Pole-mounted circuit breaker installation position offset detection method

Abstract

The invention relates to the technical field of image data processing, in particular to a pole-mounted circuit breaker installation position deviation detection method, which comprises the following steps of: extracting two edge straight lines of a telegraph pole in a circuit breaker image to obtain a vertical vanishing point; calculating the weight of each pixel point in the circuit breaker pixel point set; performing weighting processing on each pixel point in the circuit breaker pixel point set to obtain a weighted pixel point set so as to obtain a mass center and a covariance matrix of the circuit breaker; performing principal component analysis on the covariance matrix to obtain a direction vector of an attitude principal axis; the center of mass and the direction vector of the attitude principal axis jointly form an attitude principal axis straight line; connecting the center of mass with the vertical vanishing point to obtain a reference vertical line; obtaining a deviation angle according to an included angle between the attitude spindle straight line and the reference vertical line; and in response to the deviation angle being greater than the set threshold value, it is determined that the installation position of the circuit breaker deviates, so that the accuracy of the deviation detection result is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing. More particularly, the present application relates to a pole-mounted circuit breaker installation position offset detection method. BACKGROUND

[0002] The pole-mounted circuit breaker is a key protection device in the power distribution network, and its installation quality is directly related to the safe and stable operation of the power grid. During installation, a slight tilt or position deviation may be amplified over time due to wind vibration, natural disasters, and other external forces, leading to abnormal mechanical stress and potentially causing equipment failure, which poses a serious threat to power supply reliability. Therefore, in the engineering acceptance standard, it is required that the circuit breaker must maintain a strict vertical posture on the pole compared to the ground.

[0003] Currently, the industry generally uses unmanned aerial vehicles or handheld devices to take images, and then uses computer vision algorithms for analysis and detection.

[0004] However, in actual operating scenarios, in order to completely capture the circuit breaker at a high altitude into the frame, the photographer usually needs to shoot upward, resulting in perspective distortion in the captured image, which makes the pole and the circuit breaker on it, which were originally vertical to the ground in three-dimensional space, appear tilted in the two-dimensional image. The existing detection method cannot effectively distinguish between this visual offset caused by the shooting angle and the real physical installation offset of the circuit breaker, which may misjudge this installation qualified circuit breaker as a tilted state, has limitations, resulting in low accuracy of the detection result, and cannot meet the needs of accurate acceptance on site.

[0005] Therefore, how to use a single two-dimensional image taken at an arbitrary angle to eliminate the influence of perspective distortion and accurately obtain the real installation posture detection result of the pole-mounted circuit breaker is a problem that needs to be solved urgently. SUMMARY

[0006] To solve the above technical problem of how to use a single two-dimensional image taken at an arbitrary angle to eliminate the influence of perspective distortion and accurately obtain the real installation posture of the pole-mounted circuit breaker, the present application proposes a pole-mounted circuit breaker installation position offset detection method, which includes the following steps: An image of a circuit breaker installed on a telegraph pole is acquired, two edge straight lines of the telegraph pole in the circuit breaker image are extracted, and the intersection of the edge straight lines is determined as a vertical vanishing point; a set of circuit breaker pixel points in the circuit breaker image is acquired, the weight of each pixel point is calculated, the weight is positively correlated with the distance of each pixel point to the vertical vanishing point, is positively correlated with the gray scale standard deviation of the local neighborhood where each pixel point is located, and is negatively correlated with the difference between the brightness value of each pixel point and a set brightness ideal value; each pixel point in the set of circuit breaker pixel points is subjected to weighted processing to obtain a set of weighted pixel points, so as to obtain the centroid and the covariance matrix of the circuit breaker; principal component analysis is performed on the covariance matrix to obtain a direction vector of the posture principal axis; the centroid and the direction vector of the posture principal axis jointly constitute a posture principal axis straight line; a reference vertical line is obtained by connecting the centroid and the vertical vanishing point; an offset angle is obtained according to the included angle between the posture principal axis straight line and the reference vertical line; and in response to the offset angle being greater than a set threshold value, it is determined that the installation position of the circuit breaker is offset.

[0007] By eliminating the perspective distortion influence in the circuit breaker image, the installation position offset detection result of the circuit breaker can be accurately obtained. In the process of acquiring the detection result, by introducing the vertical vanishing point as an absolute vertical reference in the image and acquiring the posture principal axis straight line of the circuit breaker, the offset degree of the circuit breaker and the telegraph pole in the three-dimensional reality can be accurately obtained according to the included angle between the two, so that the installation position offset detection result of the circuit breaker can be accurately obtained. In the process of acquiring the posture principal axis straight line of the circuit breaker, by giving higher weight to the pixel points with less distortion and illumination influence and richer information, the influence of such pixel points on the centroid position can be reduced, so that the measurement error caused by the perspective distortion can be reduced, the actual installation offset of the circuit breaker can be accurately detected at any upward shooting angle, and the accuracy of the detection result is improved.

[0008] According to the column circuit breaker installation position offset detection method provided by the application, the circuit breaker image installed on the telegraph pole is acquired, and two edge straight lines of the telegraph pole in the circuit breaker image are extracted, including: taking a photo containing the telegraph pole and the circuit breaker installed on the telegraph pole and preprocessing to obtain the circuit breaker image; extracting the telegraph pole mask graph in the circuit breaker image through semantic segmentation, using Suzuki algorithm to detect the contour of the telegraph pole mask graph to obtain a contour point set, and performing straight line fitting on the contour point set to obtain two edge straight lines of the telegraph pole in the circuit breaker image.

[0009] The application considers that there may be other clutter lines in the circuit breaker image, which affect the accuracy of the extraction of the edge straight lines of the telegraph pole, therefore, the application processes the image through steps such as semantic segmentation, Suzuki algorithm and straight line fitting, so that the extracted edge straight lines belong to the telegraph pole, and preparation is made for accurately determining the vertical vanishing point subsequently.

[0010] The application provides a method for detecting installation position offset of a pole-mounted circuit breaker, wherein a YOLACT network model is pre-trained to extract a pixel set of the circuit breaker in a circuit breaker image.

[0011] The application can accurately and quickly extract the pixel set of the circuit breaker from the circuit breaker image by pre-training the YOLACT network model to extract the pixel set of the circuit breaker in the circuit breaker image, and ensure the accuracy and integrity of the pixel set of the circuit breaker.

[0012] The application provides a method for detecting installation position offset of a pole-mounted circuit breaker, wherein the weight of each pixel is calculated, including: The weight of the i-th pixel is, The distance of the i-th pixel to the vertical vanishing point is, The gray standard deviation of the local neighborhood of the i-th pixel is, The absolute value of the difference between the brightness value of the i-th pixel and the set brightness ideal value is, , , The distortion weight coefficient, the edge intensity weight coefficient and the brightness uniformity weight coefficient are respectively, The zero prevention coefficient is.

[0013] The application provides a method for detecting installation position offset of a pole-mounted circuit breaker, and the method for obtaining the centroid of the circuit breaker includes: obtaining the weighted horizontal coordinate sum value of all the pixels in the weighted pixel set, and recording the ratio of the weighted horizontal coordinate sum value to the weight sum value of all the pixels as the horizontal coordinate of the centroid of the circuit breaker; and obtaining the weighted vertical coordinate sum value of all the pixels in the weighted pixel set, and recording the ratio of the weighted vertical coordinate sum value to the weight sum value of all the pixels as the vertical coordinate of the centroid of the circuit breaker.

[0014] The application provides an accurate method for calculating the centroid of the circuit breaker, the coordinates of each pixel are weighted by the weight of each pixel, so that the finally obtained centroid can accurately reflect the center position of the high-weight pixel, and the foundation is prepared for subsequent determination of the posture main shaft straight line and the reference vertical line.

[0015] The application provides a method for detecting installation position offset of a pole-mounted circuit breaker, and the method for performing principal component analysis on the covariance matrix to obtain the direction vector of the posture main shaft includes: determining the eigenvector corresponding to the maximum eigenvalue as the direction vector of the posture main shaft.

[0016] The application provides a method for detecting installation position offset of a pole-mounted circuit breaker, and the centroid and the direction vector of the posture main shaft jointly constitute the posture main shaft straight line, including: ​ ; is a coordinate of a centroid of the circuit breaker, is an arbitrary real number, is a direction vector of the attitude principal axis.

[0017] According to the circuit breaker installation position offset detection method provided by the present application, the reference vertical line of the circuit breaker image is obtained by the method comprising: ; is a reference vertical line of the circuit breaker image, is a coordinate of a centroid of the circuit breaker, is an arbitrary real number, is a vertical vanishing point.

[0018] According to the circuit breaker installation position offset detection method provided by the present application, the installation position of the circuit breaker is determined to be qualified in response to the offset angle being not greater than a set threshold.

[0019] According to the circuit breaker installation position offset detection method provided by the present application, the installation position of the circuit breaker is determined to be offset, and then further comprising: after associating the circuit breaker image with the offset angle, an abnormal prompt is sent out.

[0020] After the circuit breaker with offset is identified, the present application sends out an abnormal prompt in time, which can timely remind the staff to handle and effectively avoid the safety hazards caused by improper installation.

[0021] The present application has the following beneficial effects: Based on the above technical solution, the circuit breaker installation position offset detection method provided by the present application can accurately obtain the installation position offset detection result of the circuit breaker by eliminating the perspective distortion influence in the circuit breaker image. In the process of obtaining the detection result, the present application introduces the vertical vanishing point as the absolute vertical reference in the image and obtains the attitude principal axis straight line of the circuit breaker. According to the included angle between the two, the offset degree of the circuit breaker and the telegraph pole in the three-dimensional reality can be accurately obtained, so that the circuit breaker installation position offset detection result can be accurately obtained. When obtaining the attitude principal axis straight line of the circuit breaker, the present application can reduce the influence of such pixel points on the centroid position by giving higher weight to the pixel points with less distortion and light influence and more abundant information, so as to reduce the measurement error caused by perspective distortion and accurately detect the actual installation offset of the circuit breaker under any upward shooting angle, thereby improving the accuracy of the detection result. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 ​is a step flow chart of a pole-mounted circuit breaker installation position offset detection method. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application.

[0024] The perspective distortion existing in the circuit breaker image installed on the power pole can identify the normally installed circuit breaker as a physical installation offset.

[0025] Based on this, the embodiment of the present application provides a pole-mounted circuit breaker installation position offset detection method, which is based on the characteristics that the power pole and the circuit breaker installation direction are parallel in reality and the power pole and the circuit breaker installation direction are both perpendicular to the ground, to obtain the actual offset angle between the true vertical line in the circuit breaker image and the circuit breaker, analyze the possibility of existing offset, and effectively improve the accuracy of the circuit breaker installation position offset detection.

[0026] Please refer to Figure 1 , Figure 1 is a step flow chart of a pole-mounted circuit breaker installation position offset detection method, and specifically includes the following steps: S1: Obtain a circuit breaker image installed on a power pole, extract two edge straight lines of the power pole in the circuit breaker image, and determine the intersection point of the edge straight lines as a vertical vanishing point.

[0027] For example, in the embodiment of the present application, obtaining a circuit breaker image installed on a power pole and extracting two edge straight lines of the power pole in the circuit breaker image include: taking a photo containing a power pole and a circuit breaker installed on the power pole and performing preprocessing to obtain a circuit breaker image; extracting a power pole mask graph in the circuit breaker image through semantic segmentation, using Suzuki algorithm to detect the contour of the power pole mask graph to obtain a contour point set, and performing straight line fitting on the contour point set to obtain two edge straight lines of the power pole in the circuit breaker image.

[0028] Specifically, after taking a photo containing a power pole and a circuit breaker installed on the power pole, the photo can be subjected to grayscale processing, noise reduction and other preprocessing operations to reduce the interference of irrelevant details.

[0029] It can be understood that there are many messy wires on the power pole, so the complete power pole edge straight line is divided into a plurality of small line segments, and therefore when the edge straight line of the power pole in the circuit breaker image is directly extracted, other straight lines may be misjudged as the edge straight line of the power pole.

[0030] Therefore, the two edge straight lines of the power pole in the circuit breaker image can be accurately extracted by semantic segmentation, Suzuki algorithm detection and straight line fitting.

[0031] It should be noted that the edge straight lines of the power pole are two completely parallel straight lines, which can be used as a reference for the installation position of the circuit breaker. However, due to the perspective distortion of the captured circuit breaker image, the two edge straight lines will become intersecting lines, and there will be an intersection point between the two edge straight lines. The intersection point is the vanishing point, which can eliminate the problem of dislocation between the two-dimensional image and the three-dimensional physical space caused by perspective distortion, and represents the only absolute vertical base point of the three-dimensional physical world in the two-dimensional image. The pixel points in different positions of the circuit breaker pixel point set are affected by the perspective distortion to different degrees.

[0032] Based on this, the embodiments of the present application take the vanishing point as the base point, weight each pixel point according to the distance between the pixel point and the base point and the gray scale distribution, and determine the main axis direction of the circuit breaker according to the weighted pixel point set, which can effectively reduce the influence of perspective distortion on the determination of the main axis direction of the circuit breaker, that is, the following steps are performed.

[0033] S2: obtaining the circuit breaker pixel point set in the circuit breaker image, and calculating the weight of each pixel point.

[0034] The weight is positively correlated with the distance of each pixel point to the vanishing point, positively correlated with the gray scale standard deviation of the local neighborhood where each pixel point is located, and negatively correlated with the difference between the brightness value of each pixel point and the set brightness ideal value.

[0035] It should be noted that the background of the captured circuit breaker image will be relatively messy, and there may be sky, branches and power lines, etc. Therefore, when analyzing the posture of the circuit breaker, the circuit breaker pixel point set can be extracted from the circuit breaker image first, and the pixel points can be weighted according to the distance between the pixel points and the vanishing point and the gray scale distribution around the pixel points, and finally the subsequent analysis is based on the weighted circuit breaker pixel point set.

[0036] For example, in the embodiments of the present application, the circuit breaker pixel point set in the circuit breaker image is extracted by a pre-trained YOLACT network model.

[0037] Specifically, when pre-training the YOLACT network model, a plurality of circuit breaker images can be collected as samples to obtain a circuit breaker image set, and the circuit breaker pixel points on each sample in the circuit breaker image set are manually labeled to obtain a labeled circuit breaker image set; the labeled circuit breaker image set is input into the YOLACT network model for training to obtain the YOLACT network model which can be used to extract the circuit breaker pixel point set in the circuit breaker image.

[0038] The circuit breaker image to be extracted is input into the pre-trained YOLACT network model, the model outputs a pixel mask, in which the pixel value of the circuit breaker body is 1 and the pixel value of the background part is 0, and the coordinates of the pixel points with a value of 1 are extracted to obtain the circuit breaker pixel point set in the circuit breaker image.

[0039] After obtaining the circuit breaker pixel point set according to the above steps, the weight of each pixel point in the circuit breaker pixel point set can be calculated according to the following steps.

[0040] It can be understood that the more close to the vertical vanishing point, the more serious the distortion of the pixel point, and in addition, due to the overexposure or underexposure phenomenon in the circuit breaker image taken in bad weather, the extraction of the circuit breaker centroid is affected. Therefore, the embodiment of the present application can comprehensively obtain the final weight by analyzing the pixel point distortion weight, the edge intensity weight and the brightness uniformity weight.

[0041] For example, in the embodiment of the present application, the weight of each pixel point in the circuit breaker pixel point set comprises: ; The weight of the i-th pixel point is The distance of the i-th pixel point to the vertical vanishing point is The gray standard deviation of the local neighborhood where the i-th pixel point is located is The absolute value of the difference between the brightness value of the i-th pixel point and the set brightness ideal value is The distortion weight coefficient is The edge intensity weight coefficient is The brightness uniformity weight coefficient is The zero prevention coefficient is It can be set to 1.

[0042] Wherein, when obtaining the local neighborhood where the current pixel point is located, a 5*5 size rectangular region can be constructed with the current pixel point as the center, and the size of the local neighborhood can be set according to actual needs, which is not limited here. The set brightness ideal value can be set to half of the maximum brightness value, i.e. 128, which can be set according to actual needs.

[0043] For example, when setting the weight coefficient, the distortion weight of the pixel point can be set to a larger distortion weight coefficient, and the edge intensity weight coefficient and the brightness uniformity weight coefficient can be set according to the edge and texture clarity in the circuit breaker image.

[0044] As an example, The distortion weight coefficient can be set to 1.5, The edge intensity weight coefficient can be set to 1.5, and The brightness uniformity weight coefficient can be set to 1.All are set to 0.8; can be set according to actual needs.

[0045] In the above calculation formula, The distortion weight is represented, the farther the distance from the i-th pixel point to the vertical vanishing point, the smaller the distortion degree of the pixel point, the greater the corresponding distortion weight, the higher the reliability of the pixel point, and the greater the corresponding weight.

[0046] The edge intensity weight is represented, the greater the gray standard deviation of the local neighborhood where the i-th pixel point is located, the richer the edge or texture around the current pixel point, the higher the reliability of the pixel point as a feature point, and the greater the corresponding weight.

[0047] The brightness uniformity weight is represented, the greater the difference between the brightness value of the i-th pixel point and the set brightness ideal value, the more serious the influence of light on the current pixel point, the more difficult to reflect the real position of the circuit breaker, and the smaller the corresponding weight needs to be, so as to reduce the influence of the pixel point.

[0048] In summary, the greater the distance from the pixel point to the vertical vanishing point and the gray standard deviation of the local neighborhood where the pixel point is located, and the smaller the difference between the brightness value of the pixel point and the set brightness ideal value, the greater the weight of the pixel point.

[0049] According to the above steps, the weight of each pixel point in the circuit breaker pixel point set can be obtained, and the mass center and the direction of the attitude principal axis of the circuit breaker can be accurately obtained by weighting the pixel points according to the weight, so as to accurately obtain the attitude principal axis straight line of the circuit breaker, that is, the following steps are executed.

[0050] S3: Weighting processing is performed on each pixel point in the circuit breaker pixel point set to obtain a weighted pixel point set, so as to obtain the mass center and the covariance matrix of the circuit breaker; principal component analysis is performed on the covariance matrix to obtain a direction vector of the attitude principal axis; the mass center and the direction vector of the attitude principal axis jointly constitute the attitude principal axis straight line.

[0051] It should be noted that in the image without perspective distortion, the role of each pixel point in the circuit breaker pixel point set in determining the mass center of the circuit breaker is consistent, but if there is perspective distortion, some pixel points may be severely deformed, so the weight obtained according to the above steps is needed to weight the pixel points to determine the real mass center and the covariance matrix of the circuit breaker, and the attitude principal axis straight line of the circuit breaker is obtained based on this. The attitude principal axis straight line of the circuit breaker quantizes the installation orientation of the circuit breaker into a straight line, if the installation position of the circuit breaker is correct, the attitude principal axis straight line of the circuit breaker is the vertical direction with the widest pixel point distribution; on the contrary, if the installation position of the circuit breaker is inclined, the direction with the widest pixel point distribution will also be inclined.

[0052] Based on this, the embodiment of the present application can obtain the centroid of the circuit breaker and the direction vector of the attitude principal axis of the circuit breaker according to the weighted pixel points, so as to construct the direction of the attitude principal axis of the circuit breaker.

[0053] For example, in the embodiment of the present application, the centroid of the circuit breaker is obtained in the following manner: the weighted horizontal coordinate sum value of all the pixel points in the weighted pixel point set is obtained, the ratio of the weighted horizontal coordinate sum value to the weight sum value of all the pixel points is recorded as the horizontal coordinate of the centroid of the circuit breaker; the weighted vertical coordinate sum value of all the pixel points in the weighted pixel point set is obtained, and the ratio of the weighted vertical coordinate sum value to the weight sum value of all the pixel points is recorded as the vertical coordinate of the centroid of the circuit breaker.

[0054] It can be understood that the greater the weight of the pixel point, the higher the influence degree on the position of the centroid.

[0055] In the centroid obtaining manner, the weight sum value of all the pixel points is the total reliability of all the pixel points in the weighted pixel point set, which is used for normalized weighted average of the position coordinates.

[0056] The centroid obtained according to the above steps can construct a 2*2 covariance matrix, which contains all the directional information of the circuit breaker weighted pixel point set, and can reflect the dispersion degree and the inclination degree of the weighted pixel point set.

[0057] Specifically, the covariance matrix is wherein, is the variance of the circuit breaker weighted pixel point set in the X direction with the centroid as the reference point, is the variance of the circuit breaker weighted pixel point set in the Y direction with the centroid as the reference point, is the weighted covariance between the horizontal coordinate and the vertical coordinate of each pixel point in the circuit breaker weighted pixel point set.

[0058] wherein, is the distribution width of the circuit breaker weighted pixel point set in the X direction around the centroid, and the greater the value, the longer the weighted pixel point set in the X direction.

[0059] is the distribution width of the circuit breaker weighted pixel point set in the Y direction around the centroid, and the greater the value, the longer the weighted pixel point set in the Y direction.

[0060] and is used to represent the dispersion degree of the weighted pixel point set, is used to represent the inclination degree of the weighted pixel point set, and if , it indicates that the principal direction vector of the weighted pixel point set is not parallel to the coordinate axis.

[0061] For example, in the embodiment of the present application, the covariance matrix is subjected to principal component analysis to obtain the direction vector of the attitude principal axis, including: determining the eigenvector corresponding to the maximum eigenvalue as the direction vector of the attitude principal axis.

[0062] The eigenvector corresponding to the maximum eigenvalue is the direction with the maximum variance, which physically corresponds to the long axis of the circuit breaker.

[0063] For example, in the embodiment of the present application, the direction vector of the attitude principal axis and the centroid jointly constitute the attitude principal axis straight line, including: ; is the attitude principal axis straight line of the circuit breaker, is the coordinate of the centroid of the circuit breaker, is any real number, is the direction vector of the attitude principal axis.

[0064] In the above calculation formula, the centroid provides a starting point for the attitude principal axis straight line, that is, the finally obtained attitude principal axis straight line passes through the real geometric centroid of the weighted pixel point set, and the direction vector of the attitude principal axis is used to provide the extension direction of the attitude principal axis straight line. By combining the centroid and the direction vector of the attitude principal axis to traverse all real numbers, an attitude principal axis straight line with the centroid as the starting point and extending along the direction vector of the principal axis can be obtained, which is used to reflect the installation angle of the circuit breaker on the pole.

[0065] S4: obtaining a reference vertical line by connecting the centroid and the vertical vanishing point; obtaining an offset angle according to the included angle between the attitude principal axis straight line and the reference vertical line; in response to the offset angle being greater than a set threshold, determining that the installation position of the circuit breaker is offset.

[0066] It should be noted that the vertical vanishing point indirectly marks the vertical direction in reality, and the direction of the reference vertical line obtained by connecting the centroid and the vertical vanishing point, that is, the reference vertical line is a straight line extending along the direction of the vertical pole in reality, so the reference vertical line is taken as the reference of complete verticality in the circuit breaker image, and by obtaining the included angle between the reference vertical line and the attitude principal axis straight line, the offset degree of the circuit breaker from the vertical direction in reality can be accurately obtained.

[0067] For example, in the embodiment of the present application, the reference vertical line of the circuit breaker image is obtained by: ; is the reference vertical line of the circuit breaker image, is the coordinate of the centroid of the circuit breaker, is any real number, is the vertical vanishing point.

[0068] For the direction vector of the reference vertical line, a direction from the centroid of the circuit breaker to the vanishing point of the vertical line is obtained.

[0069] It can be understood that the reference vertical line in the circuit breaker image and the attitude principal axis straight line of the circuit breaker can be obtained respectively according to the above steps, and by obtaining the included angle between the two, the influence of the perspective distortion can be eliminated, and therefore the obtained included angle is also the offset angle between the circuit breaker in the three-dimensional reality and the real vertical direction.

[0070] Among them, the included angle between the two straight lines can be obtained by means of direction vector dot product, Hough transform, straight line slope, etc., and the present embodiment does not make too many limitations here.

[0071] According to the above steps, the included angle between the attitude principal axis straight line and the reference vertical line can be accurately obtained, and the included angle is the offset angle.

[0072] For example, in the present embodiment, in response to the offset angle being greater than the set threshold, it is determined that the installation position of the circuit breaker is offset; and in response to the offset angle being not greater than the set threshold, it is determined that the installation position of the circuit breaker is qualified.

[0073] Among them, the set threshold can be any angle in 1° to 5°, and can be set according to actual needs, and the present embodiment does not make too many limitations here.

[0074] For example, in the present embodiment, after it is determined that the installation position of the circuit breaker is offset, it further includes: after associating the circuit breaker image with the offset angle, an abnormal prompt is sent out.

[0075] Among them, the abnormal prompt mode can be marked box highlighting, generating a separate abnormal image library, etc., and can be set according to actual needs.

[0076] It can be seen that in the embodiment of the present application, when the installation position offset detection result of the pole-mounted circuit breaker is obtained, the image of the circuit breaker installed on the electric pole can be obtained, two edge straight lines of the electric pole in the circuit breaker image are extracted, and the intersection of the edge straight lines is determined as the vertical vanishing point; the pixel point set of the circuit breaker in the circuit breaker image is obtained, the weight of each pixel point is calculated, the weight is positively correlated with the distance of each pixel point to the vertical vanishing point, is positively correlated with the gray standard deviation of the local neighborhood where each pixel point is located, and is negatively correlated with the difference between the brightness value of each pixel point and the set brightness ideal value; each pixel point in the pixel point set of the circuit breaker is weighted to obtain a weighted pixel point set, so as to obtain the centroid and the covariance matrix of the circuit breaker; the principal component analysis is performed on the covariance matrix to obtain the direction vector of the posture principal axis; the centroid and the direction vector of the posture principal axis jointly constitute the posture principal axis straight line; the reference vertical line is obtained by connecting the centroid and the vertical vanishing point; the offset angle is obtained according to the included angle between the posture principal axis straight line and the reference vertical line; in response to the offset angle being greater than a set threshold, it is determined that the installation position of the circuit breaker exists offset, the influence of the perspective distortion in the image can be reduced, and the accuracy of the offset detection result is effectively improved.

[0077] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A pole-mounted circuit breaker installation position offset detection method characterized by, The method comprises the following steps: An image of a circuit breaker installed on a telegraph pole is acquired, two edge straight lines of the telegraph pole in the circuit breaker image are extracted, and a vertical vanishing point is determined as the intersection of the two edge straight lines; A set of circuit breaker pixel points in the circuit breaker image is acquired, the weight of each pixel point is calculated, the weight is positively correlated with the distance of each pixel point to the vertical vanishing point, positively correlated with the gray standard deviation of the local neighborhood where each pixel point is located, and negatively correlated with the difference between the brightness value of each pixel point and a set brightness ideal value; Each pixel point in the set of circuit breaker pixel points is subjected to weighted processing to obtain a set of weighted pixel points, so as to obtain the centroid and the covariance matrix of the circuit breaker; principal component analysis is performed on the covariance matrix to obtain a direction vector of the posture principal axis; the centroid and the direction vector of the posture principal axis jointly constitute a posture principal axis straight line; A reference vertical line is obtained by connecting the centroid and the vertical vanishing point; an offset angle is obtained according to the included angle between the posture principal axis straight line and the reference vertical line; in response to the offset angle being greater than a set threshold, it is determined that the installation position of the circuit breaker is offset.

2. The pole-mounted circuit breaker installation position offset detection method of claim 1, wherein, The method comprises the following steps: A photo containing a telegraph pole and a circuit breaker installed on the telegraph pole is taken and preprocessed to obtain a circuit breaker image; A telegraph pole mask graph in the circuit breaker image is extracted through semantic segmentation, a contour point set is obtained by using the Suzuki algorithm to detect the contour of the telegraph pole mask graph, and two edge straight lines of the telegraph pole in the circuit breaker image are obtained by performing straight line fitting on the contour point set.

3. The method of claim 1, wherein, A set of circuit breaker pixel points in the circuit breaker image is extracted through a pre-trained YOLACT network model.

4. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The method comprises the following steps: ； Wi is the weight of the i-th pixel point, Di is the distance from the i-th pixel point to the vertical vanishing point, Si is the gray scale standard deviation of the local neighborhood of the i-th pixel point, Vi is the absolute value of the difference between the i-th pixel point and the set ideal value of the brightness, , , respectively are the distortion weight coefficient, the edge intensity weight coefficient, and the brightness uniformity weight coefficient, is the prevention of zero coefficient.

5. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The method for obtaining the centroid of the circuit breaker comprises the following steps: A weighted horizontal coordinate sum value of all pixel points in the set of weighted pixel points is obtained, and the ratio of the weighted horizontal coordinate sum value to the weight sum value of all pixel points is recorded as the horizontal coordinate of the centroid of the circuit breaker; A weighted vertical coordinate sum value of all pixel points in the set of weighted pixel points is obtained, and the ratio of the weighted vertical coordinate sum value to the weight sum value of all pixel points is recorded as the vertical coordinate of the centroid of the circuit breaker.

6. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The method comprises the following steps: The eigenvector corresponding to the maximum eigenvalue is determined as the direction vector of the posture principal axis.

7. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The centroid and the direction vector of the posture principal axis jointly constitute a posture principal axis straight line. ； is the position of the center of mass of the circuit breaker, is the position of the center of mass of the circuit breaker, is an arbitrary real number, is the direction vector of the attitude principal axis.

8. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The method for obtaining the reference vertical line of the circuit breaker image comprises the following steps: ； is a reference vertical line for the circuit breaker image, is a coordinate of the centroid of the circuit breaker, is any real number, is a vertical vanishing point.

9. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, In response to the offset angle being not greater than the set threshold, it is determined that the installation position of the circuit breaker is qualified.

10. The method of pole-mounted circuit breaker installation position offset detection of claim 1, wherein, The method further comprises the following steps after determining that the installation position of the circuit breaker is offset: After associating the circuit breaker image with the offset angle, an abnormality prompt is sent out.

Images