Extraction method of effective pressure plate area of ​​protective pressure plate based on multi-threshold and K-means clustering

Through the multi-strategy segmentation and fusion method of multi-threshold and K-mean clustering, combined with morphological processing, the substation protection plate is accurately identified, which solves the problem of low recognition rate in complex backgrounds and improves the operation and maintenance efficiency and safety of substations .

CN114998581BActive Publication Date: 2024-11-22TIANJIN SENRED ELECTRIC POWER TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210467476.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-22
Publication Date
2024-11-22
Estimated Expiration
2040-12-22

AI Technical Summary

Technical Problem

The prior art is difficult to accurately identify the drop-out status of the substation protective pressure plate in a complex background, resulting in a low recognition rate and posing a safety hazard.

Method used

A multi-strategy segmentation fusion method based on multi-threshold and K-mean clustering is adopted, combined with morphological processing, the effective pressure plate area is extracted, and the morphological characteristics of the direction angle and width-length ratio are calculated. The fusion result determines the pressure plate status.

Benefits of technology

It realizes accurate identification in complex backgrounds, improves the efficiency of substation operation and maintenance, reduces safety hazards, and provides important reference for other image segmentation methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114998581B_ABST
    Figure CN114998581B_ABST
Patent Text Reader

Abstract

The effective platen area extraction method of the protective platen based on multi-threshold and K-mean clustering includes the following steps: converting the corrected platen image to HSV color space and Lab color space respectively; performing multi-threshold segmentation on the HSV color space platen image, and performing K-mean clustering segmentation on the Lab color space platen image; firstly performing median filtering on the two segmentation processing result images, and then performing regional fusion to obtain a complete effective platen area; removing noise and small connected areas by performing morphological processing on the effective platen area. The present invention can accurately and quickly identify the platen insertion and withdrawal status under various complex background interferences, effectively improve the efficiency of substation operation and maintenance work, and help reduce safety hazards.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment state recognition, and in particular to a method for extracting an effective pressure plate area of ​​a protection pressure plate based on multi-threshold and K-means clustering. Background Art

[0002] With the rapid development of the economy, the society's demand for electric energy is increasing, and higher requirements are also put forward for the operation and maintenance of substations. In order to ensure the safe and stable operation of substations, regular inspection of the working status of substation secondary equipment is the key task. As an important part of secondary equipment inspection, the protection pressure plate can hardly meet the requirements of intelligence by relying on traditional manual inspection. Moreover, manual inspection often leads to false detection and missed detection due to visual fatigue, memory errors and other reasons, resulting in safety hazards. For this reason, it is imperative to develop a method for identifying the status of substation protection pressure plates based on new technologies and to build a deployment plan for the intelligent operation and maintenance system of substations. At present, there has been some research on the state recognition of intelligent substation protection pressure plates based on image processing, but current research often ignores the segmentation of pressure plate images, especially for pressure plate images with large background interference, which makes it difficult to ensure a high recognition accuracy. Therefore, it is of great theoretical and practical significance to study reliable segmentation and recognition methods for substation functional protection pressure plate images. Summary of the invention

[0003] In order to solve the problem of image distortion caused by shooting angle and low recognition rate caused by complex screen background and shadows caused by lighting, the present invention provides a method for identifying the insertion and withdrawal status of functional protection pressure plates in substations. The method can accurately and quickly identify the insertion and withdrawal status of pressure plates under various complex background interferences, effectively improve the operation and maintenance efficiency of substations, and help reduce safety hazards.

[0004] The technical solution adopted by the present invention is:

[0005] The method for identifying the throw-in and throw-out status of a functional protection pressure plate of a substation comprises the following steps:

[0006] S1: Collect the color image of the screen cabinet including all the pressure plates through the mobile device;

[0007] S2: preprocessing the acquired color image, intercepting and correcting the pressure plate area in the image;

[0008] S3: A multi-strategy segmentation fusion method based on multi-threshold and K-means clustering is used to extract the effective pressure plate area and perform morphological processing;

[0009] S4: Calculate the directional angle and width-to-length ratio morphological features of the effective pressure plate area, identify the pressure plate states corresponding to the two features respectively, and fuse the two pressure plate state results to determine the final pressure plate insertion and retraction state.

[0010] S2 includes the following steps:

[0011] s21: pre-process the acquired color image and mark the four vertices of the quadrilateral containing all the pressing plate areas;

[0012] s22: intercept the quadrilateral area including all the pressure plates and remove the invalid area of ​​the screen cabinet;

[0013] s23: Correct the platen area using perspective transformation.

[0014] S3 includes the following steps:

[0015] s31: Convert the corrected platen image to HSV color space and Lab color space respectively;

[0016] s32: Perform multi-threshold segmentation on the HSV color space platen image and K-means clustering segmentation on the Lab color space platen image;

[0017] s33: First, median filtering is performed on the result images processed by the two segmentation methods, and then regional fusion is performed to obtain a complete effective pressure plate area;

[0018] s34: Remove noise and small connected areas by performing morphological processing on the effective pressure plate area.

[0019] s31 includes the following steps:

[0020] s311: Convert the RGB color space platen image to the HSV color space, and calculate the H component value of the HSV color space platen image according to formula (1). Formula (1) is as follows:

[0021]

[0022] Where: R, G, B are the red, green, and blue channel values ​​respectively, max is the maximum value among R, G, and B, and min is the minimum value among R, G, and B;

[0023] s312: Converts the RGB color space platen image to the Lab color space.

[0024] s32 includes the following steps:

[0025] s321: Use the multi-threshold method to segment the HSV color space pressure plate image, where the minimum value of the H component is 0.01 and the maximum value is 0.1, and the red feature pressure plate area is obtained;

[0026] s322: Use the K-means clustering method to segment the Lab color space pressure plate image and obtain the yellow and green feature pressure plate areas.

[0027] Step s33 comprises the following steps:

[0028] s331: First, salt and pepper noise is added to the images processed by the multi-threshold and K-means clustering segmentation methods, and then the median filter is used to denoise the images;

[0029] s332: The effective color feature pressure plate area after fusion filtering is obtained to obtain a complete effective pressure plate area.

[0030] s34 includes the following steps:

[0031] s341: Use a straight line structure element with a length of 5 and an angle of 90° to perform an opening operation to remove the linear interference area;

[0032] s342: Perform a closing operation with a circular structure element with a radius of 3 to remove burrs and isolated points;

[0033] s343: Dilates the target area with a circular structure element of radius 2, fills holes in the target area, suppresses depth noise, fills depth missing holes, and improves the quality of the target area.

[0034] S4 includes the following steps:

[0035] s41: extracting the morphological features of the direction angle and width-to-length ratio of the effective pressing plate area in the image, and thereby identifying the pressing plate insertion and retraction states corresponding to the two morphological features;

[0036] s42: Combine the two status results to determine the final insertion and retraction status of the pressure plate.

[0037] S41 includes the following steps:

[0038] s411: Substitute the acquired directional angle morphological features into formula (2) for calculation and identify the platen insertion and retraction states. The insertion state is recorded as 1 and the retraction state is recorded as 0. Formula (2) is as follows:

[0039]

[0040] Where: V s In the platen state, V o is the direction angle of the pressure plate;

[0041] s412: Substitute the acquired width-to-length ratio morphological features into formula (3) for calculation and identify the platen insertion and retraction states. The insertion state is recorded as 1 and the retraction state is recorded as 0. Formula (3) is as follows:

[0042]

[0043] Where: V s In the platen state, V a is the width-to-length ratio of the platen area.

[0044] S42 includes the following steps:

[0045] s421: The platen direction angle and aspect ratio morphological feature identification results are integrated, and the final insertion and withdrawal state of the platen is determined using formula (4). The insertion state is recorded as 1 and the withdrawal state is recorded as 0. Formula (4) is as follows:

[0046]

[0047] Where: V s In the platen state, V s1 is the result of angular morphological feature recognition, V s2 is the recognition result of the aspect ratio morphological feature;

[0048] s422: Verify the identified platen insertion and retraction status results and output the platen status identification code.

[0049] The invention provides a method for identifying the throw-in and throw-out states of a functional protection pressure plate of a substation, and has the following beneficial effects:

[0050] 1) The method for identifying the throw-in / out status of a functional protection pressure plate of a substation provided by the present invention effectively solves the distortion effect caused by the shooting angle, and can accurately identify the throw-in / out status of the pressure plate under various complex background interferences, which has important theoretical and practical significance for the intelligent inspection of protection pressure plates of substations.

[0051] 2) In addition, the multi-strategy segmentation and fusion processing method proposed in the present invention is not only applicable to pressure plate image segmentation, but also has important reference value for other types of image segmentation.

[0052] 3) In identifying the state of the pressure plate, the present invention adopts multi-feature identification fusion, which effectively improves the identification reliability compared with the method of identifying the state of the pressure plate with a single feature. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0054] Figure 2 It is a flow chart of the segmentation method of the present invention.

[0055] Figure 3 This is a flow chart of the state identification method of the present invention.

[0056] Figure 4 This is the diagram of the protection plate for the substation cabinet.

[0057] Figure 5 Label the vertices of the quadrilateral that encompasses all the platen regions.

[0058] Figure 6 is the quadrilateral region that contains all the platens.

[0059] Figure 7 This is the correction processing result diagram.

[0060] Figure 8 The platen image is in HSV color space.

[0061] Fig. 9 The platen image is in Lab color space.

[0062] Fig.10 This is the result of multi-threshold segmentation processing.

[0063] Fig.11 This is the result of K-means clustering segmentation processing.

[0064] Fig.12 This is the result diagram of the complete effective platen area.

[0065] Fig.13 This is the result of morphological processing. DETAILED DESCRIPTION

[0066] To facilitate understanding by those skilled in the art, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments of the present invention:

[0067] The embodiment of the present invention provides a method for identifying the throw-in and throw-out status of a functional protection pressure plate in a substation. Figure 1 , Figure 2 , Figure 3 As shown, the method comprises the following steps:

[0068] S1: Use a mobile device to capture a color image of the screen cabinet including all the pressure plates;

[0069] The mobile terminal device uses a mobile inspection robot or a handheld inspection instrument.

[0070] S2: pre-process the obtained image, extract and correct the image pressure plate area, and solve the distortion problem caused by the shooting angle;

[0071] S3: A multi-strategy segmentation fusion method based on multi-threshold and K-means clustering is used to extract the effective pressure plate area and perform morphological processing;

[0072] S4: Calculate the morphological features of the direction angle and width-to-length ratio of the effective pressure plate area, and then identify the pressure plate states corresponding to the two features respectively, and fuse the two state results to determine the final insertion and retraction state of the pressure plate.

[0073] The specific steps of step S2 are as follows:

[0074] s21: for the obtained image Figure 4 Mark the platen area, marking the four vertices of the quadrilateral that contains all the platen areas, such as Figure 5 As shown;

[0075] s22: intercept the quadrilateral region containing all the pressure plates, such as Figure 6 As shown in the figure, the quadrilateral frame contains all the pressure plates, including the valid red, green and yellow pressure plates, invalid pressure plates of other colors and the background; the areas outside the quadrilateral frame, such as the pressure plate labels, the border of the screen cabinet, the ground, etc., are invalid areas;

[0076] s23: When collecting images of the screen cabinet pressure plate, due to the shooting angle, the collected images often have a certain offset. The pressure plate image can be corrected through perspective transformation. The perspective transformation matrix is ​​obtained from the four vertices of the pressure plate area quadrilateral obtained above. Further through the formula Realize perspective transformation and complete rotation and scaling correction of the platen area image. Figure 5 Select the four vertices of the quadrilateral containing all the platen areas as the input source coordinates, obtain the perspective transformation matrix and perform perspective transformation, and get the corrected result as shown below Figure 7 shown.

[0077] The specific steps of step S3 are as follows:

[0078] s31: Convert the corrected platen image to HSV color space and Lab color space respectively; HSV color space is a color space based on the intuitive characteristics of color. Since its components are relatively independent, HSV color space can highlight color features well and has a better effect when specifying color segmentation; Lab color space is a device-independent color system. It has a good masking effect on the background part when specifying color segmentation and can minimize noise interference;

[0079] s32: Perform multi-threshold segmentation on the HSV color space platen image and K-means clustering segmentation on the Lab color space platen image;

[0080] s33: First, median filtering is performed on the result images processed by the two segmentation methods, and then regional fusion is performed to obtain a complete effective pressure plate area;

[0081] s34: Remove noise and small connected areas by performing morphological processing on the effective pressure plate area.

[0082] The specific steps of step s31 are as follows:

[0083] s311: Call the rgb2hsv function in matlab to convert the RGB color space plate image to the HSV color space. The converted image is as follows Figure 8 Therefore, the H component value of the HSV color space plate image is calculated according to formula (1), which is as follows:

[0084]

[0085] Where: R, G, B are the red, green, and blue channel values ​​respectively, max is the maximum value among R, G, and B, and min is the minimum value among R, G, and B;

[0086] s312: Call the rgb2lab function in matlab to convert the RGB color space plate image to the Lab color space. The converted image is as follows Fig. 9 shown.

[0087] The specific steps of step s32 are as follows:

[0088] s321: Use the multi-threshold method to obtain the red feature pressure plate area, that is, the red pressure plate in the cabinet image. Because its color is close to the background border color, it is difficult to obtain an ideal segmentation effect using conventional methods. Considering the weak correlation between the components of the Lab color space, multi-threshold segmentation is used to segment the HSV color space pressure plate image, where the minimum value of the H component is 0.01 and the maximum value is 0.1. The red feature pressure plate area is segmented and binarized. The results are shown below: Fig.10 As shown;

[0089] s322: Use K-means clustering method to segment the yellow and green characteristic pressure plate areas, that is, the yellow and green pressure plates in the screen cabinet image are easily confused with the camel background, resulting in more noise areas after segmentation; considering that the Lab color space has a good masking effect on the background part, the K-means clustering segmentation method is used to segment the Lab color space pressure plate image, obtain the yellow and green characteristic pressure plate areas and perform binarization, and the results are as follows: Fig.11 shown.

[0090] The specific steps of step s33 are as follows:

[0091] s331: After obtaining the binary result images processed by the two segmentation methods of multi-threshold and K-means clustering, in order to prevent the expansion of the noise area during regional fusion, the obtained image is first denoised; by adding salt and pepper noise to the image obtained in step s321 and step s322, and then denoising by using the median filter processing method;

[0092] s332: The effective color feature pressure plate area after fusion filtering is merged so that the red, green, and yellow effective pressure plate areas are merged on the same image, that is, a complete effective pressure plate area is obtained, such as Fig.12 shown.

[0093] The specific steps of step s34 are as follows:

[0094] s341: After the complete effective platen area is obtained through step s33, there are still noise areas and linear interference areas. First, a specific straight line structure element is used to perform an open operation to remove the linear interference area; the straight line structure length is set to 5 and the angle is 90°, that is, a vertical line segment area with a length of 5;

[0095] s342: Considering that the open operation cannot remove isolated point noise, a closed operation is performed through a circular structure element to remove burrs and isolated points; the radius of the circular structure is set to 3, that is, a plane disk-shaped area with a radius of 3;

[0096] s343: To prevent the existence of missing isolated points, the circular structure element with a radius of 2 is finally expanded, and the holes in the target area are filled to suppress the depth image noise, fill the depth missing holes, and improve the quality of the target area; the final effective platen area obtained after morphological processing is as follows Fig.13 shown.

[0097] The specific steps of step S4 are as follows:

[0098] s41: extracting the morphological features of the direction angle and width-to-length ratio of the effective pressing plate area in the image, and thereby identifying the pressing plate insertion and retraction states corresponding to the two morphological features;

[0099] s42: Combine the two status results to determine the final insertion and retraction status of the pressure plate.

[0100] The specific steps of step s41 are as follows:

[0101] s411: Substitute the acquired directional angle morphological features into formula (2) for calculation and identify the platen insertion and retraction states. The insertion state is recorded as 1 and the retraction state is recorded as 0. Formula (2) is as follows:

[0102]

[0103] Where: V s In the platen state, V o is the direction angle of the pressure plate area;

[0104] s412: Substitute the acquired width-to-length ratio morphological features into formula (3) for calculation and identify the platen insertion and retraction states. The insertion state is recorded as 1 and the retraction state is recorded as 0. Formula (3) is as follows:

[0105]

[0106] Where: V s In the platen state, V a is the width-to-length ratio of the platen area.

[0107] The specific steps of step s42 are as follows:

[0108] s421: The platen direction angle and aspect ratio morphological feature identification results are integrated, and the final insertion and withdrawal state of the platen is determined using formula (4). The insertion state is recorded as 1 and the withdrawal state is recorded as 0. Formula (4) is as follows:

[0109]

[0110] Where: V s In the platen state, V s1 is the result of direction angle feature identification, V s2 is the result of the aspect ratio feature identification;

[0111] s422: Verify the identified platen insertion and retraction status results and output the platen status identification code.

[0112] Example:

[0113] The screen cabinet pressure plate image collected in this embodiment is as follows: Figure 4 As shown in the figure, the resolution is 1920×1080 pixels, the horizontal and vertical resolutions are 96dpi, and the size is 832KB. The screen cabinet has 6 rows of pressure plates, each row has 9, and the number of effective pressure plates in the entire screen is 46. It can be observed that there are many invalid areas in the screen cabinet pressure plate image, the pressure plate area only occupies the middle part, and there is also small angle distortion. First, the quadrilateral area containing all the pressure plates in the image is intercepted and corrected by perspective transformation. The processing results are shown as follows: Figure 7 Then, the multi-threshold method and K-means clustering method are used for segmentation respectively, and the segmentation result images obtained by the two methods are fused to obtain the complete segmentation result as shown in Fig.12 Then, morphological processing is used to remove noise interference and dilation is performed. The processing results are shown in Fig.13 shown.

[0114] Next, the effective platen insertion and retraction states are identified. First, the morphological features of the direction angle and width-to-length ratio of the effective platen connection area are extracted, and the platen insertion and retraction states V corresponding to these two morphological features are identified according to formulas (2) and (3). s1 and V s2 Then, according to formula (4), the final state V of the pressure plate is determined. s The relevant data are shown in Table 1. A total of 46 valid pressure plates were identified, including 18 in the input state and 28 in the output state. The identification results are completely consistent with the actual input and output states of the pressure plates in the corresponding areas.

[0115] Table 1 Multi-feature recognition results

[0116]

[0117]

Claims

1. A method for extracting effective pressure plate area of ​​protective pressure plate based on multi-threshold and K-means clustering. Features The following steps are involved: Step 1: Convert the corrected platen image to HSV color space and Lab color space respectively; Step 2: Perform multi-threshold segmentation on the HSV color space pressure plate image, and perform K-means clustering segmentation on the Lab color space pressure plate image; Step 3: First, median filtering is performed on the result images of the two segmentation methods, and then regional fusion is performed to obtain a complete effective pressure plate area; Step 4: Remove noise and small connected areas by morphological processing of the effective pressure plate area; Step 3 includes the following steps: Step 3.1: First, add salt and pepper noise to the images processed by the multi-threshold and K-means clustering methods, and then use the median filter to denoise; Step 3.2: Fuse the effective color feature pressure plate area after filtering to obtain a complete effective pressure plate area; Step 4 includes the following steps: Step 4.1: Use a straight line structure element with a length of 5 and an angle of 90° to perform an open operation to remove the linear interference area; Step 4.2: Perform a closing operation with a circular structure element with a radius of 3 to remove burrs and isolated points; Step 4.3: Dilate the image with a circular structure element of radius 2 and fill holes in the target area to suppress depth noise, fill depth missing holes, and improve the quality of the target area.

2. According to the method for extracting effective pressure plate area of ​​protective pressure plate based on multi-threshold and K-means clustering in claim 1, Features: Step 1 includes the following steps: Step 1.1: Convert the RGB color space platen image to the HSV color space, and calculate the H component value of the HSV color space platen image according to formula (1). Formula (1) is as follows: Where: R, G, B are the red, green, and blue channel values ​​respectively, max is the maximum value among R, G, and B, and min is the minimum value among R, G, and B; Step 1.2: Convert the RGB color space plate image to Lab color space.

3. According to the method for extracting effective pressure plate area of ​​protective pressure plate based on multi-threshold and K-means clustering in claim 1, Features: Step 2 includes the following steps: Step 2.1: Use the multi-threshold method to segment the HSV color space pressure plate image, where the minimum value of the H component is 0.01 and the maximum value is 0.1, and the red feature pressure plate area is obtained; Step 2.2: Use the K-means clustering method to segment the Lab color space pressure plate image and obtain the yellow and green feature pressure plate areas.