A transformer breather oil cup oil level state detection method and device

Abstract

The present application relates to the field of image processing and machine vision, and specifically discloses a kind of transformer respirator oil cup oil level state detection method and device, comprising the following steps: step one: training respirator oil cup detection model and respirator oil cup scale and oil level segmentation model;Step two: collect the visible light image near the field respirator oil cup;Step three: the position of respirator oil cup is detected using respirator oil cup detection model, and respirator oil cup image is intercepted;Step four: the scale line and oil level line are segmented using segmentation model to respirator oil cup image;Step five: for no oil level line case, color segmentation method is used to distinguish full oil state and no oil state;Step six: for the case of having oil level line, whether oil level is normal is judged according to the obtained oil level line coordinates and scale line coordinates.Can timely and automatically detect the state of multiple transformer respirator oil cups, realize rapid automatic detection, save equipment cost and manpower cost, and ensure the stable operation of transformer.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and machine vision, and particularly relates to a transformer breather cup oil level state detection method and device. BACKGROUND

[0002] At present, the transformer oil in the transformer breather cup plays the role of air isolation. The change of the temperature of the transformer will change the oil level of the breather cup. At the same time, oil seepage and water seepage will also cause abnormal change of the oil level. Abnormal oil level of the transformer breather cup will cause the transformer to be damp, so it is necessary to frequently inspect the transformer breather cup.

[0003] However, in the prior art, the abnormal detection of the oil level of the transformer breather cup is performed by manual inspection. This method has the problems of poor timeliness, untimely abnormal detection and high labor cost. SUMMARY

[0004] The present application aims to provide a transformer breather cup oil level state detection method and device, which solves the technical problems of the prior art, i.e. the abnormal detection of the oil level of the transformer breather cup is performed by manual inspection, which has the problems of poor timeliness, untimely abnormal detection and high labor cost.

[0005] To achieve the above-mentioned purpose, the present application adopts a transformer breather cup oil level state detection method, which comprises the following steps:

[0006] Step 1: training of the breather cup detection model and the breather cup scale and oil level segmentation model;

[0007] Step 2: collecting visible light images near the breather cup on site;

[0008] Step 3: using the breather cup detection model to detect the position of the breather cup and intercepting the breather cup image;

[0009] Step 4: using the segmentation model to segment the scale line and the oil level line for the breather cup image;

[0010] Step 5: for the case without oil level line, using color segmentation method to distinguish the full oil state and the no oil state;

[0011] Step 6: for the case with oil level line, judging whether the oil level is normal according to the obtained oil level line coordinates and scale line coordinates.

[0012] Since the oil cup is generally cylindrical, the oil level line is often a curved line, and the selection of the oil level line of the respirator oil cup semantic segmentation model is very critical, and improper selection may lead to subsequent errors. In the respirator oil cup scale and oil level segmentation model of the application, the oil level line is only selected from the part defined between the starting point and the ending point of the horizontal coordinates of the scale line, and the remaining part is ignored.

[0013] In step one, the following steps are included:

[0014] First, 1000 images containing respirator oil cups and backgrounds are collected;

[0015] The oil cup position is labeled using the labelImg labeling tool to obtain an oil cup detection data set;

[0016] The oil cup detection data set is sent to the target detection network for training to obtain an oil cup detection model;

[0017] The oil cup region of the above collected image is all cut off as an oil cup image;

[0018] The largest scale, the smallest scale and the oil level line of the oil cup image are labeled using labelme, wherein the oil level line is only selected from the part defined between the starting point and the ending point of the horizontal coordinates of the scale line, and the oil cup segmentation data set is obtained after labeling;

[0019] The oil cup segmentation data set is sent to the semantic segmentation network for training to obtain an oil cup scale and oil level segmentation model.

[0020] In step five, the following steps are included:

[0021] First, brightness detection is performed, and the average value of all pixels in the oil cup image is calculated as the overall brightness of the image;

[0022] If the image brightness is lower than the preset value, oil color enhancement is performed;

[0023] The image is converted to HSV color space, and oil color detection is performed in the HSV color space according to the color range of the oil color, and the Hue value range of the oil color can be selected as [13, 38], and the Saturation value range can be selected as [120, 255];

[0024] The oil color in the image is replaced with white, and the rest of the colors are replaced with black to complete the binarization processing of the image;

[0025] After binarization processing, the number of white pixel values in the binarized image is counted;

[0026] If the proportion of white is higher than that of black, it is judged as full oil, otherwise it is judged as no oil.

[0027] The step of oil color enhancement is: judging the oil cup image pixel by pixel, the color of the current pixel is represented by (R, G, B) three components, when the condition (R>1.7*B) && (G>1.3*B) is met, the oil color enhancement processing is carried out, the enhancement processing formula is R=R*10, G=G*10, and the B component remains unchanged, after enhancement, the pixel value is taken, when the pixel value is greater than 255, 255 is taken.

[0028] The application further provides a transformer breather oil cup oil level state detection device for implementing the transformer breather oil cup oil level state detection method in claims 1-5, comprising a network communication unit, a storage unit, an image acquisition and processing unit, a white light light supplement unit, a cloud platform unit and a camera unit, the network communication unit is in wire connection with the image acquisition and processing unit, the storage unit is in wire connection with the image acquisition and processing unit, the white light light supplement unit is in wire connection with the image acquisition and processing unit, the cloud platform unit is in wire connection with the image acquisition and processing unit, and the camera unit is in wire connection with the image acquisition and processing unit.

[0029] The transformer breather oil cup oil level state detection method and device have the advantages that the state of multiple transformer breathers can be automatically detected in time, rapid automatic detection is realized, equipment cost and labor cost are saved, stable operation of the transformer is ensured, and through oil color enhancement, the full oil state and the no oil state can be distinguished under poor light conditions. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0031] Figure 1 is a method step flow chart of a transformer breather oil cup oil level state detection method of the present application.

[0032] Figure 2 is a method step flow chart of step one of a transformer breather oil cup oil level state detection method of the present application.

[0033] Figure 3 is a scale and oil level segmentation model oil level line schematic diagram of the present application.

[0034] Figure 4 is a method step flow chart of step five of a transformer breather oil cup oil level state detection method of the present application.

[0035] Figure 5 is a schematic diagram of the oil color enhancement method of the present application.

[0036] Figure 6 is a method step flow chart of step six of the transformer breather cup oil level state detection method of the present application.

[0037] Figure 7 is a structural schematic diagram of the transformer breather cup oil level state detection device of the present application. DETAILED DESCRIPTION

[0038] Referring to Figures 1 to 6 , the present application provides a transformer breather cup oil level state detection method, comprising the following steps:

[0039] S1: Step one: training of the breather cup detection model and the breather cup scale and oil level segmentation model;

[0040] S2: Step two: collecting visible light images near the on-site breather cup;

[0041] S3: Step three: using the breather cup detection model to detect the position of the breather cup and intercepting the breather cup image;

[0042] S4: Step four: using the segmentation model to segment the scale line and the oil level line for the breather cup image;

[0043] S5: Step five: for the case without oil level line, using color segmentation method to distinguish the full oil state and the no oil state;

[0044] S6: Step six: for the case with oil level line, judging whether the oil level is normal according to the obtained oil level line coordinates and scale line coordinates.

[0045] Further, referring to the accompanying Figure 3 , in the breather cup scale and oil level segmentation model, the oil level line is only selected from the part defined between the starting point and the ending point of the horizontal coordinates of the scale line, and the remaining part is ignored and not selected.

[0046] Further, in step one, the following steps are included:

[0047] S11: first collect 1000 images containing the breather cup and the background;

[0048] S12: use the labelImg annotation tool to annotate the oil cup position to obtain the oil cup detection dataset;

[0049] S13: send the oil cup detection dataset into the target detection network for training to obtain the oil cup detection model;

[0050] S14: All the oil cup regions of the collected images are intercepted as oil cup images;

[0051] S15: The maximum scale, minimum scale and oil level line of the oil cup image are labeled using labelme, wherein the oil level line is only selected from the part defined by the starting point and the ending point of the horizontal coordinate of the scale line, and the oil cup segmentation dataset is obtained after the labeling is completed;

[0052] S16: The oil cup segmentation dataset is sent into a semantic segmentation network for training to obtain an oil cup scale and oil level segmentation model.

[0053] In the embodiment, the target detection model selects a Yolov4 model, and the semantic segmentation model selects a DeeplabV3+ semantic segmentation model.

[0054] Further, in step four, the maximum scale line, the minimum scale line and the oil level line are segmented from the respirator oil cup image using the segmentation model.

[0055] Further, in step five, the following steps are included:

[0056] S51: First, brightness detection is performed, and the average value of all pixels in the oil cup image is calculated as the overall brightness of the image;

[0057] S52: If the brightness of the collected image is lower than the preset value, oil color enhancement is performed;

[0058] S53: The image is converted to an HSV color space, and oil color detection is performed in the HSV color space according to the color range of the oil color. The Hue value range of the oil color can be selected as [13, 38], and the Saturation value range can be selected as [120, 255];

[0059] S54: The oil color in the image is replaced with white, and the remaining colors are replaced with black to complete the binarization processing of the image;

[0060] S55: After the binarization processing, the number of white pixel values in the binarized image is counted;

[0061] S56: If the proportion of white is higher than that of black, it is judged as full oil, otherwise it is judged as no oil.

[0062] Further, the step of oil color enhancement is: the oil cup image is judged pixel by pixel, the color of the current pixel is represented by three components (R, G, B), and when the condition (R>1.7xB) && (G>1.3xB) is met, oil color enhancement processing is performed. The enhancement processing formula is R=R*10, G=G*10, and the B component remains unchanged. After the enhancement, the pixel value is taken, and when the pixel value is greater than 255, 255 is taken.

[0063] Further, in step six, the following steps are included:

[0064] S61: Calculate the average of the Y coordinates of the maximum scale lines;

[0065] S62: Calculate the average of the Y coordinates of the minimum scale lines;

[0066] S63: Calculate the average of the Y coordinates of the oil level lines;

[0067] S64: If the average Y coordinate of the oil level lines is between the average Y coordinates of the maximum scale lines and the minimum scale lines, it indicates that the oil level is normal; if the average Y coordinate of the oil level lines is less than the average Y coordinate of the maximum scale lines, it indicates that the oil level is full; and if the average Y coordinate of the oil level lines is greater than the average Y coordinate of the minimum scale lines, it indicates that the oil level is insufficient.

[0068] Referring to Figure 7 The application also provides a transformer breather oil cup oil level state detection device, which is applied to the transformer breather oil cup oil level state detection method and comprises a network communication unit, a storage unit, an image acquisition and processing unit, a white light light supplement unit, a cloud platform unit and a camera unit.

[0069] In the embodiment, the cloud platform unit is responsible for rotating the camera unit to a specific pre-position, the camera unit is responsible for light sensing and imaging, the image acquisition and processing unit is responsible for regularly acquiring multiple transformer oil cup images, automatically identifying the breather oil cup and detecting state abnormalities, outputting results and alarm information, recording the name, position and alarm image of the alarm target, and pushing the alarm information to a superior information management platform through an HTTP protocol, and the light supplement device provides white light to ensure normal oil color.

[0070] The disclosed is only a preferred embodiment of the application, and of course cannot limit the scope of the application, and those skilled in the art can understand that all or part of the above-mentioned embodiments can be implemented, and equivalent changes made according to the claims of the application still belong to the scope covered by the application.

Claims

1. A method for detecting the oil level status of a transformer breather oil cup, characterized in that, Includes the following steps: Step 1: Train the respirator oil cup detection model and the respirator oil cup scale and oil level segmentation model; Step 2: Acquire visible light images of the vicinity of the respirator oil cup at the scene; Step 3: Use the respirator oil cup detection model to detect the position of the respirator oil cup and capture the image of the respirator oil cup; Step 4: Use a segmentation model to segment the scale lines and oil level lines in the image of the respirator oil cup; Step 5: For cases without an oil level line, use color segmentation to distinguish between full oil and no oil conditions; Step Six: For cases with an oil level line, determine whether the oil level is normal based on the obtained oil level line coordinates and scale line coordinates; Step five includes the following steps: First, brightness is detected, and the average value of all pixels in the oil cup image is calculated as the overall brightness of the image. If the brightness of the acquired image is lower than the preset value, oil color enhancement will be performed; The image is converted to the HSV color space. Oil color is detected in the HSV color space based on the color range of the oil color. The Hue value range of the oil color can be selected as [13, 38], and the Saturation value range can be selected as [120, 255]. Then, replace the oil color in the image with white and the rest of the colors with black to complete the image binarization process; After binarization, the number of white pixel values ​​in the binarized image is counted. If the proportion of white is higher than that of black, it is judged as full oil; otherwise, it is judged as no oil. The steps for enhancing the oil color are as follows: Each pixel of the oil cup image is evaluated. The color of the current pixel is represented by three components (R, G, B). When the conditions (R > 1.7 × B) && (G > 1.3 × B) are met, oil color enhancement processing is performed. The enhancement processing formula is R = R × 10, G = G × 10, and the B component remains unchanged. After enhancement, the pixel value is set. If the pixel value is greater than 255, then 255 is selected.

2. The method for detecting the oil level status of a transformer breather oil cup as described in claim 1, characterized in that, In the respirator oil cup scale and oil level segmentation model, the oil level line only selects the portion between the starting and ending points of the horizontal coordinate of the scale line, and the rest is ignored and not selected.

3. The method for detecting the oil level status of a transformer breather oil cup as described in claim 2, characterized in that, Step one includes the following steps: First, collect 1000 images containing the respirator oil cup and the background; The oil cup locations were labeled using the labelImg annotation tool to obtain the oil cup detection dataset; The oil cup detection dataset is fed into the object detection network for training to obtain the oil cup detection model; The entire oil cup area of ​​the above-collected images is cropped into an oil cup image; The maximum scale, minimum scale, and oil level line of the oil cup image are labeled using labelme. The oil level line is selected only from the part defined by the start and end points of the horizontal coordinate of the scale line. After the labeling is completed, the oil cup segmentation dataset is obtained. The oil cup segmentation dataset is fed into a semantic segmentation network for training to obtain an oil cup scale and oil level segmentation model.

4. A transformer breather oil cup oil level detection device, used to implement the transformer breather oil cup oil level detection method according to any one of claims 1 to 3, characterized in that, It includes a network communication unit, a storage unit, an image acquisition and processing unit, a white light supplementary lighting unit, a pan-tilt unit, and a camera unit. The network communication unit is connected to the image acquisition and processing unit by wires, the storage unit is connected to the image acquisition and processing unit by wires, the white light supplementary lighting unit is connected to the image acquisition and processing unit by wires, the pan-tilt unit is connected to the image acquisition and processing unit by wires, and the camera unit is connected to the image acquisition and processing unit by wires.

Images