of the power equipment image. Corner points are selected from the edge pixels of the power equipment image based on the corner adaptive threshold and the AGDD response normalization measure of each edge pixel in the power equipment image. For any corner point in the image of the power equipment, take the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices to construct a Gaussian feature triangle. Determine the main direction of the corner point based on the Gaussian feature triangle. Based on the position of each corner point and its main direction in the power equipment image, construct the PIFD feature descriptor, and perform power equipment image matching based on the PIFD feature descriptor of any two power equipment images. The step of determining the principal direction of the corner point based on the Gaussian characteristic triangle includes: The side formed by the other two vertices is taken as the base of the Gaussian characteristic triangle, and the median of the Gaussian triangle is taken as the main direction of the corner point.

2. The registration method for heterogeneous images of power equipment according to claim 1, characterized in that, The calculation of the AGDD response normalization measure of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function includes: Based on the anisotropic Gaussian kernel function, calculate the anisotropic Gaussian directional derivative response of each edge pixel in at least two power equipment images; The anisotropic Gaussian directional derivative response of each edge pixel in the image of the power equipment is standardized to obtain the standardized AGDD response measure of each edge pixel in the image of the power equipment.

3. The registration method for heterogeneous images of power equipment according to claim 2, characterized in that, The calculation of the anisotropic Gaussian directional derivative response of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function includes: According to the formula Calculate the anisotropic Gaussian directional derivative response of each edge pixel in the target power equipment image; the target power equipment image is any power equipment image. in, This represents the directional derivative of anisotropic Gaussians; Represents the anisotropy factor; Indicates Gaussian scale; Indicates the direction of the Gaussian kernel, and ; L This indicates the total number of directional derivative operators; Indicates direction as The rotation matrix; n1 represents the center pixel position of the filter window; n represents the image. k i pixels, ; The nth edge pixel in the image of the target power equipment l Derivative response of an anisotropic Gaussian sampling direction; Let represent the anisotropic Gaussian directional derivative of edge pixel n in the image of the target power equipment; where, .

4. The registration method for heterogeneous images of power equipment according to claim 2, characterized in that, The process of standardizing the anisotropic Gaussian directional derivative response of each edge pixel in the image of the power equipment to obtain the standardized AGDD response measure value of each edge pixel in the image of the power equipment includes: Based on formula Calculate the AGDD response normalized measure value of each edge pixel in the image of the power equipment; in, The nth edge pixel in the image of the target power equipment l The derivative response of an anisotropic Gaussian sampling direction This represents the normalized AGDD response measure of edge pixel n. L This represents the total number of directional derivative operators.

5. The registration method for heterogeneous images of power equipment according to claim 1, characterized in that, The step of calculating the adaptive corner threshold based on the anisotropy factor of each power equipment image includes: According to the formula Calculate the adaptive threshold for corner points in the image of the power equipment; in, The anisotropy factor represents the image of the power equipment; F This represents the adaptive threshold for corner points in the image of the power equipment.

6. The registration method for heterogeneous images of power equipment according to claim 1, characterized in that, The step of filtering corner points from the edge pixels of the power equipment image based on the corner adaptive threshold and the AGDD response normalized measure value of each edge pixel in the power equipment image includes: For each edge pixel in the image of the power equipment, the corner adaptive threshold value of the power equipment image is subtracted from the AGDD response normalization measure value of the edge pixel to obtain the corner gating value. If the corner threshold value of the edge pixel is greater than zero, then the edge pixel is determined to be a corner.

7. The registration method for heterogeneous images of power equipment according to claim 6, characterized in that, After determining that an edge pixel is a corner if its corner threshold value is greater than zero, the method further includes: The non-maximum suppression method is used to filter out the final corner points from each corner point of the power equipment image.

8. The registration method for heterogeneous images of power equipment according to claim 1, characterized in that, Before calculating the anisotropic Gaussian directional derivative response of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function, the method further includes: The Canny detector is used to extract edge pixels from the images of each power device.

9. A registration device for heterogeneous images of power equipment, characterized in that, include: The AGDD response normalization metric calculation module is used to calculate the AGDD response normalization metric of each edge pixel in at least two power equipment images based on the anisotropic Gaussian kernel function. The threshold calculation module is used to calculate the adaptive threshold for corner points based on the anisotropy factor of each power equipment image. The corner filtering module is used to filter corners from the edge pixels of the power equipment image based on the corner adaptive threshold and the AGDD response normalization measure value of each edge pixel in the power equipment image. The main direction determination module is used to construct a Gaussian feature triangle for any corner point in the image of the power equipment, taking the corner point as the first vertex and the edge pixels on both sides of the corner point whose AGDD response normalization measure value is the minimum as the other two vertices, and determine the main direction of the corner point based on the Gaussian feature triangle. The heterogeneous image matching module is used to construct PIIFD feature descriptors based on the position of each corner point and its main direction in the power equipment image, and to perform power equipment image matching based on the PIIFD feature descriptors of any two power equipment images. The main direction determination module is specifically used to: take the side formed by the other two vertices as the base of the Gaussian feature triangle, and take the median of the Gaussian triangle as the main direction of the corner point.

Images