Early warning method and system based on power transmission line image data analysis

Abstract

The invention discloses an early warning method and system based on power transmission line image data analysis, and relates to the technical field of image processing, and the method comprises the steps: marking an initial foreground region and an initial background region on a power transmission line image; and analyzing the initial foreground region and the initial background region to form initial foreground-background information, and carrying out brightness uniform processing on the image by combining the brightness distribution characteristics of the power transmission line image and the foreground-background information, thereby improving the brightness quality of the image and providing a reliable basis for subsequent denoising processing and foreground and background separation. The identification precision of the first classifier for the target object is improved by setting the confidence coefficient in the first classifier, and the image features and the monitoring parameters in the second classifier are fused by setting the weights of the image features and the monitoring parameters, so that the early warning accuracy and adaptability of the power transmission line are improved, and the safe and normal operation of the power transmission line is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an early warning method and system based on power transmission line image data analysis. Background Art

[0002] Transmission line danger warnings are mainly due to the limitations of traditional inspection methods and the development of modern information technology. Traditional manual inspection methods are not only time-consuming and labor-intensive, but also difficult to achieve all-weather and all-round coverage, especially in bad weather or remote areas, where its efficiency and safety are greatly reduced. With the accelerated advancement of smart grid construction, it has become an urgent need to use modern information technology, such as high-definition video surveillance, image recognition, and big data analysis, to achieve remote, real-time, and intelligent monitoring of transmission lines.

[0003] In the prior art, the safety of transmission lines is monitored through manual inspections, or the transmission lines are monitored only based on the status parameters of the transmission lines (current, voltage, temperature, etc.), resulting in poor accuracy and adaptability of early warning of transmission lines, and failure to ensure the safe operation of transmission lines in a timely manner.

[0004] Therefore, how to improve the accuracy and adaptability of early warning of transmission lines is a technical problem that needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of poor accuracy and adaptability of early warning of power transmission lines in the prior art, and propose an early warning method based on power transmission line image data analysis, which includes: Collecting a transmission line image, checking a brightness histogram of the transmission line image, and counting brightness distribution characteristics of the transmission line image, and marking an initial foreground area and an initial background area on the transmission line image; Analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform image brightness uniformity processing in combination with the brightness distribution characteristics of the transmission line image and the foreground-background information; Perform filtering and denoising on the initial foreground area and the initial background area through the initial foreground-background information, and separate the foreground area and the background area of ​​the processed transmission line image; The target object is identified in the foreground area by the first classifier, the monitoring parameters of the transmission line are obtained, and the monitoring parameters are matched with the transmission line image, and the monitoring parameters and the transmission line image are input into the second classifier, thereby realizing early warning of the transmission line.

[0006] In some embodiments of the present application, the brightness distribution characteristics of the power transmission line image are counted, and the initial foreground area and the initial background area on the power transmission line image are marked, including: The brightness value range, peak value and skewness are obtained in the brightness histogram of the transmission line image. The brightness value range covers all pixel values ​​from the darkest to the brightest in the image. The peak value is the brightness value with a frequency exceeding the threshold in the image. The skewness is a statistic that describes the asymmetry of the brightness histogram distribution. The transmission line image is divided into a foreground part and a background part by a preset image segmentation algorithm, and the foreground part and the background part are marked and recorded as an initial foreground area and an initial background area.

[0007] In some embodiments of the present application, the initial foreground area and the initial background area are analyzed to form initial foreground-background information, including: Determine the brightness value range of the initial foreground area and the initial background area, calculate the brightness value range deviation between the initial foreground area and the initial background area, obtain the brightness deviation value, and use the brightness value range and the brightness deviation value of the initial foreground area and the initial background area as the initial foreground-background information.

[0008] In some embodiments of the present application, the brightness distribution characteristics of the power transmission line image and the foreground-background information are combined to perform uniform brightness processing on the image, including: Determine a median value in the brightness value range, divide the peak into a left peak value and a right peak value according to the peak values ​​on the left and right sides of the median value, calculate the distances between the median value and the multiple left peak values ​​and the multiple right peak values, respectively, select the maximum distances on both sides, and determine the first brightness range with the left peak value and the right peak value; The endpoint value on the left side or the right side of the first brightness range is adjusted by the skewness, so as to obtain a second brightness range; The union range of the initial foreground area's brightness value range and the second brightness range is taken as the third brightness range, and the two endpoints of the third brightness range are adjusted based on the brightness deviation value to obtain the target brightness range. A mapping function is set according to the target brightness range, and part of the pixels in the power transmission line image are mapped through the mapping function to achieve uniform brightness processing of the image.

[0009] In some embodiments of the present application, filtering and denoising are performed on the initial foreground area and the initial background area using the initial foreground-background information, including: Analyze the noise points in the brightness histogram in the initial foreground area and the initial background area, and calculate the signal-to-noise ratio of the initial foreground area and the signal-to-noise ratio of the initial background area; The first window size range and the second window size range are determined according to the signal-to-noise ratio and brightness value range of the initial foreground area and the initial background area, respectively, and the intersection range of the first window size range and the second window size range is used as the window size of the filter of the initial foreground area and the initial background area, respectively, so as to perform filtering and denoising processing.

[0010] In some embodiments of the present application, before identifying the target object in the foreground area by the first classifier, the method further includes training the first classifier and the second classifier, and the training process includes: Collect foreground area images containing target objects, mark basic information of target objects in the foreground area images, extract image features of each object in the foreground area images, standardize the image features, divide the image features into single-pixel features and multi-pixel features, and define confidence by combining single-pixel features and multi-pixel features; ; in, is the confidence of the i-th object in the foreground image, , are the number of single-pixel features and multi-pixel features, , Respectively The confidence coefficients of the single-pixel features and the multi-pixel features, , Respectively The size of the single pixel feature and the The size of multi-pixel features, express and The maximum value in is a preset constant; Marking the confidence level on each object in the foreground area image, thereby training the classifier, evaluating and tuning, and obtaining a first classifier, the first classifier being used to identify a target object among multiple objects in the foreground area; The image data of the target object and the monitoring parameters of the transmission line are collected, and the second classifier is trained according to the image data of the target object and the monitoring parameters of the transmission line.

[0011] In some embodiments of the present application, training a second classifier based on the image data of the target object and the monitoring parameters of the power transmission line includes: The image data of the target object and the monitoring parameters of the transmission line are divided according to the state of the transmission line to obtain a data set of a normal state and a data set of an abnormal state, and the data set of the normal state and the data set of the abnormal state are classified according to the image features and the monitoring parameters; Determine the weights of image features and monitoring parameters, fuse multiple image features to obtain image fusion features, fuse multiple monitoring parameters to obtain monitoring fusion features, splice the image fusion features and monitoring fusion features, train the classifier, evaluate and tune, and obtain a second classifier. The second classifier is used to identify the state of the target object and output state information.

[0012] In some embodiments of the present application, determining the weights of image features and monitoring parameters includes: Calculate the coefficient of variation of each image feature and each monitoring parameter, quantify the abnormal state of the transmission line, calculate the correlation between each image feature and each monitoring parameter and the abnormal state of the transmission line, and determine the contribution value of each image feature or monitoring parameter based on the coefficient of variation and the correlation; ; in, It is The contribution value of an image feature or monitoring parameter, For the The correlation between the image features or monitoring parameters and the abnormal state of the transmission line, is the original contribution value, Indicated by The original contribution value is obtained by mapping the correlation between the image features or monitoring parameters and the abnormal state of the transmission line. Indicates The coefficient of variation of an image feature or monitoring parameter, Indicates A constant corresponding to an image feature or monitoring parameter; The weights of each image feature and monitoring parameter are obtained by mapping the contribution values ​​of each image feature and monitoring parameter.

[0013] Correspondingly, the present application also provides an early warning system based on power transmission line image data analysis, including: The first module is used to collect the transmission line image, check the brightness histogram of the transmission line image, and count the brightness distribution characteristics of the transmission line image, and mark the initial foreground area and the initial background area on the transmission line image; The second module is used to analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform brightness uniformity processing on the image in combination with the brightness distribution characteristics of the transmission line image and the foreground-background information; The third module is used to perform filtering and denoising processing on the initial foreground area and the initial background area according to the initial foreground-background information, and separate the foreground area and the background area of ​​the processed power transmission line image; The fourth module is used to identify the target object in the foreground area through the first classifier, obtain the monitoring parameters of the transmission line, match the monitoring parameters with the transmission line image, and input the monitoring parameters and the transmission line image into the second classifier, thereby realizing early warning of the transmission line.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. Combine the brightness distribution characteristics of the transmission line image and the foreground-background information to uniformly process the image brightness, improve the image brightness quality, and provide a reliable basis for subsequent denoising and separation of foreground and background.

[0015] 2. By setting the confidence in the first classifier, the recognition accuracy of the first classifier for the target object is improved. By setting the weights of the image features and monitoring parameters in the second classifier, the two are integrated to improve the early warning accuracy and adaptability of the transmission line, thereby ensuring the safety and normal operation of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of the early warning method based on power transmission line image data analysis proposed by the present invention; Figure 2 This is a schematic diagram of the structure of the early warning system based on power transmission line image data analysis proposed by the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0018] Reference Figure 1 , the early warning method based on the analysis of transmission line image data includes the following steps: Step S101 : collecting a power transmission line image, checking the brightness histogram of the power transmission line image, and counting the brightness distribution characteristics of the power transmission line image, and marking the initial foreground area and the initial background area on the power transmission line image.

[0019] In this embodiment, drones, cameras and other equipment are used to collect images of the transmission line. The collected images should include multiple types such as visible light images and infrared images for comprehensive analysis. Use image processing software (such as OpenCV) to view the brightness histogram of the transmission line image. The brightness histogram can intuitively display the distribution of the number of pixels with different brightness values ​​in the image, which helps to understand the brightness distribution of the image. According to the color, texture and other features in the image, an image segmentation algorithm (such as Otsu algorithm, K-means algorithm) is used to mark the initial foreground area (such as transmission lines, towers, etc.) and the initial background area (such as the sky, the ground, etc.) on the transmission line image.

[0020] It should be noted that the initial marking of the foreground area and the background area is for the subsequent brightness uniformity processing and denoising processing, and the foreground area and the background area of ​​the image are not separated. The separation is performed after the two processing are completed to improve the quality of image preprocessing.

[0021] In some embodiments of the present application, the brightness distribution characteristics of the power transmission line image are counted, and the initial foreground area and the initial background area on the power transmission line image are marked, including: The brightness value range, peak value and skewness are obtained in the brightness histogram of the transmission line image. The brightness value range covers all pixel values ​​from the darkest to the brightest in the image. The peak value is the brightness value with a frequency exceeding the threshold in the image. The skewness is a statistic that describes the asymmetry of the brightness histogram distribution. The transmission line image is divided into a foreground part and a background part by a preset image segmentation algorithm, and the foreground part and the background part are marked and recorded as an initial foreground area and an initial background area.

[0022] In this embodiment, by analyzing the brightness histogram of the transmission line image, the distribution range of the brightness values ​​in the image can be determined. This range usually covers all pixel values ​​from the darkest to the brightest. Analysis of the brightness value range helps to understand the overall brightness level of the image and whether there are areas that are too bright or too dark. The peak in the brightness histogram represents the brightness value that appears more frequently in the image. The location of the peak can reflect the concentration of the main brightness areas in the image, which helps to determine the key areas for brightness adjustment. Skewness is a statistic that describes the asymmetry of the brightness histogram distribution. Positive skewness means that the brightness value is biased towards the lower side, and negative skewness means that the brightness value is biased towards the higher side. Analysis of skewness helps to understand the balance of the image brightness distribution and whether adjustments need to be made for specific brightness areas.

[0023] Step S102: Analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform brightness uniformity processing on the image in combination with the brightness distribution characteristics of the power transmission line image and the foreground-background information.

[0024] In this embodiment, the brightness characteristics of the initial foreground area and the background area are analyzed to form the initial foreground-background information, and the brightness difference between the foreground and background areas is counted to help perform brightness uniformity processing and improve image quality. During the image acquisition process, due to changes in lighting conditions, the transmission line image may have uneven brightness. This will affect the accuracy of subsequent image processing and analysis. Dynamic Range Compression (DRC) is a commonly used technology in image processing, which is used to adjust the distribution of brightness values ​​in the image so that extremely high or extremely low brightness values ​​are compressed to a more reasonable range, thereby improving the effect of subsequent processing.

[0025] In some embodiments of the present application, the initial foreground area and the initial background area are analyzed to form initial foreground-background information, including: Determine the brightness value range of the initial foreground area and the initial background area, calculate the brightness value range deviation between the initial foreground area and the initial background area, obtain the brightness deviation value, and use the brightness value range and the brightness deviation value of the initial foreground area and the initial background area as the initial foreground-background information.

[0026] In this embodiment, the brightness range of the foreground (such as transmission lines, towers and other key targets) needs to be carefully analyzed. This usually involves performing a brightness histogram analysis on the image to determine the distribution of the foreground brightness values ​​and ensure that the target brightness range can fully cover the brightness in the foreground area. The brightness difference between the foreground and the background needs to be evaluated. This can be achieved by calculating the average brightness values ​​of the foreground and background areas and taking the difference between them.

[0027] In some embodiments of the present application, the brightness distribution characteristics of the power transmission line image and the foreground-background information are combined to perform uniform brightness processing on the image, including: Determine a median value in the brightness value range, divide the peak into a left peak value and a right peak value according to the peak values ​​on the left and right sides of the median value, calculate the distances between the median value and the multiple left peak values ​​and the multiple right peak values, respectively, select the maximum distances on both sides, and determine the first brightness range with the left peak value and the right peak value; The endpoint value on the left side or the right side of the first brightness range is adjusted by the skewness, so as to obtain a second brightness range; The union range of the initial foreground area's brightness value range and the second brightness range is taken as the third brightness range, and the two endpoints of the third brightness range are adjusted based on the brightness deviation value to obtain the target brightness range. A mapping function is set according to the target brightness range, and part of the pixels in the power transmission line image are mapped through the mapping function to achieve uniform brightness processing of the image.

[0028] In this embodiment, the skewness is used to adjust the endpoint value on the left or right side of the first brightness range to obtain the second brightness range. Different sizes of skewness correspond to different adjustment factors, which are adjusted in the form of the product of the adjustment factor and the endpoint value to obtain the second brightness range, which describes the comprehensive brightness distribution characteristics to obtain the target brightness range. The brightness deviation value corresponds to different adjustment factors to adjust the two endpoints of the third brightness range. In this way, the differences in the foreground and background areas are taken into account, and a mapping function is designed to map the extremely high or extremely low brightness values ​​in the original image to the compressed brightness range (target brightness range). This mapping function can be linear or nonlinear (such as logarithmic mapping, piecewise linear mapping, etc.), depending on the brightness distribution characteristics of the image and the requirements of the application scenario.

[0029] Step S103, filtering and denoising the initial foreground area and the initial background area using the initial foreground-background information, and separating the foreground area and the background area of ​​the processed power transmission line image.

[0030] In this embodiment, brightness information is often closely related to the type and distribution of noise. By observing the brightness histogram or distribution diagram, the abnormal or noise points of the brightness values ​​in the foreground and background are identified. If the noise in the foreground and background is mainly Gaussian noise (i.e., the brightness value is normally distributed), a Gaussian filter can be selected for denoising. If the noise is mainly salt and pepper noise (i.e., randomly distributed black or white dots appear in the image), a median filter or a mean filter can be selected for denoising. The median filter is particularly effective for removing salt and pepper noise because it can replace the noise point with the median of the neighboring pixels.

[0031] In some embodiments of the present application, filtering and denoising are performed on the initial foreground area and the initial background area using the initial foreground-background information, including: Analyze the noise points in the brightness histogram in the initial foreground area and the initial background area, and calculate the signal-to-noise ratio of the initial foreground area and the signal-to-noise ratio of the initial background area; The first window size range and the second window size range are determined according to the signal-to-noise ratio and brightness value range of the initial foreground area and the initial background area, respectively, and the intersection range of the first window size range and the second window size range is used as the window size of the filter of the initial foreground area and the initial background area, respectively, so as to perform filtering and denoising processing.

[0032] In this embodiment, the signal-to-noise ratio (SNR) is an important indicator for measuring the ratio of signal to noise in an image. In practical applications, it may be necessary to use a simulated signal or a known pure signal to accurately calculate the SNR. For a real image, the improvement of the SNR can be indirectly evaluated by comparing the image quality before and after filtering.

[0033] In this embodiment, in a digital image, brightness is usually expressed as a grayscale value or color brightness value of a pixel. An area with a wider brightness range means that the image contains more brightness changes, which is usually associated with rich details and textures. On the contrary, an area with a narrower brightness range indicates that the brightness changes in the image are small, and the details and textures may be relatively few. The filter window size is the size of the area where the filter is applied to the image. It determines how large a range of pixels in the image the filter can process. Smaller windows are usually used to retain details and edges in the image, while larger windows are more suitable for removing large areas of noise or smoothing images. In areas with a wider brightness range, in order to retain these details, a smaller filter window size needs to be selected so that the local area around each pixel can be processed more accurately. Using a smaller window can reduce the blurring of details and maintain the clarity and sharpness of the image. The same is true in reverse. The first window size range (determined by the signal-to-noise ratio) and the second window size range (determined by the brightness value range) are combined to determine the window size of each filter, and the initial foreground area and the initial background area are targeted to set the window size of the filter.

[0034] It is understandable that the foreground area and the background area are actually separated only after the brightness uniformization processing and the denoising processing are completed, and there will be some irrelevant information in the foreground area, not just the target object.

[0035] Step S104, identifying the target object in the foreground area through the first classifier, obtaining the monitoring parameters of the transmission line, matching the monitoring parameters with the transmission line image, and inputting the monitoring parameters and the transmission line image into the second classifier, thereby realizing early warning of the transmission line.

[0036] In this embodiment, the first classifier is mainly used to identify the target object from the foreground area, and the second classifier is mainly used to combine the monitoring parameters with the transmission line image to identify the current state of the transmission line.

[0037] In some embodiments of the present application, before identifying the target object in the foreground area by the first classifier, the method further includes training the first classifier and the second classifier, and the training process includes: Collect foreground area images containing target objects, mark basic information of target objects in the foreground area images, extract image features of each object in the foreground area images, standardize the image features, divide the image features into single-pixel features and multi-pixel features, and define confidence by combining single-pixel features and multi-pixel features; ; in, is the confidence of the i-th object in the foreground image, , are the number of single-pixel features and multi-pixel features, , Respectively The confidence coefficients of the single-pixel features and the multi-pixel features, , Respectively The size of the single pixel feature and the The size of multi-pixel features, express and The maximum value in is a preset constant; Marking the confidence level on each object in the foreground area image, thereby training the classifier, evaluating and tuning, and obtaining a first classifier, the first classifier being used to identify a target object among multiple objects in the foreground area; The image data of the target object and the monitoring parameters of the transmission line are collected, and the second classifier is trained according to the image data of the target object and the monitoring parameters of the transmission line.

[0038] In this embodiment, single-pixel features (color, brightness, etc.) and multi-pixel features (texture, shape, edge, etc.), single-pixel features are features that each pixel has, while multi-pixel features require multiple pixels to jointly reflect, texture: texture features are usually calculated within a certain area, shape: shape features usually describe the outline and boundary of an object or area.

[0039] In this embodiment, the image features of each object in the foreground area image are extracted, and all objects in the foreground area (including the target object and some irrelevant objects) are identified through the object detection algorithm. The confidence is the credibility score of the object and the target object. The total amount of features generated by combining single-pixel features and multi-pixel features is Indicates the correction of the pixel feature with the greatest influence on the total amount.

[0040] In some embodiments of the present application, training a second classifier based on the image data of the target object and the monitoring parameters of the power transmission line includes: The image data of the target object and the monitoring parameters of the transmission line are divided according to the state of the transmission line to obtain a data set of a normal state and a data set of an abnormal state, and the data set of the normal state and the data set of the abnormal state are classified according to the image features and the monitoring parameters; Determine the weights of image features and monitoring parameters, fuse multiple image features to obtain image fusion features, fuse multiple monitoring parameters to obtain monitoring fusion features, splice the image fusion features and monitoring fusion features, train the classifier, evaluate and tune, and obtain a second classifier. The second classifier is used to identify the state of the target object and output state information.

[0041] In this embodiment, first, the collected image data of the target object and the monitoring parameters of the transmission line are divided according to the state of the transmission line. The state here is usually divided into a normal state and an abnormal state. The divided data will form two data sets: a data set of a normal state and a data set of an abnormal state. The two data sets are further classified according to image features and monitoring parameters. Image features may include color, texture, shape, etc., while monitoring parameters may include current, voltage, temperature, etc. The image fusion features and the monitoring fusion features are spliced ​​to form a feature vector containing more comprehensive information. Use the spliced ​​feature vector to train a classifier. The choice of classifier may depend on the specific application scenario and data characteristics, such as support vector machine (SVM), random forest, neural network, etc. During the training process, the classifier needs to be evaluated to understand its performance. Evaluation indicators may include accuracy, recall rate, F1 score, etc.

[0042] In some embodiments of the present application, determining the weights of image features and monitoring parameters includes: Calculate the coefficient of variation of each image feature and each monitoring parameter, quantify the abnormal state of the transmission line, calculate the correlation between each image feature and each monitoring parameter and the abnormal state of the transmission line, and determine the contribution value of each image feature or monitoring parameter based on the coefficient of variation and the correlation; ; in, It is The contribution value of an image feature or monitoring parameter, For the The correlation between the image features or monitoring parameters and the abnormal state of the transmission line, is the original contribution value, Indicated by The original contribution value is obtained by mapping the correlation between the image features or monitoring parameters and the abnormal state of the transmission line. Indicates The coefficient of variation of an image feature or monitoring parameter, Indicates A constant corresponding to an image feature or monitoring parameter; The weights of each image feature and monitoring parameter are obtained by mapping the contribution values ​​of each image feature and monitoring parameter.

[0043] In this embodiment, it is necessary to analyze the correlation between the image features and monitoring parameters and the state of the transmission line (such as abnormality, potential fault, etc.). This can be achieved by calculating the correlation coefficient between the features or parameters and the state, such as the Pearson correlation coefficient, which directly reflects the correlation with the abnormal state and obtains the original contribution value. It indicates the correction of the original contribution value through the coefficient of variation, that is, the stability of the image feature or monitoring parameter affects the reliability of the correlation. If an image feature remains relatively stable for a long time, it may have a higher weight. For those parameters that change less and can stably reflect the line status, a higher weight should be given. Different contribution values ​​correspond to different weights. Generally speaking, the larger the contribution value, the greater the weight.

[0044] It should be noted that the weight here refers to the relative importance between image features or between monitoring parameters.

[0045] Compared with the prior art, the present invention has the following beneficial effects: 1. Combine the brightness distribution characteristics of the transmission line image and the foreground-background information to uniformly process the image brightness, improve the image brightness quality, and provide a reliable basis for subsequent denoising and separation of foreground and background.

[0046] 2. By setting the confidence in the first classifier, the recognition accuracy of the first classifier for the target object is improved. By setting the weights of the image features and monitoring parameters in the second classifier, the two are integrated to improve the early warning accuracy and adaptability of the transmission line, thereby ensuring the safety and normal operation of the transmission line.

[0047] Correspondingly, the present application also provides an early warning system based on power transmission line image data analysis, including: The first module is used to collect the transmission line image, check the brightness histogram of the transmission line image, and count the brightness distribution characteristics of the transmission line image, and mark the initial foreground area and the initial background area on the transmission line image; The second module is used to analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform brightness uniformity processing on the image in combination with the brightness distribution characteristics of the transmission line image and the foreground-background information; The third module is used to perform filtering and denoising processing on the initial foreground area and the initial background area according to the initial foreground-background information, and separate the foreground area and the background area of ​​the processed power transmission line image; The fourth module is used to identify the target object in the foreground area through the first classifier, obtain the monitoring parameters of the transmission line, match the monitoring parameters with the transmission line image, and input the monitoring parameters and the transmission line image into the second classifier, thereby realizing early warning of the transmission line.

[0048] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0049] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present invention.

[0050] Those skilled in the art will appreciate that the modules in the system in the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0051] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An early warning method based on power transmission line image data analysis, characterized in that: include, Collecting a transmission line image, checking a brightness histogram of the transmission line image, and counting brightness distribution characteristics of the transmission line image, and marking an initial foreground area and an initial background area on the transmission line image; Analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform image brightness uniformity processing in combination with the brightness distribution characteristics of the transmission line image and the foreground-background information; Perform filtering and denoising on the initial foreground area and the initial background area through the initial foreground-background information, and separate the foreground area and the background area of ​​the processed transmission line image; The target object is identified in the foreground area by the first classifier, the monitoring parameters of the transmission line are obtained, and the monitoring parameters are matched with the transmission line image, and the monitoring parameters and the transmission line image are input into the second classifier, thereby realizing early warning of the transmission line.

2. The early warning method based on power transmission line image data analysis according to claim 1 is characterized in that: And count the brightness distribution characteristics of the transmission line image, mark the initial foreground area and the initial background area on the transmission line image, including, The brightness value range, peak value and skewness are obtained in the brightness histogram of the transmission line image. The brightness value range covers all pixel values ​​from the darkest to the brightest in the image. The peak value is the brightness value with a frequency exceeding the threshold in the image. The skewness is a statistic that describes the asymmetry of the brightness histogram distribution. The transmission line image is divided into a foreground part and a background part by a preset image segmentation algorithm, and the foreground part and the background part are marked and recorded as an initial foreground area and an initial background area.

3. The early warning method based on power transmission line image data analysis according to claim 2 is characterized in that: Analyze the initial foreground area and the initial background area to form initial foreground-background information. include, Determine the brightness value range of the initial foreground area and the initial background area, calculate the brightness value range deviation between the initial foreground area and the initial background area, obtain the brightness deviation value, and use the brightness value range and the brightness deviation value of the initial foreground area and the initial background area as the initial foreground-background information.

4. The early warning method based on power transmission line image data analysis according to claim 3 is characterized in that: Combining the brightness distribution characteristics of the transmission line image and the foreground-background information, the image brightness is processed uniformly. include, Determine a median value in the brightness value range, divide the peak into a left peak value and a right peak value according to the peak values ​​on the left and right sides of the median value, calculate the distances between the median value and the multiple left peak values ​​and the multiple right peak values, respectively, select the maximum distances on both sides, and determine the first brightness range with the left peak value and the right peak value; The endpoint value on the left side or the right side of the first brightness range is adjusted by the skewness, so as to obtain a second brightness range; The union range of the initial foreground area's brightness value range and the second brightness range is taken as the third brightness range, and the two endpoints of the third brightness range are adjusted based on the brightness deviation value to obtain the target brightness range. A mapping function is set according to the target brightness range, and part of the pixels in the power transmission line image are mapped through the mapping function to achieve uniform brightness processing of the image.

5. The early warning method based on power transmission line image data analysis according to claim 3 is characterized in that: The initial foreground area and the initial background area are filtered and denoised using the initial foreground-background information. include, Analyze the noise points in the brightness histogram in the initial foreground area and the initial background area, and calculate the signal-to-noise ratio of the initial foreground area and the signal-to-noise ratio of the initial background area; The first window size range and the second window size range are determined according to the signal-to-noise ratio and brightness value range of the initial foreground area and the initial background area, respectively, and the intersection range of the first window size range and the second window size range is used as the window size of the filter of the initial foreground area and the initial background area, respectively, so as to perform filtering and denoising processing.

6. The early warning method based on power transmission line image data analysis according to claim 1 is characterized in that: Before identifying the target object in the foreground area by the first classifier, the method further includes training the first classifier and the second classifier, the training process including: Collect foreground area images containing target objects, mark basic information of target objects in the foreground area images, extract image features of each object in the foreground area images, standardize the image features, divide the image features into single-pixel features and multi-pixel features, and define confidence by combining single-pixel features and multi-pixel features; ； in, is the confidence of the i-th object in the foreground image, , are the number of single-pixel features and multi-pixel features, , Respectively The confidence coefficients of the single-pixel features and the multi-pixel features, , Respectively The size of the single pixel feature and the The size of multi-pixel features, express and The maximum value in is a preset constant; Marking the confidence level on each object in the foreground area image, thereby training the classifier, evaluating and tuning, and obtaining a first classifier, the first classifier being used to identify a target object among multiple objects in the foreground area; The image data of the target object and the monitoring parameters of the transmission line are collected, and the second classifier is trained according to the image data of the target object and the monitoring parameters of the transmission line.

7. The early warning method based on power transmission line image data analysis according to claim 6 is characterized in that: The second classifier is trained based on the image data of the target object and the monitoring parameters of the power transmission line, including: The image data of the target object and the monitoring parameters of the transmission line are divided according to the state of the transmission line to obtain a data set of a normal state and a data set of an abnormal state, and the data set of the normal state and the data set of the abnormal state are classified according to the image features and the monitoring parameters; Determine the weights of image features and monitoring parameters, fuse multiple image features to obtain image fusion features, fuse multiple monitoring parameters to obtain monitoring fusion features, splice the image fusion features and monitoring fusion features, train the classifier, evaluate and tune, and obtain a second classifier. The second classifier is used to identify the state of the target object and output state information.

8. The early warning method based on power transmission line image data analysis according to claim 7 is characterized in that: Determine the weights of image features and monitoring parameters, include, Calculate the coefficient of variation of each image feature and each monitoring parameter, quantify the abnormal state of the transmission line, calculate the correlation between each image feature and each monitoring parameter and the abnormal state of the transmission line, and determine the contribution value of each image feature or monitoring parameter based on the coefficient of variation and the correlation; ； in, It is The contribution value of an image feature or monitoring parameter, For the The correlation between the image features or monitoring parameters and the abnormal state of the transmission line, is the original contribution value, Indicated by The original contribution value is obtained by mapping the correlation between the image features or monitoring parameters and the abnormal state of the transmission line. Indicates The coefficient of variation of an image feature or monitoring parameter, Indicates A constant corresponding to an image feature or monitoring parameter; The weights of each image feature and monitoring parameter are obtained by mapping the contribution values ​​of each image feature and monitoring parameter.

9. The early warning system based on the analysis of transmission line image data is characterized by: include, The first module is used to collect the transmission line image, check the brightness histogram of the transmission line image, and count the brightness distribution characteristics of the transmission line image, and mark the initial foreground area and the initial background area on the transmission line image; The second module is used to analyze the initial foreground area and the initial background area to form initial foreground-background information, and perform brightness uniformity processing on the image in combination with the brightness distribution characteristics of the transmission line image and the foreground-background information; The third module is used to perform filtering and denoising processing on the initial foreground area and the initial background area according to the initial foreground-background information, and separate the foreground area and the background area of ​​the processed power transmission line image; The fourth module is used to identify the target object in the foreground area through the first classifier, obtain the monitoring parameters of the transmission line, match the monitoring parameters with the transmission line image, and input the monitoring parameters and the transmission line image into the second classifier, thereby realizing early warning of the transmission line.

Images