An ice-coating detection method based on image completion of three-phase transmission lines

Abstract

The application discloses an ice-coating detection method based on three-phase power transmission line image completion, and relates to the field of power transmission line state monitoring.The method comprises the following steps: acquiring a power transmission line image, and identifying three-phase conductors between two suspension points; selecting a phase with the most complete identification area in the three phases, and adopting a catenary model to fit a reference track; generating initial tracks of the other two phases by translating and copying the reference track; combining the identified partial areas of the other two phases in the image to establish anchor point constraints, and performing non-rigid deformation correction on the initial tracks so that the initial tracks coincide with the real conductors in the identified areas, thereby obtaining complete tracks of the three-phase conductors; and comparing the differences between the phases after obtaining the complete tracks, and preliminarily determining that there is ice-coating if the differences exceed a threshold value.The application solves the problem that the image is incomplete due to the fact that the three-phase conductors are shielded by the environment, and can quickly screen the ice-coating condition by using the image completion condition.

Description

Technical Field

[0001] This invention relates to the field of power transmission line condition monitoring, and more specifically, to an icing detection method based on three-phase transmission line image completion. Background Technology

[0002] When a transmission line spans a long span, the conductors, under the combined action of their own weight and tension, take on an approximate catenary shape. The three-phase conductors are usually suspended within the same tower span, with the suspension points at both ends at similar heights and symmetrical structures. Therefore, under normal operating conditions without icing, the spatial trajectories of the three-phase conductors appear to be approximately parallel and identical in shape from the same perspective.

[0003] Icing alters the equivalent load and sag of conductors, and the amount of icing in each phase is often not identical, leading to observable differences in the three-phase trajectories within the same span. Existing image-based icing identification technologies offer advantages such as low deployment cost and strong backtracking capabilities, but various interferences exist in engineering sites: conductors are partially obscured by ground features or background; backlighting, fog, and precipitation cause incomplete identification; the shooting angle introduces perspective distortion and scale differences; and the visible segments of the three-phase conductors within the same frame are of unequal length. These factors make it difficult to obtain reliable conclusions by directly comparing the identified segments of the three phases.

[0004] Due to environmental interference at the site, the following issues need to be addressed: how to reconstruct the complete trajectory of each phase within the span when only a portion of the trajectory is visible; how to make a preliminary icing determination based on the completed image; and how to further distinguish between the severity of icing and locate possible local icing areas. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an icing detection method based on three-phase transmission line image completion, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] An icing detection method based on three-phase transmission line image completion, including

[0008] Acquire images of the transmission line and identify the three-phase conductors connected between the suspension points at both ends;

[0009] In the three-phase conductors, the one with the most complete identification area is selected, and the baseline trajectory is obtained by fitting the catenary model.

[0010] The baseline trajectory is translated and copied to generate the initial trajectories of the other two phase conductors; anchor point constraints are established in combination with the regions of the other two phase conductors that have been identified in the image, and the initial trajectory is non-rigidly deformed and corrected so that the corrected trajectory coincides with the real conductor in the identified region, thereby obtaining the complete trajectory of the three-phase conductor.

[0011] After obtaining the complete trajectory of the three-phase conductors, the differences between each phase conductor are compared. When the difference between the three phase conductors exceeds a preset threshold, it is preliminarily determined that the three-phase conductors are covered with ice.

[0012] Furthermore, the non-rigid deformation correction includes:

[0013] Using the initial trajectory and the anchor point pairs of the corresponding identified areas of the guide wire as input, a set of deformation control points is established;

[0014] Based on the set of deformation control points, a deformation function is constructed using thin plate spline transformation or B-spline free deformation method;

[0015] The initial trajectory is mapped as a whole using the deformation function, so that it coincides with the identified trajectory at the anchor point.

[0016] Optionally, the process for detecting differences between the three-phase conductors includes:

[0017] Discretely sample the complete trajectory of the three-phase conductors, and define the set of sampling points as follows: Calculate the height difference between any two phase conductors at the sampling point, and obtain the overall average difference using the following formula:

[0018] ;

[0019] Indicates the first i Phase conductor and the first j Overall average difference in phase conductors; N Indicates the total number of sampling points. k For sampling point index, Indicates the first k The horizontal position of each sampling point Indicates the first i Phase conductor at sampling point The height, Indicates the first j Phase conductor at sampling point The height; among which i, j Choose 1, 2, or 3;

[0020] When the overall average difference between any two phase conductors When the threshold is exceeded, it is preliminarily determined that the three-phase conductors are covered with ice.

[0021] Preferably, the process for detecting differences between the three-phase conductors includes:

[0022] For each phase conductor, the line connecting the suspension points at both ends of that phase conductor is used as the top horizontal reference line, and the trajectory of that phase conductor is calculated accordingly. Area index enclosed by the corresponding top horizontal reference line :

[0023] ;

[0024] in, The horizontal length of the top horizontal reference line. The starting point of the horizontal coordinate is located at one of the hanging points. Indicates the first i Phase conductors on the horizontal coordinate x The height, Indicates the first i Height of suspension points at both ends of the phase conductor;

[0025] When the area index difference between any two phases When the threshold is exceeded, it is preliminarily determined that the three-phase conductors are icing. i, j Choose 1, 2, or 3.

[0026] Furthermore, the method also includes comparing the area indices of the three-phase conductors and sorting the degree of icing of the three-phase conductors from most severe to least severe according to the size of the area index, wherein a larger area index corresponds to a more severe degree of icing.

[0027] Furthermore, the method also includes placing the trajectories of any two phase conductors in the same coordinate system for comparison, and performing segmented calculations on the trajectory differences. When it is detected that a certain phase conductor is continuously lower than another phase conductor in a local area, it is preliminarily determined that there is icing in that local area.

[0028] Specifically, the process of fitting using a catenary includes:

[0029] Extract the coordinates of multiple discrete identification points of the conductor and establish the catenary function:

[0030] ;

[0031] Among them, parameters , , These represent the catenary tension factor, horizontal displacement, and vertical offset, respectively. The vertical axis is the coordinate of the axis. The horizontal axis;

[0032] The parameters are determined by fitting the error between discrete identification points and the catenary function using the least squares method. , , The optimal solution is obtained, thus yielding the complete catenary trajectory of the reference traverse.

[0033] Specifically, the deformation function is a two-dimensional coordinate mapping function constructed from the anchor point positions. ,satisfy:

[0034] ;

[0035] in Indicates the first position on the initial trajectory Anchor point coordinates, This indicates the coordinates of the corresponding identified guide wire anchor points. This represents the total number of anchor points.

[0036] Specifically, selecting the phase conductor with the most complete identification area among the three phase conductors includes: comparing the identification results of the three phase conductors, calculating the length of each phase conductor identified in the horizontal direction, and selecting the conductor with the longest length as the phase conductor with the most complete identification.

[0037] Specifically, the process of identifying the most complete phase conductor is determined by manually judging the completeness of the conductor in the image, or by calculating the length of the identification section of each phase conductor using an image recognition algorithm, and then selecting the phase conductor with the most complete identification area as the reference conductor.

[0038] The advantages of this invention over existing technologies lie in its core process. First, the most complete phase among the three phases is selected. A baseline trajectory is obtained by fitting a catenary model. This trajectory is then translated and copied to generate initial trajectories for the other two phases. Subsequently, anchor point constraints are set based on the identified segments of the other two phases in the image, and non-rigid deformation correction is applied to the initial trajectory. Finally, a complete set of trajectories for the three-phase conductors within the same span is obtained. The differences are then compared to determine if they exceed a preset threshold, thus completing the initial icing determination. This brings three advantages: First, the "baseline-transfer-correction" completion mechanism significantly mitigates the impact of occlusion and incomplete identification on the stability of the judgment. Second, it fully utilizes the structural priors of the three phases being of the same span, the same tower, and similar in shape, enabling the restoration of a highly consistent reference shape even in low visibility or with local defects, reducing false alarms and missed alarms. Third, the overall calculation chain is simple and thresholdable, facilitating rapid on-site screening and enabling rapid preliminary identification of icing.

[0039] Building upon this foundation, thin-plate splines or B-splines are further employed for free deformation to perform a smooth mapping of the initial trajectory under anchor point constraints. This ensures precise alignment with visible segments while maintaining the overall smoothness and physical reliability of the trajectory. Discrete sampling of the complete trajectory and calculation of the overall average difference provide robust global quantitative indicators even in the presence of noise. Introducing an area index, the area enclosed by each phase trajectory and the top horizontal reference line is standardized as a metric. This not only serves as an independent basis for cross-validation but also allows for direct ranking of severity from most severe to least severe based on its size, facilitating priority allocation and handling decisions for inspection resources. Piecewise difference analysis of any two phase trajectories placed in the same coordinate system further identifies suspicious sections with continuous drooping, beyond the overall judgment, enabling preliminary location of icing. This invention not only fully utilizes the prior property of parallel and similar three-phase conductors to quickly fill in missing areas of the image but also enables rapid preliminary identification of icing. Furthermore, it can prioritize the identification of the severity of icing using methods such as area analysis and mark and track suspected localized icing. Attached Figure Description

[0040] Figure 1 This is a general schematic diagram of the method of the present invention;

[0041] Figure 2 This is a schematic diagram of image completion using the method of the present invention;

[0042] Figure 3 This is a schematic diagram illustrating the preliminary determination of icing based on area index using the method of the present invention;

[0043] Figure 4 This is a schematic diagram of the ice accretion sorting method based on area index of the present invention. Detailed Implementation

[0044] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0045] This method acquires images of transmission lines and combines image processing, catenary fitting, non-rigid deformation correction, and multi-dimensional difference analysis to complete the trajectory completion and icing detection of three-phase conductors.

[0046] like Figure 1 As shown, the present invention includes the following steps:

[0047] Acquire images of the transmission line and identify the three-phase conductors connected between the suspension points at both ends;

[0048] In the three-phase conductors, the one with the most complete identification area is selected, and the baseline trajectory is obtained by fitting the catenary model.

[0049] The baseline trajectory is translated and copied to generate the initial trajectories of the other two phase conductors; anchor point constraints are established in combination with the regions of the other two phase conductors that have been identified in the image, and the initial trajectory is non-rigidly deformed and corrected so that the corrected trajectory coincides with the real conductor in the identified region, thereby obtaining the complete trajectory of the three-phase conductor.

[0050] After obtaining the complete trajectory of the three-phase conductors, the differences between each phase conductor are compared. When the difference between the three phase conductors exceeds a preset threshold, it is preliminarily determined that the three-phase conductors are covered with ice.

[0051] Figure 2 illustrates the image completion method of this invention. In a specific embodiment, firstly, an image of the power transmission line is acquired using a high-resolution camera or an image acquisition device mounted on a drone. The image must cover the three-phase conductors and their suspension points at both ends, with a resolution of no less than 1920×1080 pixels to ensure that the conductor details are clearly visible. The acquired image undergoes preprocessing steps, such as noise reduction and contrast enhancement, to improve its quality. For example, Gaussian filtering is used to remove noise, with the standard deviation σ ranging from 0.5 to 2.0.

[0052] In a further embodiment, identifying the three-phase conductors in an image can be achieved using image segmentation algorithms, such as U-Net or MaskR-CNN, to extract the conductor regions. The U-Net model consists of an encoder and a decoder. The encoder uses convolutional and pooling layers to extract features, while the decoder restores spatial resolution through upsampling and outputs a pixel-level segmentation mask for the conductors. During model training, a transmission line image dataset labeled with conductor regions is used for approximately 50 to 100 epochs, with a batch size of 16, an Adam optimizer, and a learning rate ranging from 0.0001 to 0.001. The segmentation results are the pixel regions of the three-phase conductors, denoted as W1, W2, and W3. For each phase conductor's pixel region, its continuous length in the horizontal direction is calculated by projecting the segmentation mask along the horizontal axis and counting the maximum length of consecutive pixel segments.

[0053] The longest horizontal conductor is selected as the one with the most complete identification, denoted as Wo. For example, if the horizontal identification lengths of W1, W2, and W3 are 800 pixels, 600 pixels, and 700 pixels respectively, then W1 is selected as the reference conductor.

[0054] In another embodiment, to avoid complex model training, a model may not be used. Instead, the length of the three-phase conductors that have been photographed can be measured directly by manually viewing the images, and the conductor with the longest horizontal length area photographed can be selected as the reference.

[0055] For the selected baseline traverse Wo, extract the coordinates of its multiple discrete identification points, denoted as . , The coordinates are in the horizontal direction. The coordinates are in the vertical direction, and the subscripts are the labels of specific discrete points.

[0056] In a specific implementation, n ≥ 50 can be selected to ensure fitting accuracy. A catenary model is used to fit these points; the catenary function has the following form:

[0057] ;

[0058] Among them, parameters , , These represent the catenary tension factor, horizontal displacement, and vertical offset, respectively. This is the vertical coordinate (i.e., the coordinate in the vertical direction). The horizontal coordinate is denoted by ; the tension factor reflects the conductor curvature, typically ranging from 50 to 500 meters, depending on the conductor length and slack; the horizontal displacement range is the horizontal coordinate interval between the suspension points at both ends of the conductor; the vertical offset is usually the height of the suspension point plus or minus a certain offset. All of the above parameters must be obtained through fitting.

[0059] The objective function is to minimize the error between discrete points and the catenary function using the least squares method.

[0060] ;

[0061] Optimize parameters using gradient descent. , , The number of iterations was set to 1000, and the learning rate ranged from 0.01 to 0.1. The fitted catenary function describes the baseline trajectory Lo(x) of Wo.

[0062] Using the baseline trajectory Lo(x) of Wo, the initial trajectories of the other two phase conductors are generated by translation and replication. Assume the three phase conductors are arranged parallel in space with a spacing of d, for example, 3 to 5 meters. Translate Lo(x) ±d along a horizontal direction perpendicular to the conductor extension direction (i.e., the x-axis direction) to obtain the initial trajectories. For the identified areas of W2 and W3, extract the anchor point coordinates. m = 1, 2, ..., M, where M ≥ 10 is optional.

[0063] The initial trajectory is subjected to non-rigid deformation correction to ensure it coincides with the identified areas of W2 and W3 at the anchor points. The deformation correction is applied to the anchor point pair { } is the input, where For points on the initial trajectory, These are the identified points corresponding to the conductor.

[0064] In some embodiments, the deformation function is constructed based on thin-plate spline transformation. ,satisfy .

[0065] In some embodiments, a thin-plate spline transformation function may be used. This function is composed of radial basis functions and polynomial terms, and takes the form of... ;

[0066] in It is a low-order polynomial. For radial basis functions, These are the weighting coefficients. The solution satisfies... Calculation of linear equations and The coefficient.

[0067] The deformation function T(x,y) maps the initial trajectory globally, generating a corrected trajectory. For example, if the anchor point of W2 deviates from the initial trajectory by approximately 0.5 meters, the thin-plate spline transformation can smoothly adjust the initial trajectory to the anchor point position, maintaining trajectory continuity. Since the thin-plate spline transformation is existing technology, it will not be elaborated upon further.

[0068] After obtaining the complete trajectory of the three-phase conductors, a further embodiment requires detecting their differences to determine the icing situation.

[0069] Method 1 is discrete sampling analysis: the complete trajectory of the three-phase conductors is discretely sampled, and the set of sampling points is set as follows. Calculate the height difference between any two phase conductors at the sampling point, and obtain the overall average difference using the following formula:

[0070] ;

[0071] Indicates the first i Phase conductor and the first j Overall average difference in phase conductors; N Indicates the total number of sampling points. k For sampling point index, Indicates the first k The horizontal position of each sampling point Indicates the first i Phase conductor at sampling point The height, Indicates the first j Phase conductor at sampling point The height; among which i, j Choose 1, 2, or 3;

[0072] When the overall average difference between any two phase conductors When the threshold is exceeded, it is preliminarily determined that the three-phase conductors are covered with ice.

[0073] The rationale behind this formula is that icing increases the weight of the conductor, causing the track to sag and the height difference to increase; the average value design can reduce the impact of local noise.

[0074] As shown in Figure 3, Method 2 is an area index analysis:

[0075] For each phase conductor, the line connecting the suspension points at both ends of that phase conductor is used as the top horizontal reference line, and the trajectory of that phase conductor is calculated accordingly. Area index enclosed by the corresponding top horizontal reference line :

[0076] ;

[0077] in, The horizontal length of the top horizontal reference line. The starting point of the horizontal coordinate is located at one of the hanging points. Indicates the first i Phase conductors on the horizontal coordinate x The height, Indicates the first i Height of suspension points at both ends of the phase conductor;

[0078] When the area index difference between any two phases When the threshold is exceeded, it is preliminarily determined that the three-phase conductors are icing. i, j Choose 1, 2, or 3.

[0079] The advantage of the area index is that it comprehensively considers the entire trajectory shape and is suitable for detecting uniform icing. More ice accumulation leads to a larger area. Therefore, as shown in Figure 4, further comparisons are made... The size of the area index determines the degree of icing. The larger the area index, the more severe the conductor sag and the greater the degree of icing.

[0080] In a further embodiment, the trajectories of any two phase conductors are placed in the same coordinate system, and the trajectory difference is calculated segment by segment. The conductor is divided into K segments along the horizontal axis, where K = 10 to 50, and the height difference of each segment is calculated. If within a certain segment If the value remains negative, the segment is preliminarily identified as an icing area. For example, if W2 is consistently about 0.1 meters lower than W1 within a 50-meter segment, that segment may be icy. This segmented analysis is suitable for detecting localized icing.

[0081] Through the above embodiments, the present invention realizes the entire process from image acquisition to icing detection. The present invention makes full use of the characteristic that the three-phase conductors tend to be parallel to simplify the image completion process and provides a preliminary detection method for icing, which facilitates further investigation in the later stage.

[0082] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting icing based on three-phase transmission line image completion, characterized in that, include Acquire images of the transmission line and identify the three-phase conductors connected between the suspension points at both ends; Select the one phase conductor with the most complete identification area from the three phase conductors, and use the catenary model to fit the baseline trajectory. The reference trajectory is translated and copied to generate the initial trajectories of the other two phase conductors; By combining the anchor point constraints established in the regions of the other two phase conductors that have been identified in the image, the initial trajectory is non-rigidly deformed and corrected so that the corrected trajectory coincides with the real conductor in the identified region, thereby obtaining the complete trajectory of the three-phase conductor. After obtaining the complete trajectory of the three-phase conductors, the differences between each phase conductor are compared. When the difference between the three-phase conductors exceeds a preset threshold, it is preliminarily determined that the three-phase conductors are covered with ice. The process of detecting differences between three-phase conductors includes: Discretely sample the complete trajectory of the three-phase conductors, and define the set of sampling points as follows: Calculate the height difference between any two phase conductors at the sampling point, and obtain the overall average difference using the following formula: ; Indicates the first i Phase conductor and the first j Overall average difference in phase conductors; N Indicates the total number of sampling points. k For sampling point index, Indicates the first k The horizontal position of each sampling point Indicates the first i Phase conductor at sampling point height, Indicates the first j Phase conductor at sampling point The height; among which i, j Choose 1, 2, or 3; When the overall average difference between any two phase conductors When the threshold is exceeded, it is preliminarily determined that the three-phase conductors are covered with ice.

2. The icing detection method based on three-phase transmission line image completion according to claim 1, characterized in that, The non-rigid deformation correction includes: Using the initial trajectory and the anchor point pair of the corresponding identified area of ​​the traverse as input, a set of deformation control points is established; based on the set of deformation control points, a deformation function is constructed using thin plate spline transformation or B-spline free deformation method; the deformation function is used to map the initial trajectory as a whole, so that it coincides with the identified trajectory at the anchor point.

3. The icing detection method based on three-phase transmission line image completion according to claim 1, characterized in that, The method further includes placing the trajectories of any two phase conductors in the same coordinate system for comparison, and performing segmented calculations on the trajectory differences. When it is detected that a certain phase conductor is continuously lower than another phase conductor in a local area, it is preliminarily determined that there is icing in that local area.

4. The icing detection method based on three-phase transmission line image completion according to claim 1, characterized in that, The process of fitting using a catenary includes: Extract the coordinates of multiple discrete identification points of the conductor and establish the catenary function: ; Among them, parameters , , These represent the catenary tension factor, horizontal displacement, and vertical offset, respectively. The vertical axis is... The horizontal axis; The parameters are determined by fitting the error between discrete identification points and the catenary function using the least squares method. , , The optimal solution is obtained, thus yielding the complete catenary trajectory of the reference traverse.

5. The icing detection method based on three-phase transmission line image completion according to claim 2, characterized in that, The deformation function is a two-dimensional coordinate mapping function constructed from the anchor point positions. ,satisfy: ； in Indicates the first position on the initial trajectory Anchor point coordinates, This indicates the coordinates of the corresponding identified guide wire anchor points. This represents the total number of anchor points.

6. The icing detection method based on three-phase transmission line image completion according to claim 1, characterized in that, The step of selecting the phase conductor with the most complete identification area from the three phase conductors includes: comparing the identification results of the three phase conductors, calculating the length of each phase conductor identified in the horizontal direction, and selecting the conductor with the longest length as the phase conductor with the most complete identification.

7. The icing detection method based on three-phase transmission line image completion according to claim 6, characterized in that, In the process of identifying the most complete phase conductor, the completeness of the conductor identification in the image is determined by manual interpretation, or by calculating the identification segment length of each phase conductor using an image recognition algorithm, thereby selecting the phase conductor with the most complete identification area as the reference conductor.

Images