Digital electric box intelligent monitoring method and system based on image processing
By adaptively adjusting search ranges and monitoring superpixel block changes, the method enhances pixel segmentation accuracy and real-time anomaly detection in digital electric boxes, addressing inaccuracies in existing superpixel segmentation methods.
Patent Information
- Application Number
- CN202411408038.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2044-10-10
AI Technical Summary
Existing superpixel segmentation methods in digital switchgear monitoring have inaccuracies due to uniform search range sizes, leading to incomplete or inaccurate pixel division and reduced accuracy in fault detection.
Adaptive adjustment of initial search ranges for superpixel segmentation by considering the number of searches and grayscale differences, along with positional relationships, to ensure each pixel is within multiple seed points' range, and monitoring superpixel block changes for real-time anomaly detection.
Improves pixel segmentation accuracy, reduces computational complexity, and enhances real-time anomaly detection in digital electric boxes, ensuring operational safety and reliability.
Smart Images
Figure CN118918534B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for intelligent monitoring of a digital electric box based on image processing. Background Art
[0002] As a key component in modern power systems, digital switchgear status monitoring and fault warning are crucial. Traditional switchgear monitoring methods rely on physical inspections and basic electrical parameter monitoring, but these methods often have problems such as delayed response and inability to capture subtle anomalies in real time. Infrared thermal imaging technology provides an effective solution for switchgear monitoring. By collecting infrared images of the switchgear, it can monitor the temperature distribution inside the switchgear in real time, identify temperature anomalies, overheating or potential fault points in the switchgear, and improve the operating safety and reliability of the switchgear.
[0003] The patent document with the announcement number CN104217425B discloses a method for infrared fault image segmentation of power transmission and transformation equipment based on superpixels, including: image preprocessing, preprocessing the infrared fault image of the power transmission and transformation equipment; superpixel image segmentation, using the difference between the fault area and the background area, the infrared fault image is segmented into a set number of superpixel areas by the superpixel segmentation method; fault area extraction, marking the segmented multiple superpixel areas as ; Use average brightness , Regional Concavity and regional grayscale standard deviation Three features are used to describe each region; combining the three parameters together, we get the function The area with the maximum value is selected as the extracted fault area. ; Fault area information is sorted, the temperature limit corresponding to the input infrared image is set, and the highest temperature is set , the minimum temperature is set to , combined with the color palette in the infrared image, the highest temperature value of the fault area is obtained ; The size of the area obtained by segmentation is the size of the fault area.
[0004] The above method can detect the fault area of the digital switchgear to a certain extent, but in the superpixel segmentation method it adopts, the search range size of various sub-points in each round of iteration is the same, which may cause some pixels to not be within the search range of any seed point in some rounds, or some pixels to be within the search range of only one seed point, resulting in these pixels being unable to divide the corresponding superpixel blocks or the divided superpixel blocks being inaccurate, thereby affecting the accuracy of subsequent digital switchgear fault monitoring based on superpixel block images. Summary of the invention
[0005] In order to solve the problem that the search ranges of various sub-points in superpixel segmentation are the same, resulting in some pixels being unable to divide corresponding superpixel blocks or the divided superpixel blocks being inaccurate, thereby affecting the accuracy of digital electric box fault monitoring, the present invention provides a digital electric box intelligent monitoring method and system based on image processing.
[0006] In the first aspect, the present invention provides a digital electric box intelligent monitoring method based on image processing, which adopts the following technical solutions:
[0007] The digital electric box intelligent monitoring method based on image processing comprises the following steps:
[0008] The infrared image of the digital electric box is segmented into superpixels to obtain a superpixel image; in each round of iteration of superpixel segmentation, the initial search range of various sub-points is updated, including: taking the number of initial search ranges to which each pixel point belongs as the estimated number of searches for each pixel point; taking any seed point as the target seed point, and constructing a fan-shaped area with the minimum distance from other seed points to the target seed point as the radius in each quadrant of the rectangular coordinate system with the target seed point as the origin, as the neighborhood area of the target seed point in each quadrant; taking the difference area between the neighborhood area and the initial search range of the target seed point as the judgment area of the target seed point in each quadrant; determining the search necessity of each pixel point in the judgment area The search necessity is negatively correlated with the expected number of searches, the grayscale difference between the pixel and the target seed point, and positively correlated with the grayscale difference between the pixel and the seed point corresponding to the minimum distance; the judgment area is divided into multiple categories according to the search necessity; the growth necessity of the category is obtained according to the search necessity of the pixel points in the category and the positional relationship between the categories, and the growth necessity is positively correlated with the mean value of the search necessity of the pixel points in the category; the initial search range of the target seed point is updated according to the size of the growth necessity; the abnormal degree of the superpixel block change is determined according to the area change and temperature change of the superpixel block between the superpixel images of the infrared images at adjacent moments, so as to perform digital electric box abnormality monitoring.
[0009] The present invention performs abnormal monitoring of digital electric boxes based on superpixel images of infrared images. Compared with directly performing abnormal monitoring of digital electric boxes based on infrared images, the present invention can more accurately identify and analyze temperature anomalies in different areas of the electric box, and can more accurately determine the boundaries of abnormal areas, thereby improving the accuracy of abnormal area detection. At the same time, superpixel blocks reduce the amount of data that needs to be processed, reduce the computational complexity of identifying abnormal areas of digital electric boxes, and improve the efficiency of abnormal monitoring of digital electric boxes. In the process of superpixel segmentation of infrared images, the present invention corrects the initial search range of various sub-points to ensure that each pixel point in the infrared image can be located within the search range of as many seed points as possible, thereby improving the accuracy of segmentation of each pixel point and the efficiency of superpixel segmentation iteration, so that the obtained Superpixel images are more precise, thereby improving the accuracy and real-time performance of abnormal monitoring of digital electric boxes. The present invention determines the necessity of searching pixel points based on the estimated number of pixel searches and the grayscale difference between pixel points and seed points, obtains the growth necessity of categories based on the necessity of searching pixel points in each category and the positional relationship between categories, and expands the initial search range of seed points based on the growth necessity, ensuring that each pixel point can be searched by as many seed points as possible, thereby improving the accuracy of superpixel segmentation. When performing abnormal monitoring of digital electric boxes, the present invention takes into account the changes in superpixel blocks between adjacent moments, and can promptly discover potential abnormalities in digital electric boxes, thereby improving the real-time performance of abnormal monitoring of digital electric boxes and ensuring the safety and reliability of the operation of the electric box.
[0010] Preferably, the search necessity satisfies the expression:
[0011] ;in, The target seed point is In the judgment area of the quadrant Necessity of searching for pixels; The target seed point is In the judgment area of the quadrant The estimated number of searches for each pixel; is the gray value of the target seed point; The target seed point is In the judgment area of the quadrant The gray value of each pixel; For the The gray value of the seed point closest to the target seed point in the quadrant; is the natural exponential function; is a hyperparameter; Indicates the absolute value symbol; is the maximum value function.
[0012] The present invention determines the necessity of searching a pixel point in combination with the estimated number of searches for the pixel point, and can ensure that the pixel point is searched by as many seed points as possible, thereby improving the accuracy of pixel division; the present invention further corrects the necessity of searching the pixel point according to the grayscale difference between the pixel point and the target seed point, and the grayscale difference between the pixel point and the seed point closest to the target seed point in the corresponding quadrant, and can reduce the number of unnecessary searches, improve the efficiency of superpixel segmentation, and at the same time improve the division accuracy of the superpixel block to which the pixel point belongs.
[0013] Preferably, the step of dividing the judgment area into multiple categories according to the search necessity includes: The search necessity of all pixels in the judgment area of the quadrant is All pixels in the judgment area of the quadrant are clustered by mean shift to obtain multiple categories, among which is the quadrant number, .
[0014] Preferably, the growth necessity satisfies the expression:
[0015] ;in, The target seed point is The quadrant corresponds to Categories Necessity of growth; for The mean of the search necessity of all pixels in ; for As the connection type, the corresponding The mean of the search necessity of all pixels in the isolated category; for The number of isolated categories corresponding to the connected categories; is the difference function: ; for As the connection type, the corresponding The weight of an isolated class.
[0016] Preferably, the weight satisfies the expression: ;in, for As the connection type, the corresponding The distance between the centroid of an isolated class and the target seed point; The target seed point is The quadrant corresponds to The distance between the centroid of each category and the target seed point; From the target seed point to the The distance between the seed points closest to the target seed point in the quadrant; The target seed point is The number of categories that the quadrant corresponds to.
[0017] Preferably, the updating of the initial search range of the target seed point according to the size of the growth necessity includes: for the target seed point in the For any category corresponding to the quadrant, if the category is not an isolated category and the growth necessity of the category is not less than the preset growth threshold, all the pixels in the category are added to the initial search range of the target seed point; if the category is an isolated category, and the growth necessity of the isolated category and all its connected categories is not less than the growth threshold, all the pixels in the isolated category are added to the initial search range of the target seed point, where is the quadrant number, .
[0018] Since the superpixel block corresponding to a seed point is a whole when the superpixel block is divided, it is impossible for the superpixel block corresponding to a seed point to contain multiple discontinuous small blocks. Therefore, in order to divide the pixel points in the category far away from the target seed point into the superpixel block corresponding to the target seed point, the connection category corresponding to the category must also be included in the search range of the target seed point. The present invention expands the initial search range of the seed point according to the positional relationship between the categories and the growth necessity of each category, improves the accuracy of the expansion of the search range of the target seed point, and makes the subsequent superpixel image more accurate.
[0019] Preferably, the method for obtaining the isolated category and the connected category is as follows: for any category, when there is no pixel point in the category that is adjacent to the pixel point within the initial search range of the target seed point, the category is taken as an isolated category; for any isolated category, the pixel point with the smallest distance from the centroid of the isolated category within the initial search range of the target seed point is obtained, and the category to which each pixel point on the line between the pixel point and the centroid of the isolated category belongs is obtained, and each category among these categories except the isolated category is taken as the connected category of the isolated category.
[0020] Preferably, the abnormal degree of change satisfies the expression:
[0021] ;in, is the superpixel image of the infrared image at the current moment The abnormal degree of change of super pixel blocks; is the superpixel image of the infrared image at the current moment The area of a superpixel block; is the superpixel image of the infrared image at the current moment The area of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment; is the superpixel image of the infrared image at the current moment The temperature of each superpixel; is the superpixel image of the infrared image at the current moment The temperature of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment.
[0022] The present invention determines the abnormal degree of change of the superpixel block based on the area change and temperature change of the superpixel block between adjacent moments. According to the abnormal degree of change, the area where the temperature in the electric box is abnormally increasing or spreading can be discovered in time, and the potential abnormalities in the digital electric box can be discovered in time, thereby improving the real-time performance of abnormal monitoring of the digital electric box.
[0023] Preferably, the method for performing abnormal monitoring of the digital electric box includes: in response to the presence of a superpixel block in the superpixel image of the infrared image at the current moment having a temperature exceeding a preset temperature threshold, or a superpixel block having a degree of abnormal change greater than a preset change threshold, indicating that there is an abnormality in the digital electric box at the current moment, and treating the superpixel block whose temperature exceeds the preset temperature threshold and the superpixel block whose degree of abnormal change is greater than the preset change threshold as abnormal areas.
[0024] The present invention uses the temperature and abnormal degree of change of the super-pixel block as the judgment standard, which can not only detect the overheating anomaly in the digital electrical box, but also detect the potential anomaly of temperature growth or diffusion, which can improve the detection rate of potential faults and improve the operating safety and reliability of the electrical box.
[0025] In a second aspect, the present invention provides a digital electric box intelligent monitoring system based on image processing, which adopts the following technical solutions:
[0026] The digital electric box intelligent monitoring system based on image processing comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the digital electric box intelligent monitoring method based on image processing is implemented.
[0027] By adopting the above technical solution, the above-mentioned digital electric box intelligent monitoring method based on image processing is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is made according to the memory and the processor for easy use.
[0028] The present invention has the following technical effects:
[0029] 1. The present invention can more accurately identify and analyze temperature anomalies in different areas of the electric box, improve the detection accuracy of abnormal areas, reduce calculation complexity, and improve the efficiency of abnormal monitoring of digital electric boxes;
[0030] 2. The present invention improves the accuracy of pixel segmentation and the efficiency of super-pixel segmentation iteration, and the obtained super-pixel image is more accurate, which improves the accuracy and real-time performance of abnormal monitoring of digital electric boxes;
[0031] 3. The present invention can timely discover potential anomalies in the digital electric box, improve the real-time performance of anomaly monitoring of the digital electric box, and ensure the operational safety and reliability of the electric box. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.
[0033] Figure 1 is a method flow chart of a digital electric box intelligent monitoring method based on image processing in an embodiment of the present invention;
[0034] Figure 2 It is an infrared image of a digital electric box;
[0035] Figure 3 It is a schematic diagram of the neighborhood area of the target seed point in the first quadrant;
[0036] Figure 4 It is a schematic diagram of the judgment area of the target seed point in the first quadrant;
[0037] Figure 5 A super-pixel image is obtained by performing super-pixel segmentation on an infrared image using the method of the present invention;
[0038] Figure 6 It is a super-pixel image obtained by performing super-pixel segmentation on the infrared image using the traditional super-pixel segmentation algorithm. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0040] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.
[0041] The embodiment of the present invention discloses a digital electric box intelligent monitoring method based on image processing, referring to Figure 1 , comprising steps S1 to S4:
[0042] S1: Collect infrared images of digital electric box.
[0043] An infrared thermal imager is installed in the digital electrical box, facing the electrical components in the digital electrical box, and the infrared thermal imager is used to collect infrared images of the digital electrical box in real time. Figure 2 It is an infrared image of a digital electric box. In the embodiment of the present invention, the acquisition frequency is once per minute. In other embodiments, the implementer can set the acquisition frequency of the infrared image according to the actual implementation situation.
[0044] At this point, the acquisition of infrared images of the digital electrical box has been achieved.
[0045] S2: Perform superpixel segmentation on the infrared image of the digital electric box to obtain a superpixel image. In each round of iteration of superpixel segmentation, the search necessity of the pixel points in the judgment area is determined according to the expected search times of the pixel points and the grayscale difference between the pixel points and the seed points. The judgment area is divided into multiple categories. The growth necessity of the category is obtained according to the search necessity of the pixel points in the category and the positional relationship between the categories. The initial search range of the seed point is updated according to the growth necessity.
[0046] It should be noted that in the traditional superpixel segmentation algorithm, the search range size of various sub-points in each round of iteration is the same, which may cause some pixels to be not located in the search range of any seed point, or some pixels to be located in the search range of only one seed point in some rounds, which will cause these pixels to be unable to divide the corresponding superpixel blocks or the divided superpixel blocks to be inaccurate. Therefore, in the process of superpixel segmentation of infrared images, the present invention corrects the search range of various sub-points to ensure that each pixel in the infrared image can be located in the search range of as many seed points as possible, thereby improving the accuracy of segmentation of each pixel.
[0047] Specifically, the infrared image is segmented into superpixels to obtain a superpixel image of the infrared image. In the present invention, the number of seed points during superpixel segmentation is set to , the initial search range of each seed point is ,in , is the length of the infrared image, is the width of the infrared image, Indicates the rounding symbol. The number of seed points can be set by the implementer according to the actual implementation situation, and there is no specific limit. For example The superpixel segmentation algorithm adopted in the present invention is a simple linear iterative clustering algorithm (SLIC).
[0048] In each round of iteration of superpixel segmentation of infrared images using the SLIC algorithm, all seed points of this round of iteration are obtained according to the superpixel blocks obtained in the previous round of iteration. The initial search range size of each seed point is For any pixel in the infrared image, determine how many seed points the pixel is within the initial search range, and use the number of initial search ranges to which the pixel belongs as the estimated number of searches for the pixel.
[0049] Take any seed point as the target seed point, take the target seed point as the origin, the horizontal right direction as the horizontal axis direction, and the horizontal upward direction as the vertical axis direction to construct a rectangular coordinate system.
[0050] For the rectangular coordinate system Quadrant ( ), the first The seed point closest to the target seed point in the quadrant is taken as the first seed point of the target seed point. Adjacent seed points. With the target seed point as the center, The distance between adjacent seed points is the radius. Draw a fan-shaped area in the quadrant as the target seed point in the The neighborhood area of the quadrant, the two sides of the fan-shaped area coincide with the horizontal axis and the vertical axis respectively. Figure 3 is a schematic diagram of the neighborhood area of the target seed point in the first quadrant. Figure 3 The black dots represent seed points.
[0051] Get the target seed point in The neighboring area of the quadrant, but not all pixels in the search range of the target seed point, constitute the target seed point in the first quadrant. The judgment area of the quadrant. Figure 4 This is a schematic diagram of the judgment area of the target seed point in the first quadrant. Figure 4The middle sector area is the neighborhood area of the target seed point in the first quadrant, the area formed by the dotted line is the initial search range of the target seed point, the gray area is the judgment area of the target seed point in the first quadrant, and the black dots represent the seed points.
[0052] When When there is no judgment area in the quadrant, the initial search range for the target seed point is The parts in the quadrant are not updated.
[0053] When When there is a judgment area in the quadrant, for the target seed point in For any pixel point in the judgment area of the quadrant, the necessity of searching for the pixel point is obtained based on the estimated search times of the pixel point and the difference between the pixel point and the target seed point:
[0054]
[0055] in, The target seed point is In the judgment area of the quadrant The necessity of searching for pixels is used to reflect the The necessity of adding pixels to the initial search area of the target seed point; The target seed point is In the judgment area of the quadrant The estimated number of searches for each pixel; is the gray value of the target seed point; The target seed point is In the judgment area of the quadrant The gray value of each pixel; For the The gray value of the seed point closest to the target seed point in the quadrant, that is, the gray value of the target seed point Gray values of adjacent seed points; is the natural exponential function; is a hyperparameter used to prevent The value of is large, resulting in Always close to 0, in this embodiment The empirical value of is 4. In other embodiments, the implementer can set it according to the actual implementation situation; Indicates the absolute value symbol; is the maximum value function, Used to The value range is limited to [-1,1].
[0056] When The fewer the estimated number of searches for each pixel, the more pixel points cannot be searched by more seed points, which may cause the The pixel point is divided into the super pixel block corresponding to the seed point with a large difference, which affects the effect of super pixel segmentation. When the estimated number of searches for a pixel is 0, The pixel cannot be divided into any super pixel block. Therefore, in order to ensure pixel points can be accurately divided into corresponding super pixel blocks. In the present invention, when the The fewer the number of predicted searches for each pixel, the The greater the pixel search necessity.
[0057] is the target seed point and its In the judgment area of the quadrant The difference between the gray values of pixels The target seed point The adjacent seed points and the target seed point are In the judgment area of the quadrant The difference between the gray values of pixels is due to the The pixel point is located between the target seed point and the target seed point. Between adjacent seed points, when the difference Less than difference At that time, The closer the features of the pixel point are to the target seed point, the The more pixels should be divided into the superpixel block corresponding to the target seed point, the larger the The necessity of searching for pixels makes the subsequent superpixel block division pixel points can be searched by the target seed point, so the present invention As The index of Correction due to differences Less than difference , so that In between, thus making is a decimal between (0,1), then the exponent Can achieve The increasing effect of The smaller the difference When, yes The greater the increase of The greater the necessity of searching for pixels, the greater the difference Greater than the difference At that time, The pixel point and the target seed point The closer the features of adjacent seed points are, the closer the first The more pixels should be classified into the target seed point, the In the superpixel block corresponding to the adjacent seed point, the The necessity of searching for pixels makes the subsequent superpixel block division The pixel points do not need to be searched by the target seed point, so the present invention As The index of Correction due to differences Greater than the difference , so that In between, thus making is a number greater than 1, then the exponent Can achieve The reduction effect of Bigger than the difference When, yes The greater the reduction of The less necessary the search is, the smaller the difference is. Equal to difference hour, is 1, then No amendment, The necessity of searching for pixels depends on .
[0058] According to the target seed point The search necessity of all pixels in the judgment area of the quadrant is All pixels in the judgment area of the quadrant are clustered by mean shift to obtain multiple categories. For any category, when there is no pixel in the category adjacent to the pixel in the initial search range of the target seed point, the category is regarded as an isolated category.
[0059] For any isolated category, obtain the pixel point with the smallest distance from the centroid of the isolated category within the initial search range of the target seed point, obtain the category to which each pixel point on the line between the pixel point and the centroid of the isolated category belongs, and use each category among these categories except the isolated category as the connection category of the isolated category.
[0060] The growth necessity of each category is obtained based on the search necessity of pixels in each category and the positional relationship between categories:
[0061]
[0062] Among them, the target seed point is The quadrant corresponds to Category express, for Necessity of growth; for The mean value of the search necessity of all pixels in is denoted as Necessity of search; for As the connection type, the corresponding The mean value of the search necessity of all pixels in an isolated category is denoted as As the connection type, the corresponding the necessity of searching for isolated categories; for The number of isolated categories that correspond to connected categories.
[0063] when Necessity of search The bigger the The more pixels in the target seed point should be included in the search range. The greater the need for growth.
[0064] for and The difference function between: ,when As the connection type, the corresponding Necessity of searching for isolated categories Greater than Necessity of search hour, ,otherwise, .when As the connection type, the corresponding Necessity of searching for isolated categories The larger the The pixels in the isolated category should be included in the search range of the target seed point, so that the subsequent The pixels in the isolated category can be divided into the superpixel block corresponding to the target seed point. However, when dividing the superpixel block, the superpixel block corresponding to a seed point is a whole, and it is impossible for a superpixel block corresponding to a seed point to contain multiple discontinuous small blocks. Therefore, the first To divide the pixels in the first isolated category into the superpixel block corresponding to the target seed point, it is necessary to The connected categories of the isolated categories are also included in the search range of the target seed point. As the search necessity of all isolated categories corresponding to the connected categories, The search necessity is modified when there is The search necessity of the corresponding isolated category is greater than When searching for necessity, according to The search necessity of the corresponding isolated category is The difference function of the search necessity is The search necessity is expanded so that The need for growth increases, ensuring that When the corresponding isolated category is to be included in the search range of the target seed point, It can also be included in the search range of the target seed point.
[0065] for As the connection type, the corresponding Weights for isolated classes: ,in, for As the connection type, the corresponding The distance between the centroid of an isolated class and the target seed point; The target seed point is The quadrant corresponds to The distance between the centroid of each category and the target seed point; From the target seed point to the The distance between adjacent seed points is used to , Perform normalization; The target seed point is The number of categories corresponding to the quadrant. When dividing superpixel blocks, not only the grayscale difference between each pixel point and the seed point within the seed point search range must be considered, but also the distance between each pixel point and the seed point. The longer the distance, the smaller the possibility of each pixel point being divided into the superpixel block corresponding to the seed point. Therefore, the present invention sets a weight for each isolated category according to the distance between the centroid of each isolated category and the target seed point. When the distance between the centroid of the isolated category and the target seed point is closer, the weight is The larger, according to The difference function between the corresponding isolated categories The greater the necessity of the search to be corrected, the greater the weight The smaller the time, the smaller the The difference function between the corresponding isolated categories The smaller the need for correction of the search necessity, the less necessary it is.
[0066] For the target seed point For any category corresponding to the quadrant, if the category is not an isolated category and the growth necessity of the category is not less than the preset growth threshold, all pixels in the category are added to the initial search range of the target seed point; if the category is an isolated category, and the growth necessity of the isolated category and all its connected categories is not less than the growth threshold, all pixels in the isolated category are added to the initial search range of the target seed point to update the initial search range. The growth threshold is set by the implementer according to the actual implementation situation, for example, 0.7.
[0067] Similarly, according to the growth necessity of each category corresponding to the target seed point in each quadrant, the initial search range is updated multiple times, and the updated initial search range finally obtained is used as the target search range of the target seed point.
[0068] Similarly, the target search range of each seed point of this round of iteration is obtained. The superpixel block of this round is obtained according to the target search range of each seed point. It should be noted that obtaining the superpixel block of this round according to the target search range of each seed point is a well-known technology in the superpixel segmentation algorithm and will not be described in detail here.
[0069] It should be noted that in each round of iteration of superpixel segmentation of infrared images, the present invention adaptively expands the search range of various sub-points to avoid the situation where some pixels cannot be searched or are located within the search range of fewer seed points, resulting in inaccurate division of these pixels. The present invention speeds up the iteration efficiency of superpixel segmentation and increases the accuracy of superpixel segmentation. For example, Figure 5 In order to utilize the method of the present invention Figure 2 The super-pixel image obtained by super-pixel segmentation of the infrared image shown, Figure 6 In order to use the traditional superpixel segmentation algorithm SLIC Figure 2 The super-pixel image is obtained by super-pixel segmentation of the infrared image shown. It can be seen that Figure 6 There are large differences in the grayscale values of pixels in some superpixel blocks, and the segmentation effect is poor. Figure 5 Relative to Figure 6 The segmentation effect is better.
[0070] At this point, the superpixel segmentation of the infrared image is achieved and the superpixel image is obtained.
[0071] S3: Determine the abnormal degree of change of the super-pixel block according to the area change and temperature change of the super-pixel block between the super-pixel images of the infrared images at adjacent moments.
[0072] Specifically, for the super-pixel image of the infrared image at the current moment, the average value of the temperature corresponding to each pixel point in each super-pixel block is used as the temperature of each super-pixel block.
[0073] Take any superpixel block in the superpixel image of the infrared image at the current moment as the target superpixel block, calculate the distance between the centroid of the target superpixel block and the centroid of each superpixel block in the superpixel image at the previous moment, and take the superpixel block in the superpixel image at the previous moment with the smallest distance to the centroid of the target superpixel block as the superpixel block corresponding to the target superpixel block in the superpixel image at the previous moment.
[0074] Similarly, the corresponding superpixel blocks in the superpixel image at the previous moment for each superpixel block in the superpixel image of the infrared image at the current moment are obtained.
[0075] According to the difference between each superpixel block in the superpixel image of the infrared image at the current moment and the corresponding superpixel block in the superpixel image at the previous moment, the abnormal degree of change of each superpixel block in the superpixel image of the infrared image at the current moment is obtained:
[0076]
[0077] in, is the superpixel image of the infrared image at the current moment The abnormal degree of change of super pixel blocks; is the superpixel image of the infrared image at the current moment The area of the super pixel block, that is, the area of the super pixel block of the infrared image at the current moment The number of pixels contained in a superpixel block; is the superpixel image of the infrared image at the current moment The area of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment; is the superpixel image of the infrared image at the current moment The temperature of each superpixel; is the superpixel image of the infrared image at the current moment The temperature of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment; is an S-type function, expressed as , used for Normalize.
[0078] when , and the larger it is, the The more the temperature of the electrical components in the electrical box corresponding to each super pixel block rises, the The greater the abnormality of the superpixel block, the greater the abnormality and the greater the abnormality. is the superpixel image of the infrared image at the current moment The area growth rate of superpixel blocks is , explain The area of the superpixel block does not increase. The abnormal degree of change of the superpixel block depends on the The temperature rise of the electrical components in the electrical box corresponding to each super pixel block depends on ;when , explain The area of the superpixel block increases. The temperature of the electrical components in the electrical box corresponding to the superpixel block remains unchanged or rises, indicating that The temperature of the electrical components in the electrical box corresponding to each super pixel block diffuses. The larger the abnormality of the change of the super pixel block, the larger the abnormality of the change of the super pixel block. right Correction is made. When the area growth rate is larger, The smaller, the The greater the enhancement, the The greater the abnormality of the change of the superpixel block.
[0079] At this point, the abnormal degree of change of each super-pixel block in the super-pixel image of the infrared image at the current moment is obtained.
[0080] S4: The temperature of the super-pixel block in the super-pixel image of the infrared image at the current moment and the degree of abnormal change are used to monitor the abnormality of the digital electric box.
[0081] In response to the superpixel block in the superpixel image of the infrared image at the current moment having a temperature exceeding a preset temperature threshold, or a superpixel block having a change abnormality greater than a preset change threshold, the digital electric box at the current moment has an abnormality, and the superpixel block whose temperature exceeds the preset temperature threshold and the superpixel block whose change abnormality is greater than the preset change threshold are regarded as abnormal areas of the digital electric box, and the relevant staff is notified to check the electrical components corresponding to the abnormal area in the digital electric box. The temperature threshold and the change threshold are set by the implementer according to the actual implementation situation, for example, the temperature threshold is 50°C and the change threshold is 0.73.
[0082] At this point, intelligent monitoring of digital electrical boxes has been realized.
[0083] An embodiment of the present invention also discloses a digital electric box intelligent monitoring system based on image processing, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a digital electric box intelligent monitoring method based on image processing according to the present invention is implemented.
[0084] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.
[0085] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
[0086] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. Intelligent monitoring method of digital electric box based on image processing, It is characterized in that include: Perform super-pixel segmentation on the infrared image of the digital electric box to obtain a super-pixel image; During each iteration of superpixel segmentation, the initial search ranges of various sub-points are updated, including: The number of initial search ranges to which each pixel belongs is taken as the estimated number of searches for each pixel; any seed point is taken as the target seed point, and in each quadrant of the rectangular coordinate system with the target seed point as the origin, a fan-shaped area is constructed with the minimum distance from other seed points to the target seed point as the radius, as the neighborhood area of the target seed point in each quadrant; the difference area between the neighborhood area and the initial search range of the target seed point is taken as the judgment area of the target seed point in each quadrant; the search necessity of each pixel point in the judgment area is determined to satisfy the expression: ； in, The target seed point is In the judgment area of the quadrant Necessity of searching for pixels; The target seed point is In the judgment area of the quadrant The estimated number of searches for each pixel; is the gray value of the target seed point; The target seed point is In the judgment area of the quadrant The gray value of each pixel; For the The gray value of the seed point closest to the target seed point in the quadrant; is the natural exponential function; is a hyperparameter; Indicates the absolute value symbol; is the maximum value function; The judgment area is divided into multiple categories according to the search necessity; the growth necessity of the category is obtained according to the search necessity of the pixel points in the category and the positional relationship between the categories, satisfying the expression: ； in, The target seed point is The quadrant corresponds to Categories Necessity of growth; for The mean of the search necessity of all pixels in ; for As the connection type, the corresponding The mean of the search necessity of all pixels in the isolated category; for The number of isolated categories corresponding to the connected categories; is the difference function: ; for As the connection type, the corresponding The weight of the isolated category; The weights satisfy the expression: ； in, for As the connection type, the corresponding The distance between the centroid of an isolated class and the target seed point; The target seed point is The quadrant corresponds to The distance between the centroid of each category and the target seed point; From the target seed point to the The distance between the seed points closest to the target seed point in the quadrant; The target seed point is The number of categories corresponding to the quadrant; Update the initial search range of the target seed point according to the size of the growth necessity; The abnormal degree of the super-pixel block change is determined according to the area change and temperature change of the super-pixel block between the super-pixel images of the infrared images at adjacent moments, so as to perform abnormal monitoring of the digital electric box.
2. According to the image processing-based digital electric box intelligent monitoring method of claim 1, It is characterized in that The judgment area is divided into multiple categories according to the search necessity, including: According to the target seed point The search necessity of all pixels in the judgment area of the quadrant is All pixels in the judgment area of the quadrant are clustered by mean shift to obtain multiple categories, among which is the quadrant number, .
3. According to the image processing-based digital electric box intelligent monitoring method of claim 1, It is characterized in that The updating of the initial search range of the target seed point according to the size of the growth necessity includes: For the target seed point For any category corresponding to the quadrant, if the category is not an isolated category and the growth necessity of the category is not less than the preset growth threshold, all the pixels in the category are added to the initial search range of the target seed point; if the category is an isolated category, and the growth necessity of the isolated category and all its connected categories is not less than the growth threshold, all the pixels in the isolated category are added to the initial search range of the target seed point, where is the quadrant number, .
4. According to the digital electric box intelligent monitoring method based on image processing according to claim 3, It is characterized in that The method for obtaining the isolated category and the connected category is: For any category, when there is no pixel in the category adjacent to the pixel in the initial search range of the target seed point, the category is regarded as an isolated category; For any isolated category, obtain the pixel point with the smallest distance from the centroid of the isolated category within the initial search range of the target seed point, obtain the category to which each pixel point on the line between the pixel point and the centroid of the isolated category belongs, and use each category among these categories except the isolated category as the connection category of the isolated category.
5. According to the digital electric box intelligent monitoring method based on image processing according to claim 1, It is characterized in that The abnormal degree of the change satisfies the expression: ； in, is the superpixel image of the infrared image at the current moment The abnormal degree of change of super pixel blocks; is the superpixel image of the infrared image at the current moment The area of a superpixel block; is the superpixel image of the infrared image at the current moment The area of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment; is the superpixel image of the infrared image at the current moment The temperature of each superpixel; is the superpixel image of the infrared image at the current moment The temperature of the superpixel block corresponding to the superpixel block in the superpixel image at the previous moment.
6. According to the digital electric box intelligent monitoring method based on image processing according to claim 1, It is characterized in that The method for performing abnormal monitoring of the digital electric box includes: In response to the fact that the temperature of a superpixel block in the superpixel image of the infrared image at the current moment exceeds a preset temperature threshold, or the abnormal degree of change of a superpixel block is greater than a preset change threshold, there is an abnormality in the digital switchgear at the current moment, and the superpixel blocks whose temperature exceeds the preset temperature threshold and the superpixel blocks whose abnormal degree of change is greater than the preset change threshold are regarded as abnormal areas.
7. Digital electric box intelligent monitoring system based on image processing, It is characterized in that include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the digital electric box intelligent monitoring method based on image processing according to any one of claims 1 to 6 is implemented.