Substation UAV Inspection Image Edge Recognition Method, System and Storage Medium

Abstract

The present invention relates to the technical field of image recognition, and discloses a method, a system and a storage medium for edge recognition of inspection images of a substation by an unmanned aerial vehicle. The method includes: determining the gray value of an inspection image collected by the unmanned aerial vehicle, and reconstructing the gray value to obtain the pixel gray value of the inspection image; determining the edge point set of the inspection image according to the pixel gray value of the inspection image; determining the edge point threshold information of the inspection image according to the edge point set; and determining the edge information of the inspection image based on the edge point threshold information and a preset convolutional network model, so as to quickly and accurately identify the edge information.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular, to a method, a system, and a storage medium for edge recognition of substation drone inspection images. Background Art

[0002] The safe and stable operation of a substation is the foundation of the national power grid construction and, more importantly, the foundation for ensuring the normal operation of the power system. The inspection service of the substation is one of the key tasks for ensuring the safe and stable operation of the power grid. By monitoring the operation status of the substation in real time and grasping the real-time operation situation of the substation, defects in the operation of the power grid system can be discovered in a timely manner, and early warnings can be issued in a timely manner to avoid safety accidents.

[0003] Currently, most substations adopt the method of drone inspection. During the inspection process, the target image information of the substation is collected to ensure that the target image information is always on the inspection route. Since the drone remains in a flying state during the inspection process, the collected images are all dynamic images, which contain a lot of invalid information, resulting in the inability to accurately identify substation faults. Summary of the Invention

[0004] The present invention provides a method, a system, and a storage medium for edge recognition of substation drone inspection images to solve the problems existing in the prior art.

[0005] To achieve the above object, the present invention is implemented through the following technical solutions:

[0006] In a first aspect, the present invention provides a method for edge recognition of substation drone inspection images, including:

[0007] S1: Determine the gray value of the inspection image collected by the drone, and reconstruct the gray value to obtain the pixel gray value of the inspection image;

[0008] S2: Determine the edge point set of the inspection image according to the pixel gray value of the inspection image;

[0009] S3: Determine the edge point threshold information of the inspection image according to the edge point set;

[0010] S4: Determine the edge information of the inspection image based on the edge point threshold information and a preset convolutional network model.

[0011] Optionally, the S1 includes:

[0012] S11: Set the local similarity function of the dynamic inspection image as H ij , which is expressed by the following formula:

[0013]

[0014] In the formula, H s-ij represents the pixel value similarity between the inspection images i and j, and H g-ij represents the gray value similarity between the inspection images i and j in space;

[0015] The similarity of the gray values of the UAV inspection images is expressed as:

[0016]

[0017] In the formula, x i represents the pixel gray value of the dynamic inspection image in the core area, and x j represents the gray value collected by the inspection image in the core area, λ g represents the influencing factor of the pixel value similarity of the inspection image, and δ g-i represents the density function, which is defined as follows:

[0018]

[0019] In the formula, N i represents the pixel points of the dynamic inspection image i collected by the UAV in the adjacent area, and N R represents the number of pixel points of the dynamic inspection image i collected by the UAV in the adjacent area, and δ g-i represents the density function of the dynamic inspection image collected by the UAV in the core area;

[0020] The reconstruction of the gray value of the inspection image is expressed as:

[0021]

[0022] In the formula, η i represents the gray value of the i-th image pixel in the reconstructed inspection image η.

[0023] Optionally, the S2 includes:

[0024] S21: Describe each pixel within the specification of M×N in the inspection image group as follows:

[0025] p(x,y) = f R (x,y)i + f G (x,y)j + f B (x,y)k;

[0026] In the formula, (x,y) represents the image position of the inspection image within M×N, and f R (x,y), f G (x,y), f B (x,y) respectively represent the specific pixel values of R, G, and B in the inspection image, and (i,j,k) represent 3 virtual units of the spatial features of the inspection image;

[0027] S22: Normalize each pixel p(x, y) in the group of inspection images, and calculate the Grassmann product of each dynamic inspection image r and the image gray value r0:

[0028] P r P r0 = -q1·q2 + q1×q2;

[0029] In the formula, q1 and q2 represent the pixel values of the dynamic inspection image, and P r P r0 includes a pixel value S[P r P r0 and a dynamic image V[P r P r0 ;

[0030] S23: Determine the similarity discrimination function between r and r0:

[0031]

[0032] In the formula, t represents the gray value of the inspection image. Based on the similarity discrimination function, calculate the value of the dynamic inspection image pixel r0 in the USAN region as n(r0), and compare n(r0) with the vector g to determine the edge point set E of the UAV inspection image.

[0033] Optionally, the S3 includes:

[0034] S31: Calculate the edge point threshold of the inspection image, and define the threshold relationship formula between the edge points and non-edge points in the inspection image, represents the effective area of the inspection image, used to describe the number of pixels with a USAN area of in the group of inspection images, represents the total number of pixels in the image group, then occupies the probability of the effective edge points of the UAV inspection image is expressed as follows:

[0035]

[0036] In the formula, ω1 and ω2 respectively represent the area and number of effective edge information in the UAV inspection image, and μ1 and μ2 represent the proportion of effective edge information in the UAV inspection image;

[0037] S32: Calculate the variance of the effective edge points and non-edge points in the UAV inspection image

[0038]

[0039] In the formula, μ represents the edge mean value of the UAV inspection image. As can be seen from the above formula, the larger the variance value of the inspection image, the greater the threshold difference between the two. Define the threshold at this time as κ to obtain the threshold information of the UAV inspection image and achieve effective segmentation of non-edge points and weak edge points in the UAV inspection image:

[0040]

[0041] In the formula, R(P) represents the edge point threshold function information of the UAV inspection image, and g l represents the lower limit of the effective segmentation information of the edge points of the inspection image, and g h represents the upper limit of the effective segmentation information of the edge points of the inspection image.

[0042] Optionally, the S4 includes:

[0043] S41: Set the coefficient of variation of the inspection image as follows:

[0044] C ij =Δ ij / I ij ;

[0045] In the formula, Δ ij and I ij respectively represent the standard variance and threshold of the edge ij of the inspection image. Among them, C ij =β ij , and β ij represents the connection strength of the inspection image. Then the discrete matrix of the inspection image is as follows:

[0046]

[0047] In the formula, n represents the number of iterations of the inspection image, F ij [n] and U ij [n] respectively represent the information value and dynamic value of the edge of the inspection image, θ ij [n] represents the weight of the edge of the inspection image, α and β represent the weight membership degree and connection coefficient of the edge points of the inspection image, θ0 represents the gray value of the edge of the inspection image. When β ij ≠0, use the preset convolutional network model to obtain the edge information of the inspection image, define the edge matrix T, and perform iterative processing on the UAV inspection image. There is the following expression:

[0048]

[0049] S42: Define the discrete mass points in the plane of the inspection image; and perform normalization processing on the edge information of the inspection image;

[0050] S43: Based on the grayscale values of the UAV inspection images obtained above, determine a model for the adaptive recognition of the edge information of the substation UAV inspection images, and the expression is as follows:

[0051] δ = i c0 + j c0 +(T ij [n]·θ0);

[0052] In the formula, δ represents the edge recognition result, and i c0 , j c0 respectively represent the results of normalizing the edge information of the inspection image, and T ij [n] represents the result of iterative processing.

[0053] In a second aspect, the present application also provides a UAV inspection image edge recognition system, including: a processor and a memory; the memory is used to store a computer program, and the processor runs the computer program to enable the mobile terminal to execute the method steps described above.

[0054] In a third aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the method steps described above are implemented.

[0055] Beneficial effects:

[0056] For the substation UAV inspection image edge recognition method provided by the present invention, first determine the grayscale values of the inspection images collected by the UAV, and reconstruct the grayscale values to obtain the pixel grayscale values of the inspection images; determine the edge point set of the inspection images according to the pixel grayscale values of the inspection images; determine the edge point threshold information of the inspection images according to the edge point set; and determine the edge information of the inspection images based on the edge point threshold information and a preset convolutional network model. In this way, on the basis of calculating the similarity of the grayscale values of the UAV inspection images and reconstructing the image pixel grayscale values, detect the edge area of the image by segmenting the non-edge points and weak edges of the image, and then use the convolutional network to obtain the edge information of the image, and recognize the edge information on the basis of normalization processing. The edge information can be recognized quickly and accurately. Description of the drawings

[0057] Figure 1 is a flowchart of a substation UAV inspection image edge recognition method provided by a preferred embodiment of the present invention;

[0058] Figure 2 is a test result diagram of the method provided by a preferred embodiment of the present invention and the method of the prior art in terms of the discrete coefficient of image edge recognition;

[0059] Figure 3This is a test result graph of the method provided by the preferred embodiment of the present invention and the method of the prior art in terms of the high-quality coefficient of image edge recognition. Detailed implementation manners

[0060] The technical solutions of the present invention will be described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts belong to the scope of protection of the present invention.

[0061] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the art to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Similarly, the terms such as "a" or "one" do not denote a quantity limitation, but mean that there is at least one. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0062] Please refer to Figure 1 , the embodiment of the present application provides a method for edge recognition of substation drone inspection images, including:

[0063] S1: Determine the gray value of the inspection image collected by the drone, and reconstruct the gray value to obtain the pixel gray value of the inspection image;

[0064] S2: Determine the edge point set of the inspection image according to the pixel gray value of the inspection image;

[0065] S3: Determine the edge point threshold information of the inspection image according to the edge point set;

[0066] S4: Determine the edge information of the inspection image based on the edge point threshold information and a preset convolutional network model.

[0067] The above method for edge recognition of substation drone inspection images, on the basis of calculating the gray value similarity of the drone inspection image and reconstructing the image pixel gray value, detects the edge area of the image by segmenting non-edge points and weak edges of the image, and then obtains the edge information of the image by using a convolutional network, and recognizes the edge information on the basis of normalization processing. It can quickly and accurately recognize the edge information.

[0068] Specifically, assuming that the grayscale values of two dynamic inspection images collected by the substation UAV are relatively close, these two dynamic inspection images can be classified into the same category. The local similarity function of the dynamic inspection image is set as H ij , which can be expressed by the following formula:

[0069]

[0070] In formula (1), H s-ij represents the pixel value similarity between the UAV inspection images i and j. H g-ij represents the grayscale value similarity between the UAV inspection images i and j in space.

[0071] The similarity H of pixel values s-ij is expressed as:

[0072]

[0073] In formula (2), λ k represents the influencing factor of the pixel value similarity of the UAV inspection image and is the factor determining the conversion of H s-ij . The value of λ k is affected by the pixel value similarity of the inspection image.

[0074] The similarity of the grayscale value of the UAV inspection image is expressed as:

[0075]

[0076] In formula (3), x i represents the pixel grayscale value of the dynamic inspection image in the core area, and x j represents the grayscale value collected by the inspection image in the core area. λ g represents the influencing factor of the pixel value similarity of the inspection image, and δ g-i can be defined as:

[0077]

[0078] In formula (4), N i represents the pixel points of the dynamic inspection image i collected by the UAV in the adjacent area, and N R represents the number of pixel points of the dynamic inspection image i collected by the UAV in the adjacent area. δ g-i represents the density function of the dynamic inspection image collected by the UAV in the core area.

[0079] The reconstruction of the grayscale value of the UAV inspection image is expressed as:

[0080]

[0081] In formula (5), ηi It represents the gray value of the i-th image in the reconstructed inspection image η. The gray values of the pixels of the UAV inspection image are reconstructed through the above process.

[0082] According to the spatial characteristics of the UAV inspection image, algebraic operations are used to detect the image edges of the inspection image group. Each pixel of the dynamic inspection image in the UAV inspection image group within M×N is described as follows:

[0083] p(x,y) = f R (x,y)i + f G (x,y)j + f B (x,y)k (6)

[0084] In formula (6), (x,y) represents the image position of the UAV dynamic inspection image within M×N, and f R (x,y), f G (x,y), f B (x,y) respectively represent the specific pixel values of R, G, and B in the UAV dynamic inspection image, and (i,j,k) represent 3 virtual units of the spatial characteristics of the inspection image. Normalize each pixel p(x,y) in the UAV inspection image group, and calculate the Grassmann product of each dynamic inspection image r and the image gray value r0:

[0085] P r P r0 = -q1·q2 + q1×q2 (7)

[0086] In formula (7), q1 and q2 represent the pixel values of the dynamic inspection image, and P r P r0 includes a pixel value S[P r P r0 and a dynamic image V[P r P r0 . After normalization, the discriminant variance of the gray values r and r0 of the UAV inspection image group is (-1, 0). It can be seen that the higher the similarity between the two UAV inspection dynamic images, the closer the value of the discriminant variance is to (-1, 0). The similarity discriminant function between r and r0 is represented by the following formula:

[0087]

[0088] Based on the similarity discrimination function given by the above formula, calculate the value of the pixel r0 in the USAN region of the dynamic inspection image as n(r0), compare n(r0) with the vector g, and determine the edge point set E of the UAV inspection image [8]. S and V respectively represent the clear image and the failure image in the UAV inspection image, and t represents the gray value of the inspection image. Then, use the Otsu algorithm to calculate the edge point threshold of the UAV inspection image, and define the threshold relationship formula between the edge points and non-edge points in the UAV inspection image. represents the effective area of the UAV inspection image. used to describe the number of pixels with a USAN area of in the UAV inspection image group. represents the total number of pixels in the image group, then occupies the probability of the effective edge points of the UAV inspection image and can be expressed by the following formula:

[0089]

[0090] In formula (8), ω1 and ω2 respectively represent the area and number of effective edge information in the UAV inspection image, and μ1 and μ2 represent the proportion of effective edge information in the UAV inspection image. Use the following formula to calculate the variance between the effective edge points and non-edge points in the UAV inspection image:

[0091]

[0092] In formula (10), μ represents the edge mean of the UAV inspection image. As can be seen from the above formula, the larger the variance value between the two inspection images, the greater the threshold difference between the two [9]. Define the threshold at this time as κ, obtain the threshold information of the UAV inspection image, and realize the effective segmentation of the non-edge points and weak edge points of the UAV inspection image:

[0093]

[0094] In formula (11), R(P) represents the edge point threshold function information of the UAV inspection image, and g h represents the effective segmentation information of the edge points of the inspection image.

[0095] In order to detect the edge points of the UAV inspection image, determine the threshold information of the edge points of the UAV inspection image through the above steps, and realize the effective detection of the edge of the substation UAV inspection image.

[0096] When identifying the edge information of the UAV inspection image, introduce the coefficient of variation of the UAV dynamic inspection image

[10] :

[0097] C ij =Δ ij / I ij(12)

[0098] In formula (12), Δ ij and I ij represent the standard variance and threshold of the edge ij of the UAV inspection image respectively. Formula (12) can analyze the connection strength β ij of the UAV inspection image to obtain C ij = β ij , and obtain the discrete matrix of the UAV inspection image:

[0099]

[0100] In formula (13), n represents the number of iterations of the inspection image, and F ij [n] and U ij [n] represent the information value and dynamic value of the edge of the UAV inspection image respectively [11 - 12]. θ ij [n] represents the weight of the edge of the UAV inspection image. α and β represent the weight membership degree and connection coefficient of the edge points of the UAV inspection image. θ0 represents the gray value of the edge of the UAV inspection image. When β ij ≠ 0, the edge information of the UAV inspection image is obtained by using the convolutional network. Define the edge matrix T and perform iterative processing on the UAV inspection image

[13] , and there is the following expression:

[0101]

[0102] Formula (14) reflects the spatial information and time information of the UAV inspection image. Define the discrete mass points in the inspection image plane as:

[0103]

[0104] In formula (15), m r represents the discrete mass point information at the inspection image coordinate point (x r , m r ). (x c , y c ) represents the center of the discrete mass points on the edge of the inspection image, which can reflect the shape characteristics of the edge of the UAV inspection image

[14] .

[0105] In order to effectively identify the edge information of the UAV inspection image, it is necessary to perform normalization processing on it

[15] . The process is as follows:

[0106]

[0107] In formula (16), M and N represent the dynamic inspection images collected by the UAV. Formula (16) can perform noise interference processing on the dynamic inspection images. Combining the gray values of the UAV inspection images obtained above, it can achieve adaptive recognition of the edge information of the substation UAV inspection images. The expression is as follows:

[0108] δ = i c0 + j c0 +(T ij [n]·θ0) (17)

[0109] In an example, to verify the actual application performance of the above-designed edge recognition method for substation UAV inspection images, the following experiment was carried out:

[0110] During the experiment, in order to exclude the influence of noise factors on the quality of substation UAV inspection images, the experimental parameters shown in Table 1 were set.

[0111] Table 1 Statistical table of experimental parameter settings

[0112]

[0113] During the experiment, the discrete coefficient was first introduced to measure the recognition effect of the edges of substation UAV inspection images. The larger the discrete coefficient value, the better the recognition effect of the inspection image edges, and vice versa.

[0114] Then, the quality coefficient was used to measure the recognition performance of the edges of substation UAV inspection images. The quality coefficient was defined as:

[0115]

[0116] In the formula, n0 represents the number of points recognized in the ideal state, n d represents the number of points recognized in the normal state, represents the proportionality coefficient, d γ represents the distance between the recognized edge point γ and the ideal edge point.

[0117] Furthermore, an image edge recognition method based on Franklin moments and an image edge recognition method based on neutrosophic theory were introduced for comparison.

[0118] The test results of the three methods in terms of the discrete coefficient of image edge recognition are as Figure 2 shown. Analysis Figure 2As can be seen from the results, when the image edge recognition method based on Franklin moments is adopted, the discrete coefficient in identifying the edge of the inspection image is between 2 and 8. As the number of recognition times increases, the discrete coefficient also becomes larger and larger, resulting in an unsatisfactory recognition effect. When the image edge recognition method based on neutrosophic theory is adopted, the discrete coefficient in identifying the edge of the inspection image is larger than that of the image edge recognition method based on Franklin moments. However, when the recognition coefficient is less than 5 times, the discrete coefficient of image edge recognition is less than 10, which cannot meet the recognition requirements of the inspection image edge. When the method of this paper is adopted, the discrete coefficient in identifying the edge of the inspection image is between 15 and 20, which is significantly higher than that of the image edge recognition method based on Franklin moments and the image edge recognition method based on neutrosophic theory. Thus, it can be shown that the designed substation UAV inspection image edge recognition method in this paper has a better effect in identifying the edge of the inspection image.

[0119] The test results of the three methods in terms of the high-quality coefficient of image edge recognition are as Figure 3 shown. From Figure 3 the results, it can be seen that the change trends of the high-quality coefficients of the image edge recognition method based on Franklin moments and the image edge recognition method based on neutrosophic theory in identifying the edge of the inspection image are basically the same. However, the image edge recognition method based on neutrosophic theory is higher than that based on Franklin moments. When the method of this paper is adopted, as the number of recognition times increases, the high-quality coefficient of identifying the edge of the inspection image is always higher than 17. Thus, it can be shown that the designed substation UAV inspection image edge recognition method in this paper has higher performance in inspection recognition.

[0120] In summary, the substation UAV inspection image edge recognition method provided in this application. Through testing, it is found that this method not only has a better recognition effect in identifying the edge of the substation UAV inspection image, but also can improve the performance of image edge recognition.

[0121] The embodiment of this application also provides a UAV inspection image edge recognition system, including: a processor and a memory; the memory is used to store a computer program, and the processor runs the computer program to enable the mobile terminal to execute the above method steps. This UAV inspection image edge recognition system can implement each embodiment of the above UAV inspection image edge recognition method and can achieve the same beneficial effects. Here, it will not be elaborated.

[0122] The embodiment of this application also provides a computer-readable storage medium, on which a computer program is stored. When this program is executed by a processor, it implements the above method steps. This readable storage medium can implement each embodiment of the above battery replacement method and can achieve the same beneficial effects. Here, it will not be elaborated.

[0123] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field according to the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art shall fall within the protection scope determined by the claims.

Claims

1. A method for edge recognition of substation UAV inspection images, characterized in that, Including: S1: Determine the gray values of the inspection images collected by the drone, and reconstruct the gray values to obtain the pixel gray values of the inspection images; S2: Determine the edge point set of the inspection image according to the pixel gray values of the inspection image; S3: Determine the edge point threshold information of the inspection image according to the edge point set; S4: Determine the edge information of the inspection image based on the edge point threshold information and a preset convolutional network model; The S1 includes: S11: Set the local similarity function of the dynamic patrol image as H ij , which is expressed by the following formula: where, H s-ij represents the pixel value similarity between the inspection images i and j, and H g-ij represents the gray value similarity between the inspection images i and j in space; The gray value similarity between the drone inspection images i and j is expressed as: where x i represents the pixel gray value of the inspection image i in the core area, and x j represents the pixel gray value of the inspection image j in the core area, λ g represents the influencing factor of the similarity of the inspection image pixel values, and δ g-i represents the density function, which is defined as follows: Where N i represents the pixel points of the dynamic inspection image i collected by the UAV in the adjacent area, and N R represents the number of pixel points of the dynamic inspection image i collected by the UAV in the adjacent area, and δ g-i represents the density function of the dynamic inspection image collected by the UAV in the core area; Reconstructing the gray values of the inspection images is expressed as: where η i represents the gray value of the i-th image pixel in the reconstructed inspection image η.

2. The method for edge recognition of substation UAV inspection images according to claim 1, characterized in that, The S2 includes: S21: Describe each pixel within the specification of M×N in the inspection image group as follows: p(x,y) = f R (x,y)i + f G (x,y)j + f B (x,y)k; where (x, y) represents the image position of the inspection image within M×N, and f R (x, y), f G (x, y), f B (x, y) represent the specific pixel values of R, G, and B in the inspection image respectively, and (i, j, k) represent three virtual units of the spatial features of the inspection image; S22: Normalize each pixel p(x,y) in the inspection image group, and calculate the Grassmann product of each dynamic inspection image r and the image gray value r0: P r P r0 = -q1·q2 + q1×q2; Wherein, q1 and q2 represent the pixel values of the dynamic inspection image, P r P r0 includes a pixel value S[P r P r0 and a dynamic image V[P r P r0 ; S23: Determine the similarity discrimination function between r and r0: In the formula, t represents the gray value of the inspection image. Based on the similarity discrimination function, calculate the value of the dynamic inspection image pixel r0 in the USAN region as n(r0), compare n(r0) with the vector g, and determine the edge point set E of the drone inspection image.

3. The method for edge recognition of substation UAV inspection images according to claim 2, characterized in that, The S3 includes: S31: Calculate the edge point threshold of the inspection image, and define the threshold relationship between the edge points and non-edge points in the inspection image. represents the effective area of the inspection image. used to describe the number of pixels with a USAN area of in the inspection image group. represents the total number of pixels in the image group, then the probability of the effective edge points in the UAV inspection image is expressed as follows: In the formula, ω1 and ω2 respectively represent the area and number of valid edge information in the drone inspection image, and μ1 and μ2 represent the proportion of valid edge information in the drone inspection image; S32: Calculate the variance of valid edge points and non-edge points in the UAV inspection image In the formula, μ represents the edge mean of the drone inspection image. It can be seen from the above formula that the larger the variance value of the inspection image, the greater the threshold difference between the two. Define the threshold at this time as κ, obtain the threshold information of the drone inspection image, and realize the effective segmentation of non-edge points and weak edge points of the drone inspection image: Wherein, R(P) represents the edge point threshold function information of the UAV inspection image, and g l represents the lower limit of the effective segmentation information of the edge points of the inspection image, and g h represents the upper limit of the effective segmentation information of the edge points of the inspection image.

4. The edge recognition method for substation UAV inspection images according to claim 1, wherein, The S4 includes: S41: Set the coefficient of variation of the inspection image as follows: C ij = Δ ij / I ij ; where, Δ ij and I ij represent the standard variance and threshold of the edge ij of the inspection image respectively, where C ij = β ij , and β ij represents the connection strength of the inspection image. Then, the discrete matrix of the inspection image is as follows: Wherein, n represents the iteration number of the inspection image, F ij [n], U ij [n] respectively represent the information value and the dynamic value of the edge of the inspection image, θ ij [n] represents the weight value of the edge of the inspection image, α and β represent the weight membership degree and the connection coefficient of the edge points of the inspection image, θ0 represents the gray value of the edge of the inspection image. When β ij ≠0, the edge information of the inspection image is obtained by using a preset convolutional network model, an edge matrix T is defined, and the inspection image of the unmanned aerial vehicle is iteratively processed; S42: Define the discrete mass points in the inspection image plane; and normalize the edge information of the inspection image; S43: Combine the obtained gray values of the drone inspection image to determine the model for adaptive recognition of the edge information of the substation drone inspection image, and the expression is as follows: δ = i c0 + j c0 +(T ij [n]·θ0); where δ represents the edge recognition result, and i c0 , j c0 respectively represent the results of normalizing the edge information of the inspection image, and T ij [n] represents the result of iterative processing.

5. A UAV inspection image edge recognition system, wherein, Including: A processor and a memory; The memory is used to store a computer program, and the processor runs the computer program to enable the mobile terminal to execute the method steps described in any one of claims 1 to 4.

6. A computer-readable storage medium, on which a computer program is stored, wherein, When the program is executed by the processor, it implements the method steps described in any one of claims 1 - 4.

Images