Method for detecting state of isolating switch of power transformation equipment

Abstract

The invention discloses a power transformation equipment disconnecting switch state detection method, and belongs to the field of image processing. According to the method, firstly, four template images are selected in each scene in a normally-working disconnecting switch reference video, a corresponding mask is added, only a disconnecting switch area needing to be concerned is reserved in the mask image, AND operation is carried out on each template image and the mask image to obtain a template saliency disconnecting switch area, and then the template saliency disconnecting switch area is obtained; and then performing AND operation on the to-be-detected image frame and the mask image to obtain a to-be-detected saliency isolation switch area, performing structural similarity matching on the to-be-detected saliency isolation switch area and the template saliency isolation switch area, and for the image frame which is matched to be normal in opening and normal in closing, judging whether the to-be-detected saliency isolation switch area is normal or not. And local template matching is carried out with opening and closing target region templates, and finally, whether opening and closing are normal or not is further determined according to comparison between normalization coefficient matching similarity and a threshold value. According to the invention, the recognition accuracy and robustness are improved.

Description

Technical Field

[0001] The invention relates to the field of image processing, and in particular to a method for detecting the state of an isolating switch of a transformer equipment. Background Art

[0002] High-voltage disconnectors are the most used switchgear in power systems. After long-term operation, atmospheric corrosion, loss of grease in the transmission mechanism joints, accumulation of foreign matter and dust, etc., will cause disconnectors to rust and jam, operational failure, switch failure, refusal to operate and other mechanical failures. Therefore, it is very necessary to detect high-voltage disconnector failures in a timely manner. Once a failure occurs, it will pose a major threat to the power system and personal safety. Therefore, regular inspection and maintenance of high-voltage disconnectors is crucial, but relying solely on manual inspections is extremely costly and has lags.

[0003] With the continuous advancement of science and technology, it has become possible to use video to monitor the status of high-voltage disconnectors in real time, greatly reducing the work pressure and safety risks of maintenance personnel, ensuring that any abnormal opening and closing of high-voltage disconnectors can be discovered in time at any time, so that corresponding emergency measures can be taken to prevent the expansion of faults and protect the safe and stable operation of power equipment and the entire power system.

[0004] In the prior art, after the image is binarized, the Hough line segments and the tilt angle information or the connecting rod pixels between the two end connectors are analyzed to determine whether the disconnector device is in the closed or open state, thereby realizing the identification of the state of the disconnector device. However, due to the complexity of the use scenario, there may be similar straight or broken line pipelines, wires, other power structure equipment, lightning and other complex situations in the background, which will overlap with the disconnector in the binary image, making it impossible to distinguish whether it is the disconnector part or other structures in the background. The use of this image binarization solution cannot realize the state analysis of the disconnector in complex use scenarios. Summary of the invention

[0005] The invention provides a method for detecting the state of an isolating switch of a power transformation device, which improves the recognition accuracy and robustness.

[0006] The present invention provides the following technical solutions:

[0007] A method for detecting the state of a power transformer isolating switch, comprising:

[0008] S10: a camera capable of capturing the entire isolating switch is provided for each isolating switch to be monitored;

[0009] S20: Using the camera to collect reference videos of each disconnector from normal opening to normal closing and from normal closing to normal opening;

[0010] S30: extracting a normal opening template image, a normal closing template image, an abnormal opening template image, and an abnormal closing template image from a reference video of each disconnector;

[0011] S40: corresponding to each normal opening template image, normal closing template image, abnormal opening template image, and abnormal closing template image, respectively obtain a significant disconnector region template image that only retains the disconnector region that needs attention;

[0012] S50: collecting a video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals; obtaining the image to be detected and retaining only the significant isolating switch area to be detected of the isolating switch area that needs to be paid attention to;

[0013] S60: Perform similarity matching on the image to be detected in the significant isolating switch area with the template images of the significant isolating switch area of ​​the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, and divide the state of the image to be detected into the normal opening state, the normal closing state, the abnormal opening state or the abnormal closing state.

[0014] Furthermore, the method also includes:

[0015] S70: When the state of the image to be detected is the normal open state or the normal closed state, local similarity matching is performed on the local area of ​​the knife switch connection of the image to be detected and the local area of ​​the knife switch connection of the normal open template or the normal closed template. When the local similarity matching passes, the state of the image to be detected is considered to be the normal open state or the normal closed state, otherwise, the state of the image to be detected is considered to be the abnormal open state or the abnormal closed state.

[0016] Furthermore, the method further comprises:

[0017] S80: When it is detected that the state of the image to be detected changes within two consecutive time intervals and does not conform to the specified state, an alarm message is issued.

[0018] Furthermore, the camera is located at a direction of 45 degrees directly below the corresponding isolating switch, and a low-computing power analysis chip is integrated in the camera.

[0019] Further, the S40 includes:

[0020] S41: corresponding to each group of normal switching template diagram, normal closing template diagram, abnormal opening template diagram and abnormal closing template diagram, a mask diagram is prepared;

[0021] The mask image is a binary image, the pixels of the smallest circumscribed rectangular area where the isolating switch is located in the mask image are 255, and the pixels of other background areas are 0;

[0022] S42: Perform AND operations on the corresponding mask image with the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, to generate template images of the significant disconnecting switch areas of the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, which only retain the disconnecting switch areas that need to be paid attention to.

[0023] Further, the S50 includes:

[0024] S51: collecting a video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals;

[0025] S52: performing an AND operation on the mask image and the image to be detected to generate an image to be detected that only retains the significant isolating switch area of ​​the isolating switch area that needs to be paid attention to.

[0026] Furthermore, the formula for similarity matching is as follows:

[0027]

[0028]

[0029] Among them, x and y represent the two images to be matched for similarity, SSIM(x,y) represents the matching similarity between the two images x and y, and μ x , μ x represents the mean of two images x and y, σ x , σ y represents the variance of two images x and y, σ xy represents the covariance of two images x and y, c 1 and c 2 is a constant, c 1 =(k 1 L) 2 , c 2 =(k 2 L) 2 , L is the dynamic range of pixel values, k 1 =0.01,k 2 =0.03,x i and i Represents the pixel value of the i-th pixel of the two images x and y respectively, i = 1, 2, …, N, N is the total number of pixels of image x or y.

[0030] Further, the S70 includes:

[0031] S71: When the state of the image to be detected is a normal opening state or a normal closing state, a local area of ​​the knife switch connection of the normal opening template image or the normal closing template image is marked as a target area template;

[0032] S72: On the image to be detected, the area corresponding to the target area template is expanded up, down, left, and right by a set size to obtain a search area;

[0033] S73: In the search area, using the size of the target area template as the size of the sliding window, performing sliding window matching on the image to be detected with a set step size, performing local similarity matching on each sliding window and the target area template, and finding the maximum local matching similarity;

[0034] S74: When the maximum local matching similarity is greater than a set threshold, it is considered that the state of the image to be detected is a normal open state or a normal closed state; otherwise, it is considered that the state of the image to be detected is an abnormal open state or an abnormal closed state.

[0035] Furthermore, the formula for local similarity matching is:

[0036]

[0037] Where R is the local matching similarity, T(x′, y′) is the pixel value at (x′, y′) corresponding to the target area template, I(x+x′, y+y′) is the pixel value at (x′, y′) of the sliding window relative to the (x, y) coordinates of the image to be detected, Δx=x+x′, Δy=y+y′.

[0038] Furthermore, the threshold is set by the following process:

[0039] S1: Collect a large number of target videos of normal closing and opening processes;

[0040] S2: Perform local similarity matching on each frame of the target video with the target area templates of normal closing and normal opening to obtain the maximum local matching similarity corresponding to each frame;

[0041] S3: Count the distribution of each maximum local matching similarity, and select the value of the maximum local matching similarity inflection point position corresponding to normal closing and normal opening as the set threshold corresponding to normal closing and normal opening.

[0042] The present invention has the following beneficial effects:

[0043] According to the characteristics of the use scenario, the present invention first selects 4 template images for each scene in the reference video of the normally working disconnector, adds a corresponding mask to each template image, and the mask image only retains the disconnector area that needs to be paid attention to. Each template image and the mask image are ANDed to obtain the template significant disconnector area, and then the newly collected image frame to be detected in each scene is first ANDed with the mask image to obtain the significant disconnector area to be detected, and then the significant disconnector area to be detected is matched with the template significant disconnector area for structural similarity, and for the image frame matched to the normal opening and closing, the local template matching is performed with the template of the opening and closing target area, and finally the normalized coefficient matching similarity is compared with the threshold value to further determine whether the opening and closing are normal. The present invention improves the recognition accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flow chart of an example of a method for detecting the state of a power transformer isolating switch of the present invention;

[0045] Figure 2 The flowchart is another example of the method for detecting the state of the isolating switch of the substation according to the present invention. DETAILED DESCRIPTION

[0046] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0047] The embodiment of the present invention provides a method for detecting the state of a transformer isolation switch. Figure 1 As shown, the method includes:

[0048] S10: A camera capable of capturing the entire isolating switch is provided for each isolating switch that needs to be monitored.

[0049] In one example, a camera is fixedly installed at a position 45 degrees directly below each isolating switch that needs to be monitored, capable of photographing the complete isolating switch, and a low-computing chip is integrated in the camera for algorithm analysis.

[0050] S20: A reference video of each disconnector from normal opening to normal closing and from normal closing to normal opening is collected through a camera.

[0051] When necessary, a large number of target videos of normal closing and opening processes can be collected as much as possible for use in statistically determining the set threshold value during subsequent local similarity matching.

[0052] S30: extracting a normal opening template image, a normal closing template image, an abnormal opening template image, and an abnormal closing template image from the reference video of each disconnector, and storing them in the corresponding camera.

[0053] S40: corresponding to each normal opening template diagram, normal closing template diagram, abnormal opening template diagram and abnormal closing template diagram, respectively obtain a significant disconnector area template image that only retains the disconnector area that needs to be paid attention to.

[0054] In one example, the specific implementation means of this step include:

[0055] S41: corresponding to each group of normal switching template diagram, normal closing template diagram, abnormal opening template diagram and abnormal closing template diagram, a corresponding mask diagram is prepared.

[0056] The mask image is a binary image. The pixels on the mask image corresponding to the minimum bounding rectangular area where the isolation switch is located in the template image are 255, and the pixels in other background areas are 0.

[0057] S42: Perform AND operations on the corresponding mask image with the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, to generate template images of the significant disconnecting switch areas of the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, which only retain the disconnecting switch areas that need to be paid attention to.

[0058] The formula for mask map and operation is as follows:

[0059] S(i,j)=I(i,j)&M(i,j),0≤i <w,0≤j<h

[0060] Where w and h are the width and height of the two images involved in the AND operation, respectively; I is the original image of the normal template image of the opening switch, the normal template image of the closing switch, the abnormal template image of the opening switch and the abnormal template image of the closing switch; M is the mask image; S is the corresponding template image of the significant disconnector area; the connected area with non-zero pixel value in S is the significant target area that needs attention.

[0061] S50: collecting the video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals; obtaining the image to be detected and retaining only the significant isolating switch area to be detected of the isolating switch area that needs to be paid attention to.

[0062] Among them, this step includes:

[0063] S51: collecting the video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals of a certain period of time.

[0064] The interval time can be set to 2s, that is, 50 frames of images are set.

[0065] S52: performing an AND operation on the mask image and the image to be detected to generate an image to be detected that only retains the significant isolating switch area of ​​the isolating switch area that needs to be paid attention to.

[0066] When performing AND operation, the image to be detected is respectively ANDed with the mask images corresponding to the normal template image of the opening switch, the normal template image of the closing switch, the abnormal template image of the opening switch and the abnormal template image of the closing switch, to obtain the corresponding images to be detected of the significant disconnector area.

[0067] S60: Perform similarity matching on the image to be detected in the significant disconnector area with the template images of the significant disconnector area of ​​the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, and divide the state of the image to be detected into the normal opening state, the normal closing state, the abnormal opening state or the abnormal closing state.

[0068] Specifically, the formula for similarity matching is as follows:

[0069]

[0070]

[0071] Among them, x and y represent the two images to be matched for similarity, SSIM(x,y) represents the matching similarity between the two images x and y, and μ x , μ y represents the mean of two images x and y, σ x , σ y represents the variance of two images x and y, σ xy represents the covariance of two images x and y, c 1 and c 2 is a constant, c 1 =(k 1 L) 2 , c 2 =(k 2 L) 2 , L is the dynamic range of pixel values, k 1 =0.01,k 2 =0.03,x i and i Represents the pixel value of the i-th pixel of the two images x and y respectively, i = 1, 2, …, N, N is the total number of pixels of image x or y.

[0072] The above formula requires pixel-by-pixel calculation. In order to speed up the calculation, a Gaussian kernel function (i.e., Gaussian convolution) can be used to calculate the mean, variance, and covariance of the image.

[0073] The corresponding image to be detected in the saliency isolation switch area is similarly matched with the corresponding template image to obtain the corresponding matching similarity. The state corresponding to the value with the largest matching similarity is the state of the image to be detected.

[0074] Preferably, before performing structural similarity matching, the significant isolation switch region may be scaled to accelerate the operation.

[0075] After the above structural similarity matching, the video image frames to be detected will be initially divided into normal opening, normal closing, abnormal opening, and abnormal closing states. If the video image frames to be detected are normal opening or normal closing, the next step will be entered for reconfirmation. Figure 2 shown.

[0076] S70: When the state of the image to be detected is the normal open state or the normal closed state, local similarity matching is performed on the local area of ​​the knife switch connection of the image to be detected and the local area of ​​the knife switch connection of the normal open template or the normal closed template. When the local similarity matching passes, the state of the image to be detected is considered to be the normal open state or the normal closed state, otherwise, the state of the image to be detected is considered to be the abnormal open state or the abnormal closed state.

[0077] A specific implementation of this step is:

[0078] S71: When the state of the image to be detected is the open normal state or the closed normal state, a local area of ​​the knife switch connection of the open normal template image or the closed normal template image is marked as a target area template.

[0079] S72: On the image to be detected, the area corresponding to the target area template is expanded up, down, left, and right by a set size to obtain a search area.

[0080] For example, the left and right and top and bottom expansion sizes of the target area template are [0.1*box w ,0.1*box h ], box w and *box h are the width and height of the target area template respectively.

[0081] S73: In the search area, the size of the target area template is used as the size of the sliding window, and the image to be detected is subjected to sliding window matching with a set step size, and each sliding window is matched with the target area template for local similarity to find the maximum local matching similarity.

[0082] The formula for local similarity matching is as follows:

[0083]

[0084] Where R is the local matching similarity, T(x′, y′) is the pixel value at (x′, y′) corresponding to the target area template, and I(x+x′, y+y′) is the pixel value at (x′, y′) of the sliding window relative to the (x, y) coordinates of the image to be detected. Δx and Δy represent the coordinates of all pixels in the sliding window of the I image, that is, Δx=x+x′, Δy=y+y′.

[0085] S74: When the maximum local matching similarity is greater than a set threshold (eg, 0.9), the state of the image to be detected is considered to be a normal open state or a normal closed state; otherwise, the state of the image to be detected is considered to be an abnormal open state or an abnormal closed state.

[0086] The threshold can be set based on experience or by statistical methods. The process of setting by statistical methods is as follows:

[0087] S1: Collect a large number of target videos of normal closing and opening processes.

[0088] S2: Perform local similarity matching on each frame of the target video and the target area templates with normal closing and opening conditions to obtain the maximum local matching similarity corresponding to each frame.

[0089] S3: Count the distribution of each maximum local matching similarity, and select the value of the maximum local matching similarity inflection point position corresponding to normal closing and normal opening as the set threshold corresponding to normal closing and normal opening.

[0090] The present invention uses the above method to extract a frame of the image to be detected from the video image to be detected at intervals to detect the switch state, and can automatically alarm through the following steps.

[0091] S80: If at a specified time, when it is detected that the state of the image to be detected changes within two consecutive time intervals (ie, multiple consecutive detections) and does not conform to the specified state, an alarm message is issued.

[0092] According to the characteristics of the usage scenarios, the present invention first selects 4 template images for each scene in a reference video of a normally working isolating switch, adds a corresponding mask to each template image, and the mask image only retains the isolating switch area that needs to be paid attention to. Each template image and the mask image are ANDed to obtain the template significant isolating switch area, and then the image frame to be detected newly collected in each scene is first ANDed with the mask image to obtain the significant isolating switch area to be detected, and then the significant isolating switch area to be detected is matched with the template significant isolating switch area for structural similarity, and for the image frames matched to normal opening and closing, local template matching is performed with the opening and closing target area templates, and finally, the normalized coefficient matching similarity is compared with the threshold to further determine whether the opening and closing are normal.

[0093] The present invention has the following beneficial effects:

[0094] The present invention uses a mask to extract the significant isolation switch area for structural similarity matching, which can suppress the influence of the background area, especially for lightning and rain line weather background images. The correct matching of structural similarity plays a certain role in improving the matching.

[0095] Video image frames are extracted at equal intervals for analysis, and structural similarity matching is used to determine the abnormal states of opening and closing at the scene level. For images that are suspected to be opening and closing normally, they are further confirmed through secondary target-level local area template matching to ensure recognition accuracy.

[0096] Big data statistics are used to determine the thresholds for distinguishing normal and abnormal opening and closing, ensuring the accuracy and robustness of identifying normal opening and closing.

[0097] Using the method proposed in this paper, the actual measurement on the Qualcomm msm8909 chip (quad-core Cortex-A7, main frequency 1.1GHz, 1GB DDR3) can achieve an isolation switch status recognition speed of 280ms / frame, which can fully meet the requirements of real-time status monitoring. Moreover, through the secondary confirmation of target area template matching, it can achieve a recognition accuracy of 99.0+% in the usage scenario.

[0098] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting the state of a power transformer isolating switch, characterized in that: include: S10: a camera capable of capturing the entire isolating switch is provided for each isolating switch to be monitored; S20: Using the camera to collect reference videos of each disconnector from normal opening to normal closing and from normal closing to normal opening; S30: extracting a normal opening template image, a normal closing template image, an abnormal opening template image, and an abnormal closing template image from a reference video of each disconnector; S40: corresponding to each normal opening template image, normal closing template image, abnormal opening template image, and abnormal closing template image, respectively obtain a significant disconnector region template image that only retains the disconnector region that needs attention; S50: collecting a video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals; obtaining the image to be detected and retaining only the significant isolating switch area to be detected of the isolating switch area that needs to be paid attention to; S60: Perform similarity matching on the image to be detected in the significant isolating switch area with the template images of the significant isolating switch area of ​​the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, and divide the state of the image to be detected into the normal opening state, the normal closing state, the abnormal opening state or the abnormal closing state.

2. The method for detecting the state of the isolating switch of the substation according to claim 1, characterized in that: The method also includes: S70: When the state of the image to be detected is the normal open state or the normal closed state, local similarity matching is performed on the local area of ​​the knife switch connection of the image to be detected and the local area of ​​the knife switch connection of the normal open template or the normal closed template. When the local similarity matching passes, the state of the image to be detected is considered to be the normal open state or the normal closed state, otherwise, the state of the image to be detected is considered to be the abnormal open state or the abnormal closed state.

3. The method for detecting the state of the isolating switch of the substation according to claim 2, characterized in that: The method further comprises: S80: When it is detected that the state of the image to be detected changes within two consecutive time intervals and does not conform to the specified state, an alarm message is issued.

4. The method for detecting the state of the isolating switch of the substation according to claim 3, characterized in that: The camera is located at a direction of 45 degrees directly below the corresponding isolating switch, and a low-computing power analysis chip is integrated in the camera.

5. The method for detecting the state of the isolating switch of the substation according to claim 3, characterized in that: The S40 includes: S41: corresponding to each group of normal switching template diagram, normal closing template diagram, abnormal opening template diagram and abnormal closing template diagram, a mask diagram is prepared; The mask image is a binary image, the pixels of the smallest circumscribed rectangular area where the isolating switch is located in the mask image are 255, and the pixels of other background areas are 0; S42: Perform AND operations on the corresponding mask image with the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, to generate template images of the significant disconnecting switch areas of the normal opening template image, the normal closing template image, the abnormal opening template image and the abnormal closing template image, respectively, which only retain the disconnecting switch areas that need to be paid attention to.

6. The method for detecting the state of the isolating switch of the substation according to claim 3, characterized in that: The S50 includes: S51: collecting a video image to be detected, and extracting a frame of the image to be detected from the video image to be detected at intervals; S52: performing an AND operation on the mask image and the image to be detected to generate an image to be detected that only retains the significant isolating switch area of ​​the isolating switch area that needs to be paid attention to.

7. The method for detecting the state of the isolating switch of the substation according to claim 3, characterized in that: The formula for similarity matching is as follows: Among them, x and y represent the two images to be matched for similarity, SSIM(x,y) represents the matching similarity between the two images x and y, and μ c , μ y represents the mean of two images x and y, σ x , σ y represents the variance of two images x and y, σ xy Represents the covariance of two images x and y, c1 and c2 are constants, c1=(k1L) 2 , c2=(k2L) 2 , L is the dynamic range of pixel values, k1 = 0.01, k2 = 0.03, x i and i Represents the pixel value of the i-th pixel of the two images x and y respectively, i = 1, 2, …, N, N is the total number of pixels of image x or y.

8. The method for detecting the state of the isolating switch of the substation according to claim 3, characterized in that: The S70 includes: S71: When the state of the image to be detected is a normal opening state or a normal closing state, a local area of ​​the knife switch connection of the normal opening template image or the normal closing template image is marked as a target area template; S72: On the image to be detected, the area corresponding to the target area template is expanded up, down, left, and right by a set size to obtain a search area; S73: In the search area, using the size of the target area template as the size of the sliding window, performing sliding window matching on the image to be detected with a set step size, performing local similarity matching on each sliding window and the target area template, and finding the maximum local matching similarity; S74: When the maximum local matching similarity is greater than a set threshold, the state of the image to be detected is considered to be a normal open state or a normal closed state; otherwise, the state of the image to be detected is considered to be an abnormal open state or an abnormal closed state.

9. The method for detecting the state of the isolating switch of the substation according to claim 8, characterized in that: The formula for local similarity matching is: Among them, R is the local matching similarity, T(x ′ ,y ′ ) is the target area template corresponding to (x ′ ,y ′ ), I(x+x ′ ,y+y ′ ) is the position of the sliding window relative to the (x,y) coordinate of the image to be detected. ′ ,y ′ ) pixel value, Δx=x+x ′ ,Δy=y+y ′ .

10. The method for detecting the state of the isolating switch of the substation according to claim 9, characterized in that: The threshold is set by the following process: S1: Collect a large number of target videos of normal closing and opening processes; S2: Perform local similarity matching on each frame of the target video with the target area templates of normal closing and normal opening to obtain the maximum local matching similarity corresponding to each frame; S3: Count the distribution of each maximum local matching similarity, and select the value of the maximum local matching similarity inflection point position corresponding to normal closing and normal opening as the set threshold corresponding to normal closing and normal opening.

Images