Telegraph pole inclination detection method based on artificial intelligence

Abstract

The invention relates to the field of artificial intelligence, in particular to a telegraph pole inclination detection method based on artificial intelligence. Comprising the following steps: based on an acquired telegraph pole image data sequence, extracting edge features by using a dynamic adaptive convolution kernel edge enhancement algorithm, forming an edge feature graph, and extracting an edge point set; fitting the edge point set by using a self-adaptive dynamic regression model, selecting an optimal regression order, calculating to obtain a slope of the self-adaptive dynamic regression model, and calculating to obtain an initial inspection inclination angle of the telegraph pole by using an arc tangent function; calculating a final inspection inclination angle of the telegraph pole through a nonlinear sensor data correction optimization algorithm based on the initial inspection inclination angle of the telegraph pole; and an inclination threshold value is set and compared with the final inspection inclination angle of the telegraph pole to judge whether the inclination of the telegraph pole is normal or not. The technical problem that a traditional convolution algorithm cannot adaptively adjust a processing mode according to image local features, so that the processing effect difference between an edge region and a smooth region is large is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an electric pole tilt detection method based on artificial intelligence. Background Art

[0002] As an important part of the power transmission system, utility poles undertake the key task of supporting and transmitting power lines. They are widely used in urban and rural power infrastructure to ensure a stable supply of electricity. However, over time, utility poles may tilt or collapse due to natural factors (such as wind, earthquakes, corrosion) or human factors (such as construction, traffic accidents), which will not only cause power outages and affect daily life and industrial production, but may also cause serious safety accidents. Therefore, timely and accurate detection of the tilt of utility poles is of great significance for maintaining the stable operation of the power transmission system and ensuring public safety.

[0003] Traditional pole tilt detection mainly relies on manual inspection, which requires a lot of manpower, especially in areas with wide coverage of power grids. The inspection tasks are heavy and the cycle is long. In addition, inspectors need to climb poles for inspection, which poses certain safety hazards, especially when working in bad weather or high-altitude environments. Although instrument-based detection methods improve measurement accuracy, the equipment is expensive and the operation is complicated, the cost is difficult to control, and regular maintenance and calibration are required, which increases the operation and maintenance burden. With the rapid development of artificial intelligence (AI) technology, an AI-based pole tilt detection method has emerged, which has broad application prospects in pole tilt detection. Artificial intelligence technology has powerful data processing and pattern recognition capabilities, and can extract effective information from massive data to achieve automated and intelligent detection and analysis.

[0004] However, the existing methods for detecting the tilt of utility poles have the following technical problems: the traditional convolution algorithm cannot adaptively adjust the processing method according to the local features of the image, resulting in large differences in the processing effects between the edge area and the smooth area; the tilt angle of the utility pole not only depends on the image data, but also requires comprehensive sensor data for correction. It is difficult to ensure the accuracy of detection using a single image or sensor. Summary of the invention

[0005] The present invention provides a method for detecting the inclination of electric poles based on artificial intelligence to solve the technical problem that the traditional convolution algorithm cannot adaptively adjust the processing method according to the local features of the image, resulting in a large difference in the processing effect between the edge area and the smooth area; the inclination angle of the electric pole not only depends on the image data, but also needs to be corrected by comprehensive sensor data, and it is difficult to ensure the accuracy of detection using a single image or sensor detection.

[0006] The present invention provides a method for detecting the inclination of electric poles based on artificial intelligence, which specifically includes the following technical solutions:

[0007] A method for detecting the inclination of a utility pole based on artificial intelligence comprises the following steps:

[0008] S1: Based on the acquired telegraph pole image data sequence, the edge features of the telegraph pole image are extracted using the dynamic adaptive convolution kernel edge enhancement algorithm, and an edge feature map is constructed. Based on the obtained edge feature map, the edge point set is extracted;

[0009] S2: Use the adaptive dynamic regression model to fit the edge point set, select the optimal regression order, calculate the slope of the adaptive dynamic regression model according to the optimal regression order, and use the inverse tangent function to calculate the initial inspection inclination angle of the electric pole according to the slope of the adaptive dynamic regression model;

[0010] S3: Based on the initial inspection tilt angle of the utility pole, the final inspection tilt angle of the utility pole is calculated by a nonlinear sensor data correction optimization algorithm; and a tilt threshold is set and compared with the final inspection tilt angle of the utility pole to determine whether the tilt of the utility pole is normal.

[0011] Preferably, the S1 specifically includes:

[0012] In the implementation process of the dynamic adaptive convolution kernel edge enhancement algorithm, the local gradient of the image at each position is calculated, and the intensity of image change is quantified by calculating the gradient amplitude of the local area of ​​the image; according to the size of the gradient amplitude, a spatial enhancement factor is assigned to each image position.

[0013] Preferably, the S1 specifically includes:

[0014] In the implementation process of the dynamic adaptive convolution kernel edge enhancement algorithm, the size of the convolution kernel is dynamically adjusted based on the gradient amplitude to extract the edge features. The specific formula is:

[0015]

[0016] Wherein, C(x, y, t) represents the edge feature extracted at the image position (x, y) and time point t, which is used to form the edge feature map; K(x, y, t) represents the convolution kernel size at the image position (x, y) and time point t; u and v represent the sliding range of the convolution kernel on the image; It represents the weighted summation of the pixel values ​​in the specified area of ​​the image; I(x+u, y+v, t) represents the pixel value at the image position (x+u, y+v) and time point t; A(x+u, y+v, t) represents the spatial enhancement factor at the image position (x+u, y+v) and time point t.

[0017] Preferably, the S2 specifically includes:

[0018] In the adaptive dynamic regression model, an optimization strategy based on error analysis is adopted. The fitting effect of the adaptive dynamic regression model is evaluated by calculating the error of the adaptive dynamic regression model on the edge point set, and the optimal regression order is selected by combining the regularization parameter and the optimization objective function.

[0019] Preferably, the S2 specifically includes:

[0020] The formula for calculating the optimal regression order is:

[0021]

[0022] Among them, n o represents the optimal regression order; represents the order n of the adaptive dynamic regression model that minimizes the objective function; Represents the error term, which is the adaptive dynamic regression model for each edge point (x i ,y i )’s fitting error; For every pair of edge points (x i ,y i ) to sum; y i Indicates the vertical coordinate value corresponding to the i-th edge point; c j represents the j-th order regression coefficient in the adaptive dynamic regression model; represents the jth power of the horizontal coordinate value corresponding to the i-th edge point; α represents the regularization parameter; n represents the order of the adaptive dynamic regression model.

[0023] Preferably, the S2 specifically includes:

[0024] The slope of the adaptive dynamic regression model is calculated according to the optimal regression order. The specific formula is:

[0025]

[0026] Where m represents the slope of the adaptive dynamic regression model; represents the sum operation; It represents the slope contribution of each order term in the adaptive dynamic regression model; Represents the horizontal coordinate value corresponding to the i-th edge point, -1 power;

[0027] The inclination angle of the electric pole is calculated using the inverse tangent function according to the slope of the adaptive dynamic regression model. The specific formula is:

[0028] θ p =arctan(m)

[0029] Among them, θ p represents the initial tilt angle of the electric pole; arctan represents the inverse tangent function.

[0030] Preferably, the S3 specifically includes:

[0031] In the implementation process of the nonlinear sensor data correction optimization algorithm, the influence of each sensor signal is calculated separately, different correction factors are applied according to the error characteristics of each sensor, and logarithmic transformation and square term transformation are introduced to calculate the final inspection inclination angle of the utility pole.

[0032] Preferably, the S3 specifically includes:

[0033] The calculation formula of the final inspection inclination angle of the electric pole is:

[0034]

[0035] Among them, θ q Indicates the final inspection inclination angle of the electric pole; θ p Indicates the initial inspection tilt angle of the pole; represents the weighted summation of N sensor signals; S k (t) represents the kth sensor signal; α k represents the correction factor of the kth sensor signal; β k represents the nonlinear correction factor that controls the kth sensor signal after logarithmic correction; γ k represents the quadratic correction factor of the kth sensor signal; ln(1+|S k (t)|) means taking the natural logarithm of the absolute value of the kth sensor signal plus one; (S k (t)) 2 represents the square term of the kth sensor signal.

[0036] Preferably, the S3 specifically includes:

[0037] When the final detection inclination angle of the electric pole is greater than the inclination threshold θ th When the pole is tilted abnormally, the pole is tilted abnormally; when the final tilt angle of the pole is less than or equal to the tilt threshold, the pole is tilted normally; the specific implementation is:

[0038]

[0039] Among them, Q represents the judgment result, that is, whether the inclination of the electric pole is abnormal or normal; θ th Indicates the tilt threshold.

[0040] The beneficial effects of the technical solution of the present invention are:

[0041] 1. The dynamic adaptive convolution kernel edge enhancement algorithm dynamically adjusts the size of the convolution kernel and combines the gradient information of the image. The area with larger gradient (i.e., edge area) uses a smaller convolution kernel to capture fine edge information, while the area with smaller gradient uses a larger convolution kernel to capture a wider range of features. The spatial enhancement factor is used to enhance the features of the edge area, thereby achieving accurate edge extraction.

[0042] 2. An adaptive dynamic regression model is used to fit the set of edge points extracted from the edge feature map. By selecting the optimal regression order, the problem of overfitting or underfitting can be avoided, thereby ensuring the robustness and accuracy of the adaptive dynamic regression model.

[0043] 3. By using the sensor signal and combining it with the initial inspection tilt angle of the utility pole, a nonlinear sensor data correction optimization algorithm is introduced. The calculation of the tilt angle can be further corrected according to the dynamic response signal provided by the sensor, ensuring that the final inspection tilt angle of the utility pole has extremely high accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of a method for detecting electric pole tilt based on artificial intelligence described in the present invention. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0047] The following is a detailed description of a specific scheme of a method for detecting the tilt of a utility pole based on artificial intelligence provided by the present invention in conjunction with the accompanying drawings.

[0048] See attached Figure 1 , which shows a flow chart of a method for detecting the inclination of a utility pole based on artificial intelligence provided by an embodiment of the present invention, the method comprising the following steps:

[0049] S1. Based on the acquired telegraph pole image data sequence, a dynamic adaptive convolution kernel edge enhancement algorithm is used to extract edge features of the telegraph pole image and form an edge feature map. Based on the obtained edge feature map, a set of edge points is extracted.

[0050] The camera is used to capture images of electric poles in real time, and a continuous electric pole image data sequence I (x, y, t) is obtained, where (x, y) is the spatial coordinate of the electric pole image, and t is the time index, indicating electric pole images collected at different time points;

[0051] Multimodal sensors, such as accelerometers, gyroscopes, and stress sensors, are installed on utility poles and their surroundings to monitor the dynamic response of utility poles in real time and obtain the output signal S of each sensor. k (t);

[0052] Based on the acquired continuous telegraph pole image data sequence, the dynamic adaptive convolution kernel edge enhancement algorithm is used to improve the extraction accuracy of the edge features of the telegraph poles in the image by dynamically adjusting the convolution kernel size and performing spatial enhancement according to the image gradient, and the edge features of the telegraph pole images are extracted;

[0053] The dynamic adaptive convolution kernel edge enhancement algorithm not only considers the pixel value of the image when performing convolution calculation, but also combines the dynamically adjusted convolution kernel size and the spatial enhancement factor. For each image position, the dynamically adjusted convolution kernel is used to perform weighted summation on the surrounding area. Each element of the convolution kernel is multiplied by the pixel value of the corresponding position in the image, and each product is also weighted according to the spatial enhancement factor. All weighted products are summed to obtain the convolution results at different positions, ensuring accurate edge extraction effect.

[0054] When processing an image, the edge area will show a large brightness change, while the smooth area will show a small brightness change. In order to effectively extract edge features, the dynamic adaptive convolution kernel edge enhancement algorithm will calculate the local gradient of the image at each position. The gradient represents the rate of change of the image at different positions. The intensity of the image change is quantified by calculating the gradient amplitude of the local area of ​​the image. According to the size of the gradient amplitude, a spatial enhancement factor is assigned to each image position. The size of the spatial enhancement factor is in a negative exponential relationship with the gradient amplitude, so that the area with a very large gradient (i.e., the edge area) will obtain a very high spatial enhancement factor, while the area with a very small gradient (i.e., the smooth area) will obtain a very small spatial enhancement factor, which can effectively distinguish the edge area from the smooth area in the image;

[0055] Traditional convolutional neural networks use fixed-size convolution kernels for image processing, which leads to differences in processing effects between edge areas and smooth areas. In order to adapt to the characteristics of different areas in the image, the dynamic adaptive convolution kernel edge enhancement algorithm dynamically adjusts the size of the convolution kernel by calculating the local gradient amplitude of the image. Specifically, the size of the convolution kernel is associated with the local gradient amplitude of the image. A small convolution kernel is used in areas with large gradients to accurately capture edge details, while in areas with small gradients, the convolution kernel becomes larger to capture global information, so that the convolution operation can be performed adaptively according to the local characteristics of the image, thereby improving the flexibility and effect of image processing;

[0056] The edge feature extraction formula is:

[0057]

[0058] Among them, C(x, y, t) represents the edge feature extracted at the image position (x, y) and time point t, which is used to form the edge feature map; K(x, y, t) represents the size of the convolution kernel at the image position (x, y) and time point t. The size of the convolution kernel is adjusted according to the size of the gradient amplitude. The specific calculation formula is:

[0059]

[0060] Among them, K0 represents the basic convolution kernel size, which can be set according to the specific implementation scenario and is not limited here; B represents the hyperparameter for adjusting the change of the convolution kernel size, which is used to control the influence of the gradient change on the convolution kernel size adjustment; represents the gradient at the image position (x, y) and time point t, which reflects the degree of local change in the image and is used to determine the edge position in the image. The calculation method of the gradient is a technical means well known to those skilled in the art and will not be described in detail here; represents the modulus of the gradient (gradient amplitude), which is used to measure the degree of change of the image, that is, the strength of the edge of the image. The calculation method of the modulus of the gradient is a technical means well known to those skilled in the art and will not be described in detail here; u and v represent the sliding range of the convolution kernel on the image; It represents the weighted summation of the pixel values ​​in the specified area of ​​the image (determined by the convolution kernel size K(x, y, t)). The convolution kernel covers every pixel area of ​​the image and performs weighted processing on each pixel value. I(x+u, y+v, t) represents the pixel value at the image position (x+u, y+v) and time point t. A(x+u, y+v, t) represents the spatial enhancement factor at the image position (x+u, y+v) and time point t. The calculation formula is:

[0061]

[0062] Among them, A(x, y, t) represents the spatial enhancement factor at the image position (x, y) and time point t, which is used to dynamically adjust the response of the convolution kernel to enhance the extraction effect of edge features. By adjusting the spatial enhancement factor, the features of the edge area are highlighted, the noise of the non-edge area is suppressed, and the edge information is ensured to be clear; λ represents the sensitivity adjustment parameter of the spatial enhancement factor, which controls the degree of edge enhancement and is used to adjust the sensitivity of the gradient amplitude to the spatial enhancement factor; exp represents an exponential function, which uses the negative exponential decay of the gradient amplitude to calculate the spatial enhancement factor, which is used to strengthen the edge features, giving a stronger weight to the area with a larger gradient (i.e., the edge of the telephone pole), and a smaller weight to the area with a smaller gradient (i.e., the background area);

[0063] According to the obtained edge feature map, the edge point set P is extracted from it, that is,

[0064] P={(x i ,y i )|C(x i ,y i , t)≥a}

[0065] Among them, x i and i represent the pixel coordinates of the ith edge point in the horizontal and vertical directions respectively; a represents the edge threshold, which can be set according to the specific implementation scenario and is not limited here; C(x i ,y i , t)≥a means that the pixel points greater than or equal to the edge threshold are edge points; C(x i ,y i , t) indicates that at the edge point (x i ,y i ) and the edge features at time point t:

[0066] S2. Fit the edge point set using an adaptive dynamic regression model, select an optimal regression order, calculate the slope of the adaptive dynamic regression model according to the optimal regression order, and calculate the initial inspection inclination angle of the electric pole using an inverse tangent function according to the slope of the adaptive dynamic regression model;

[0067] Use the adaptive dynamic regression model to fit the edge point set, and the adaptive dynamic regression model sets the target variable y i (i.e. the vertical coordinate of the pole) can be obtained by inputting the variable x i (i.e. the horizontal coordinates of the electric pole) and the linear combination of power terms. In the adaptive dynamic regression model, each order will have a corresponding regression coefficient, which reflects the impact of each order term on the target variable.

[0068] In order to find the optimal regression order, an optimization strategy based on error analysis is adopted. Specifically, the error of the adaptive dynamic regression model on the edge point set under each possible order is calculated to evaluate the fitting effect of the adaptive dynamic regression model. A regularization parameter is used to penalize the complexity of the adaptive dynamic regression model to avoid the adaptive dynamic regression model being too complex. An optimal regression order is selected by optimizing the objective function.

[0069] The formula for calculating the optimal regression order is:

[0070]

[0071] Among them, n o Represents the optimal regression order, which can optimize the fitting accuracy and complexity of the adaptive dynamic regression model and avoid overfitting or underfitting; represents the order n of the adaptive dynamic regression model that minimizes the objective function; Represents the error term, which is the adaptive dynamic regression model for each edge point (x i ,y i )’s fitting error; For every pair of edge points (x i ,y i ) to sum; y i Indicates the vertical coordinate value corresponding to the i-th edge point; c j It represents the regression coefficient of the jth order in the adaptive dynamic regression model, reflecting the regression weight of each order; represents the power of the horizontal coordinate value corresponding to the ith edge point; α represents the regularization parameter, which is used to balance the complexity and fitting accuracy of the adaptive dynamic regression model to prevent overfitting; n represents the order of the adaptive dynamic regression model, that is, the maximum degree of the regression polynomial;

[0072] The selection of the optimal regression order not only improves the fitting accuracy, but also ensures that the calculated slope and tilt angle are robust and accurate in practical applications;

[0073] The slope of the adaptive dynamic regression model is calculated according to the optimal regression order. The specific formula is:

[0074]

[0075] Among them, m represents the slope of the adaptive dynamic regression model, which reflects the inclination angle of the utility pole; represents the sum operation; It represents the slope contribution of each order term in the adaptive dynamic regression model; Represents the j-1th power of the horizontal coordinate value corresponding to the i-th edge point;

[0076] The inclination angle of the electric pole is calculated using the inverse tangent function according to the slope of the adaptive dynamic regression model. The specific formula is:

[0077] θ p =arctan(m)

[0078] Among them, θ p represents the initial inspection tilt angle of the electric pole, that is, the tilt angle of the electric pole relative to the vertical direction; arctan represents the inverse tangent function, which is used to convert the slope of the adaptive dynamic regression model into the tilt angle of the electric pole;

[0079] S3. Based on the initial inspection tilt angle of the utility pole, the final inspection tilt angle of the utility pole is calculated by a nonlinear sensor data correction optimization algorithm; and a tilt threshold is set and compared with the final inspection tilt angle of the utility pole to determine whether the tilt of the utility pole is normal;

[0080] Based on the initial inspection tilt angle of the utility pole, the nonlinear sensor data correction optimization algorithm uses the sensor signal to undergo fine nonlinear correction, and through weighted and nonlinear adjustment, the final inspection tilt angle of the utility pole is calculated;

[0081] In the nonlinear correction process, the influence of each sensor signal on the final inspection inclination angle of the utility pole is taken into account, the influence of each sensor signal is calculated separately, and different correction factors are applied according to the error characteristics of each sensor. The correction factor can be set according to the specific implementation scenario and is not limited here;

[0082] The calculation of the correction factor depends on the actual performance of the sensor signal. When the sensor signal value is extremely small or close to zero, the logarithmic transformation can smooth the sensor signal, reduce the impact of the extremely small sensor signal on the correction result, and prevent over-correction or error in the low signal range;

[0083] When the sensor signal is extremely large, the square transformation helps capture the high-order nonlinear characteristics of the sensor signal, ensuring that the sensor signal can be corrected when a large deviation occurs;

[0084] The calculation formula of the final inspection inclination angle of the electric pole is:

[0085]

[0086] Among them, θ q Indicates the final inspection inclination angle of the electric pole; θ p Indicates the initial inspection tilt angle of the pole; represents the weighted summation of N sensor signals; S k (t) represents the kth sensor signal; α k represents the correction factor of the kth sensor signal; βk represents the nonlinear correction factor that controls the kth sensor signal after logarithmic correction, which is used to correct the error of the sensor signal in the low value range; γ k represents the quadratic correction factor of the kth sensor signal, which controls the square correction effect of the sensor signal; ln(1+|S k (t)|) represents the natural logarithm of the absolute value of the kth sensor signal plus one, which is used to reduce the extreme changes of the sensor signal. In particular, when the absolute value of the sensor signal is extremely small or close to zero, the logarithmic transformation can smooth the sensor signal and avoid overcorrection in the extremely small sensor signal range; (S k (t)) 2 Represents the square term of the kth sensor signal, eliminating the nonlinear deviation generated by the sensor;

[0087] Set the tilt threshold θ th , and compare it with the final inspection inclination angle of the pole to determine whether the inclination of the pole is normal or abnormal; when the final inspection inclination angle of the pole is greater than the inclination threshold θ th When , it is abnormal, otherwise, it is normal; the tilt threshold can be set according to the specific implementation scenario, and is not limited here;

[0088]

[0089] Among them, Q represents the judgment result, that is, whether the inclination of the electric pole is abnormal or normal; θ th represents the tilt threshold;

[0090] In summary, an artificial intelligence-based method for detecting the tilt of electric poles is completed.

[0091] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0092] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0093] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for detecting the inclination of electric poles based on artificial intelligence, characterized in that: The following steps are involved: S1: Based on the acquired telegraph pole image data sequence, the edge features of the telegraph pole image are extracted using the dynamic adaptive convolution kernel edge enhancement algorithm, and an edge feature map is constructed. Based on the obtained edge feature map, the edge point set is extracted; S2: Use the adaptive dynamic regression model to fit the edge point set, select the optimal regression order, calculate the slope of the adaptive dynamic regression model according to the optimal regression order, and use the inverse tangent function to calculate the initial inspection inclination angle of the electric pole according to the slope of the adaptive dynamic regression model; S3: Based on the initial inspection tilt angle of the utility pole, the final inspection tilt angle of the utility pole is calculated by a nonlinear sensor data correction optimization algorithm; and a tilt threshold is set and compared with the final inspection tilt angle of the utility pole to determine whether the tilt of the utility pole is normal.

2. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 1, characterized in that: The S1 specifically includes: In the implementation process of the dynamic adaptive convolution kernel edge enhancement algorithm, the local gradient of the image at each position is calculated, and the intensity of image change is quantified by calculating the gradient amplitude of the local area of ​​the image; according to the size of the gradient amplitude, a spatial enhancement factor is assigned to each image position.

3. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 2, characterized in that: The S1 specifically includes: In the implementation process of the dynamic adaptive convolution kernel edge enhancement algorithm, the size of the convolution kernel is dynamically adjusted based on the gradient amplitude to extract the edge features. The specific formula is: Wherein, C(x, y, t) represents the edge feature extracted at the image position (x, y) and time point t, which is used to form the edge feature map; K(x, y, t) represents the convolution kernel size at the image position (x, y) and time point t; u and v represent the sliding range of the convolution kernel on the image; It represents the weighted summation of the pixel values ​​in the specified area of ​​the image; I(x+u, y+v, t) represents the pixel value at the image position (x+u, y+v) and time point t; A(x+u, y+v, t) represents the spatial enhancement factor at the image position (x+u, y+v) and time point t.

4. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 1, characterized in that: The S2 specifically includes: In the adaptive dynamic regression model, an optimization strategy based on error analysis is adopted. The fitting effect of the adaptive dynamic regression model is evaluated by calculating the error of the adaptive dynamic regression model on the edge point set, and the optimal regression order is selected by combining the regularization parameter and the optimization objective function.

5. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 4, characterized in that: The S2 specifically includes: The formula for calculating the optimal regression order is: Among them, n o represents the optimal regression order; represents the order n of the adaptive dynamic regression model that minimizes the objective function; Represents the error term, which is the adaptive dynamic regression model for each edge point (x i ,y i )’s fitting error; For every pair of edge points (x i ,y i ) to sum; y i Indicates the vertical coordinate value corresponding to the i-th edge point; c j represents the j-th order regression coefficient in the adaptive dynamic regression model; represents the jth power of the horizontal coordinate value corresponding to the i-th edge point; α represents the regularization parameter; n represents the order of the adaptive dynamic regression model.

6. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 1, characterized in that: The S2 specifically includes: The slope of the adaptive dynamic regression model is calculated according to the optimal regression order. The specific formula is: Where m represents the slope of the adaptive dynamic regression model; represents the sum operation; It represents the slope contribution of each order term in the adaptive dynamic regression model; Represents the j-1th power of the horizontal coordinate value corresponding to the i-th edge point; The inclination angle of the electric pole is calculated using the inverse tangent function according to the slope of the adaptive dynamic regression model. The specific formula is: θ p =arctan(m) Among them, θ p represents the initial tilt angle of the electric pole; arctan represents the inverse tangent function.

7. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 1, characterized in that: The S3 specifically includes: In the implementation process of the nonlinear sensor data correction optimization algorithm, the influence of each sensor signal is calculated separately, different correction factors are applied according to the error characteristics of each sensor, and logarithmic transformation and square term transformation are introduced to calculate the final inspection inclination angle of the utility pole.

8. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 7, characterized in that: The S3 specifically includes: The calculation formula of the final inspection inclination angle of the electric pole is: Among them, θ q Indicates the final inspection inclination angle of the electric pole; θ p Indicates the initial inspection tilt angle of the pole; represents the weighted summation of N sensor signals; S k (t) represents the kth sensor signal; α k represents the correction factor of the kth sensor signal; β k represents the nonlinear correction factor that controls the kth sensor signal after logarithmic correction; γ k represents the quadratic correction factor of the kth sensor signal; ln(1+|S k (t)|) means taking the natural logarithm of the absolute value of the kth sensor signal plus one; (S k (t)) 2 represents the square term of the kth sensor signal.

9. The method for detecting the inclination of electric poles based on artificial intelligence according to claim 1, characterized in that: The S3 specifically includes: When the final detection inclination angle of the electric pole is greater than the inclination threshold θ th When the pole is tilted abnormally, the pole is tilted abnormally; when the final tilt angle of the pole is less than or equal to the tilt threshold, the pole is tilted normally; the specific implementation is: Among them, Q represents the judgment result, that is, whether the inclination of the electric pole is abnormal or normal; θ th Indicates the tilt threshold.

Images