A power transmission line periphery hidden danger dynamic detection system based on image difference recognition

Abstract

The application provides a power transmission line periphery hidden danger dynamic detection system based on image difference recognition, and relates to the technical field of data processing, and comprises a data acquisition module, a video and image data of the periphery of a power transmission line are acquired by the data acquisition module, the data is derived from video monitoring equipment arranged in the periphery of the power transmission line, an image preprocessing module, the acquired video and image data are preprocessed by the image preprocessing module to obtain preprocessed video and image data, an image difference recognition module, the preprocessed image is recognized by difference by the image difference recognition module by using a machine learning algorithm, the abnormal value of the periphery environment of the power transmission line is detected by comparing the images of continuous frames, the abnormal value of the periphery environment of the power transmission line is adjusted according to a dynamic factor to obtain a correction index, and a hidden danger analysis module, the correction index is analyzed by the hidden danger analysis module to determine the type, position and severity of the hidden danger. The application improves the efficiency and accuracy of hidden danger detection in the periphery of the power transmission line.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a power transmission line periphery hidden danger dynamic detection system based on image difference recognition. BACKGROUND

[0002] With the rapid development of the power industry, the safe operation of power transmission lines is crucial to ensuring the stable supply of the entire power system. However, since power transmission lines often pass through complex geographical environments, changes in the surrounding environment can pose potential threats to the line safety. Therefore, timely detection and identification of hidden dangers in the periphery of power transmission lines is of great significance to prevent accidents and ensure the continuity of power supply.

[0003] Some existing power transmission line periphery hidden danger detection systems can automatically identify some hidden dangers, but lack consideration of dynamic changes between consecutive frames of images, leading to missed or false detection of some transient or progressive hidden dangers. At the same time, the robustness of these systems in handling complex backgrounds and variable lighting conditions needs to be improved. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a power transmission line periphery hidden danger dynamic detection system based on image difference recognition, which improves the efficiency and accuracy of power transmission line periphery hidden danger detection.

[0005] To solve the above technical problems, the technical solution of the present application is as follows:

[0006] In a first aspect, a power transmission line periphery hidden danger dynamic detection system based on image difference recognition includes:

[0007] A data acquisition module for acquiring video and image data in the periphery of a power transmission line, the data being derived from video monitoring equipment arranged in the periphery of the power transmission line;

[0008] An image preprocessing module for preprocessing the acquired video and image data to obtain preprocessed video and image data;

[0009] An image difference recognition module for recognizing differences in preprocessed images using a machine learning algorithm, detecting abnormal values in the periphery of the power transmission line by comparing consecutive frames of images; adjusting the abnormal values in the periphery of the power transmission line according to dynamic factors to obtain a correction index;

[0010] A hidden danger analysis module for analyzing the correction index to determine the type, location and severity of the hidden danger;

[0011] A warning prompt module for generating warning information according to the type, location and severity of the hidden danger, and real-time prompting the operation and maintenance personnel through the user interface.

[0012] Further, video and image data of the power transmission line periphery are collected, and the data is derived from video monitoring devices arranged in the power transmission line periphery, including:

[0013] In the arrangement space of the video monitoring devices of the power transmission line periphery, a certain number of grey wolf individuals are randomly initialized, i.e. candidate video monitoring device positions; each grey wolf individual represents a solution, including the position, orientation and focal length parameters of the video monitoring device;

[0014] A target function is defined, which is used to evaluate the pros and cons of the grey wolf individual, i.e. the quality of the data collected by the video monitoring device under the configuration of the specific position and parameters;

[0015] The grey wolf individual updates its own position according to the current final solution, simulates the hunting behavior of the grey wolf, searches for the corresponding solution in the solution space, and the grey wolf individual gradually approaches the current corresponding solution. When the grey wolf individual approaches the final solution, search is performed to find the global final solution. According to the output result of the grey wolf optimization algorithm, the position and parameter configuration of the video monitoring device are adjusted;

[0016] Under the updated configuration of the video monitoring device, actual video and image data collection is performed to obtain the video and image data of the power transmission line periphery.

[0017] Further, the collected video and image data are preprocessed to obtain preprocessed video and image data, including:

[0018] In the space of the preprocessing parameters, a certain number of particles are randomly initialized, i.e. candidate preprocessing parameter combinations. Each particle represents a set of preprocessing parameters, including the type, size and threshold of the filter;

[0019] A fitness function is defined to evaluate the quality of the preprocessed video and image data;

[0020] According to the individual historical final position and the group historical final position, the speed of each particle is updated, and according to the updated speed, the position of each particle, i.e. the preprocessing parameter, is adjusted;

[0021] The preprocessing operation is performed on each new position, and the preprocessing effect is evaluated using the fitness function. The speed updating, position updating and evaluation updating steps are repeatedly executed until a preset number of iterations is reached to obtain the preprocessing parameter combination corresponding to the group historical final position;

[0022] The preprocessing operation is performed on the collected video and image data using the preprocessing parameter combination corresponding to the final position to obtain the preprocessed video and image data.

[0023] Further, the pre-processed images are subjected to difference recognition using a machine learning algorithm, and by comparing successive frames of images, abnormal values in the environment surrounding the power transmission line are detected, including:

[0024] Successive image frames are extracted from the video stream at a set frame rate;

[0025] Feature extraction is performed on each image frame to obtain a feature vector representing the image content;

[0026] The number of initial frames used to establish a Gaussian mixture model is determined, and the feature vectors of the initial frames are input into the Gaussian mixture model as training data, and the distribution of the feature vectors is learned through the Gaussian mixture model algorithm to establish a Gaussian mixture model representing the "normal" state;

[0027] The parameters of the Gaussian mixture model are set, including the number of Gaussian distributions and the type of covariance matrix;

[0028] For each new image frame in the video stream, its feature vector is extracted, and the feature vector of the new image is input into the established Gaussian mixture model, and the difference value between the new image feature vector and each Gaussian distribution in the Gaussian mixture model is calculated to obtain the difference value between the new image and the Gaussian mixture model.

[0029] Further, feature extraction is performed on each image frame to obtain a feature vector representing the image content, including:

[0030] The pre-processed single-frame image data is received as input, and the SIFT feature detection algorithm is initialized, including setting the number of layers and scale factors for scale space extreme value detection;

[0031] By performing Gaussian blur and downsampling on the input image at different scales, a scale space of the image is constructed to find stable feature points at different scales;

[0032] In the scale space, by comparing the pixel values of each point with adjacent scales and adjacent position points, local extreme points are found, and the extreme points are considered as candidate key points;

[0033] The candidate key points detected in the previous step are screened and positioned, and by calculating the gradient direction and amplitude of the pixels around the key points, a main direction is assigned to each key point;

[0034] In the neighborhood around the key point, according to the assigned main direction, the gradient information of the pixels is counted to generate a fixed-length descriptor vector;

[0035] All the descriptor vectors of the key points are sorted into a feature vector set, and the feature vector set represents the content of the entire image.

[0036] Further, the distribution of the feature vectors is learned by a Gaussian mixture model algorithm to establish a Gaussian mixture model representing the "normal" state, including:

[0037] Collecting feature vector data representing the "normal" state, determining the number of Gaussian distributions used, i.e. the number of components of the mixture model; initializing parameters for each Gaussian distribution, including mean vector, covariance matrix and mixing coefficient, i.e. the weight of each Gaussian distribution in the mixture model;

[0038] Iteratively updating the Gaussian mixture model parameters using the expectation maximization algorithm, calculating the posterior probability of each data point belonging to each Gaussian distribution, i.e. the responsibility, according to the current Gaussian mixture model parameters;

[0039] Based on the calculated posterior probability, updating the mean, covariance and mixing coefficient of each Gaussian distribution, and iteratively converging until the Gaussian mixture model parameters converge;

[0040] Using the silhouette coefficient to evaluate the fitting effect of the Gaussian mixture model to obtain an evaluation result;

[0041] According to the evaluation result, determining the corresponding Gaussian mixture model as the Gaussian mixture model representing the "normal" state.

[0042] Further, the calculation formula of the dynamic factor is:

[0043]

[0044] Where A(t) represents the dynamic adjustment factor; α represents the adjustment coefficient; I i (t) and I i (t+1) represent the gray value of the i-th pixel in image I(t) and image I(t+1) respectively; N is the total number of pixels in the image; β represents the adjustment coefficient; v(t) represents the wind speed at time t; κ is a constant, representing the magnification multiple of wind speed on image difference; γ represents the adjustment coefficient; b i (t) is the actual value of the i-th pixel in image b(t); is the denoising estimated value of the i-th pixel in image b(t); i represents the index value.

[0045] Further, the abnormal values of the environment around the transmission line are adjusted according to the dynamic factor to obtain a correction index, including:

[0046] The dynamic factor is multiplied by the corresponding difference value to obtain the correction index.

[0047] The second aspect is a transmission line surrounding hidden danger dynamic detection method based on image difference recognition, which comprises the following steps:

[0048] Collect video and image data around the power transmission line, the data is from the video monitoring device arranged around the power transmission line;

[0049] Preprocess the collected video and image data to obtain preprocessed video and image data;

[0050] Differential recognition is performed on the preprocessed image by using a machine learning algorithm, and by comparing the images of consecutive frames, the abnormal values of the environment around the power transmission line are detected; The abnormal values of the environment around the power transmission line are adjusted according to the dynamic factor to obtain a correction index;

[0051] The correction index is analyzed to determine the type, location and severity of the hidden danger;

[0052] According to the type, location and severity of the hidden danger, generate early warning information, and real-time prompt operation and maintenance personnel through user interface.

[0053] The third aspect is a computing device, comprising:

[0054] One or more processors;

[0055] Storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors realize the method.

[0056] The fourth aspect is a computer readable storage medium, the computer readable storage medium stores a program, the program is executed by the processor to realize the method.

[0057] The above scheme of the present application at least includes the following beneficial effects:

[0058] Through the real-time data collection of the video monitoring device arranged around the power transmission line, the system can continuously monitor the environment around the line, and timely find the hidden danger, this automatic data collection method greatly improves the detection efficiency and real-time performance, reduces the frequency and cost of manual inspection.

[0059] The image differential recognition module uses machine learning algorithm to recognize the difference of the preprocessed image, which can accurately detect the slight change of the environment around the power transmission line, and effectively identify the abnormal value by comparing the consecutive frames, this technical means improves the accuracy and sensitivity of hidden danger identification.

[0060] The detected abnormal values are adjusted by the dynamic factor to obtain the correction index, which helps to reduce false positives and false negatives, makes the system more intelligent and adaptive, and the dynamic adjustment mechanism makes the system can cope with complex and changeable environmental conditions, improves the robustness of the system.

[0061] The hidden danger analysis module analyzes the correction index, can accurately determine the type, position and severity of the hidden danger, which provides detailed hidden danger information for the operation and maintenance personnel, and helps them to make quick and accurate response. The early warning prompt module can generate early warning information according to the specific situation of the hidden danger, and real-time prompt operation and maintenance personnel through the user interface. This instant feedback mechanism greatly shortens the time from discovering hidden dangers to taking measures, and improves the safety and stability of the power system. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 is a schematic diagram of a power transmission line peripheral hidden danger dynamic detection system based on image difference recognition provided by an embodiment of the present application.

[0063] Figure 2 is a flowchart of a power transmission line peripheral hidden danger dynamic detection method based on image difference recognition provided by an embodiment of the present application. DETAILED DESCRIPTION

[0064] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0065] As Figure 1 shown, an embodiment of the present application proposes a power transmission line peripheral hidden danger dynamic detection system based on image difference recognition, comprising:

[0066] A data acquisition module is configured to acquire video and image data around the power transmission line, wherein the data is obtained from video monitoring equipment arranged around the power transmission line.

[0067] An image preprocessing module is configured to preprocess the acquired video and image data to obtain preprocessed video and image data.

[0068] An image difference recognition module is configured to use a machine learning algorithm to recognize differences in the preprocessed image, detect abnormal values of the environment around the power transmission line by comparing consecutive frames of images, and adjust the abnormal values of the environment around the power transmission line according to dynamic factors to obtain a correction index.

[0069] A hidden danger analysis module is configured to analyze the correction index and determine the type, position and severity of the hidden danger.

[0070] An early warning prompt module is configured to generate early warning information according to the type, position and severity of the hidden danger, and real-time prompt operation and maintenance personnel through the user interface.

[0071] In the embodiments of the present application, by arranging video monitoring devices around the power transmission line, real-time and continuous video and image data around the power transmission line can be collected, providing a rich data source for subsequent image processing and hidden danger detection, which helps to timely discover and handle potential safety hazards, and improves the safety and stability of the power transmission line. Preprocessing of the collected raw video and image data, such as denoising, enhancement, scaling, etc., can improve image quality, reduce the influence of environmental factors (such as light changes, shadows, etc.) on subsequent difference identification, and the preprocessed image data is clearer and more accurate, which helps to improve the accuracy and efficiency of hidden danger detection. Using machine learning algorithms to identify differences in preprocessed images, by comparing the image changes of consecutive frames, abnormal values of the environment around the power transmission line can be sensitively captured, and this dynamic detection method can quickly discover hidden dangers such as rapid tree growth, illegal building of buildings, external damage, etc. At the same time, according to the dynamic factor, the abnormal value is adjusted to obtain a correction index, which can further improve the accuracy and reliability of the detection. Through in-depth analysis of the correction index, the system can accurately determine the type (such as natural disasters, human damage, etc.), specific location and severity of the hidden danger, which helps the maintenance personnel to quickly understand the hidden danger situation and develop a reasonable handling plan, so as to timely eliminate safety hazards and prevent accidents. According to the results of hidden danger analysis, the system can automatically generate warning information and real-time prompt to the maintenance personnel through the user interface, which can ensure that the maintenance personnel knows the hidden danger situation in the first time and takes prompt measures, effectively improving the safety management level of the power transmission line. At the same time, the detail and accuracy of the warning information also help to improve the pertinence and efficiency of the maintenance work.

[0072] In a preferred embodiment of the present application, video and image data around the power transmission line are collected, which are derived from video monitoring devices arranged around the power transmission line, comprising:

[0073] In the arrangement space of the video monitoring device around the power transmission line, a certain number of grey wolf individuals are randomly initialized, i.e. candidate video monitoring device positions; each grey wolf individual represents a solution, including the position, orientation and focal length parameters of the video monitoring device, specifically including: determining the spatial range where the video monitoring device can be arranged, which is usually based on the orientation of the power transmission line, the surrounding environment and safety specifications, etc. Within this spatial range, a certain number of grey wolf individuals are randomly generated, each representing a candidate video monitoring device position and parameter configuration. Each grey wolf individual not only contains position information (such as latitude and longitude or three-dimensional coordinates), but also includes the orientation (such as azimuth angle, pitch angle) and focal length parameters of the device. These parameters will affect the field of view and monitoring effect of the monitoring device.

[0074] A target function is defined to evaluate the quality of data collected by the video surveillance device at a specific location and parameter configuration, and the calculation formula of the target function is:

[0075]

[0076] where F(x) represents the target function, which is used to evaluate the quality of data collected by the video surveillance device at a specific location and parameter configuration; w1, w2 and w3 are weight coefficients, which are used to balance the importance of image sharpness, monitoring range coverage and weighted target visibility; these weights can be adjusted according to actual needs; represents the image sharpness, which is measured by calculating the sum of gradients in x and y directions of each pixel point in the image; the larger the gradient, the clearer the image edge, and thus the higher the image sharpness; This part represents the coverage rate of the monitoring range, where A is the overlapping area of the monitoring device field of view and the area to be monitored, and T is the total area of the area to be monitored; the larger this ratio, the wider the coverage range of the monitoring device; This part represents the unweighted target visibility, where P is the number of pixels of the target in the image, S is the total number of pixels in the image, C is the average contrast of the target, M is the maximum possible contrast, and d is the distance between the monitoring device and the key target; the larger this value, the more obvious the target in the image; μ is set as the center position of the key target (generally 0), and σ is the standard deviation of the Gaussian function, which is used to control the speed of weight decay with distance; this formula makes the device configuration closer to the key target obtain a higher evaluation; ∑ i,j is a summation symbol, which represents the summation of all pixel points (i, j) in the image, where i and j represent the horizontal and vertical coordinates of the pixel point, respectively; This is the gradient of image I in the x direction at pixel point (i, j) (i.e. the rate of brightness change in the horizontal direction), which reflects the degree and direction of change of pixel values in the image, and is one of the important indicators for evaluating image sharpness. If the image has clear edges and rich details in the horizontal direction, the gradient value will be relatively large; This is the gradient of image I in the y direction at pixel point (i, j) (i.e. the rate of brightness change in the vertical direction), which is similar to , it reflects the sharpness of the image in the vertical direction.

[0077] The gray wolf individual updates its own position according to the current final solution, simulates the hunting behavior of the gray wolf, searches for a corresponding solution in the solution space, and the gray wolf individual gradually approaches the current corresponding solution. When the gray wolf individual approaches the final solution, a search is performed to find the global final solution. According to the output result of the gray wolf optimization algorithm, the position and parameter configuration of the video monitoring device are adjusted, specifically including: generating a gray wolf individual as an initial wolf group, evaluating the fitness of each gray wolf individual according to the target function, and selecting the individual with the highest fitness as the leader wolf (i.e. the current optimal solution); other gray wolf individuals update their positions according to the position of the leader wolf and their own positions, simulate the hunting behavior of the gray wolf, and repeat the process of evaluating fitness and updating positions until the stopping condition is met (such as reaching the maximum number of iterations or the fitness improvement being less than a certain threshold). When the gray wolf individual approaches the final solution, a fine search is performed to find the global optimal solution. According to the output result of the gray wolf optimization algorithm (i.e. the position and parameter configuration of the optimal gray wolf individual), the position, orientation and focal length of the actual video monitoring device are adjusted. After adjusting the configuration, actual video monitoring testing is performed to verify whether the new configuration meets the expected effect.

[0078] Under the updated video monitoring device configuration, actual video and image data collection is performed to obtain video and image data around the power transmission line, specifically including: under the updated and verified video monitoring device configuration, starting the device to perform actual video and image data collection, saving the collected video and image data, and performing necessary preprocessing and analysis work for subsequent use or further analysis.

[0079] In the embodiments of the present application, by randomly initializing a certain number of gray wolf individuals (i.e. candidate video monitoring device positions), the diversity and comprehensiveness of the search process can be ensured, thereby increasing the possibility of finding the global optimal solution. This initialization method helps to avoid the algorithm falling into a local optimal solution and improves the overall optimization effect. Each gray wolf individual is designed to include the position, orientation and focal length parameters of the video monitoring device. Such a design can more comprehensively reflect the actual deployment of the video monitoring device, making the optimization process more in line with actual needs. By considering multiple parameters simultaneously, comprehensive optimization of video monitoring device configuration can be achieved, improving the quality of data acquisition. The quality of data acquisition by the video monitoring device under a specific position and parameter configuration is evaluated by defining an objective function, which is directly related to the quality requirements of data acquisition, ensuring that the optimization process is consistent with the actual application requirements. The position of the gray wolf individual is updated by simulating the hunting behavior of the gray wolf. This search strategy can effectively find better solutions in the solution space. The bionics principle of the gray wolf optimization algorithm makes the search process adaptive and intelligent, which can improve the search efficiency and the probability of finding the global optimal solution. The position and parameter configuration of the video monitoring device are adjusted according to the output results of the gray wolf optimization algorithm, which ensures that the optimization results can be directly applied to the actual system to improve the performance of the system. By using the algorithm to guide the actual configuration adjustment, precise optimization of the video monitoring system can be achieved, improving the efficiency and accuracy of data acquisition. Actual video and image data acquisition is performed under the updated video monitoring device configuration, which can verify the effectiveness of the optimization algorithm and ensure that the collected data meets the application requirements. The optimization effect is fed back by the actual acquisition data, forming a closed-loop optimization process, which helps to continuously improve and enhance the performance of the video monitoring system.

[0080] In a preferred embodiment of the present application, the collected video and image data is preprocessed to obtain preprocessed video and image data, including:

[0081] In the space of preprocessing parameters, a certain number of particles are randomly initialized, i.e. candidate combinations of preprocessing parameters. Each particle represents a set of preprocessing parameters, including the type, size and threshold of the filter. Specifically, the preprocessing parameter range is determined, which includes the type of filter (such as Gaussian filter, median filter, etc.), the size of filter (i.e. the size of the kernel) and the threshold (for some types of filters, such as edge detection filter); a certain number of particles are randomly generated in the space of preprocessing parameters, each particle representing a candidate combination of preprocessing parameters. Each particle is assigned an initial position and velocity. Here, "position" refers to the specific value of the preprocessing parameter, and "velocity" refers to the direction and step size of parameter adjustment.

[0082] A fitness function is defined for evaluating the quality of the pre-processed video and image data, wherein the fitness function is calculated according to the following formula:

[0083]

[0084] wherein μ X and μ Y are the mean values of the original image and the pre-processed image respectively; σ X and σ Y are the standard deviations of the original image and the pre-processed image respectively; σ XY is the covariance of the original image and the pre-processed image; C1 and C2 are constants; G x (i,j) and G y (i,j) are the gradient values of the image in the x and y directions respectively; MxN is the total number of pixels of the image; σ n is the noise standard deviation in the original image (calculated without denoising); σ p is the noise standard deviation of the pre-processed image; w1, w2 and w3 are weight coefficients, and the fitness function will guide the particles to move towards the optimal pre-processing parameter combination to achieve the best data pre-processing effect.

[0085] The velocity of each particle is updated according to the individual historical final position and the group historical final position, and the position of each particle, i.e. the pre-processing parameter, is adjusted according to the updated velocity, which specifically includes: for each particle, recording the position with the highest fitness function value in its history, i.e. the individual historical best position, finding the position with the highest fitness function value among all particles, i.e. the group historical best position, updating the velocity of each particle according to the individual historical best position, the group historical best position and the current velocity of the particle through the standard velocity update formula of PSO, and adjusting the position of each particle, i.e. the value of the pre-processing parameter, according to the updated velocity.

[0086] The pre-processing operation is performed on each new position, and the pre-processing effect is evaluated using the fitness function, and the steps of velocity updating, position updating and evaluation updating are repeatedly executed until a preset iteration number is reached, so as to obtain the pre-processing parameter combination corresponding to the group historical final position, which specifically includes: performing the pre-processing operation, such as filtering and denoising, on each new position (i.e. the new pre-processing parameter combination), evaluating the quality of the pre-processed video and image data using the fitness function to obtain a fitness value; constantly repeating the steps of updating the velocity and position of the particle and evaluating the pre-processing effect until a preset iteration number is reached or other stopping conditions are met, and after the iteration is completed, outputting the pre-processing parameter combination corresponding to the group historical best position.

[0087] The collected video and image data are preprocessed using the preprocessed parameter combination corresponding to the final position to obtain preprocessed video and image data, specifically including: using the output optimal preprocessed parameter combination to preprocess the collected original video and image data, and after preprocessing, video and image data with improved quality are obtained.

[0088] In the embodiments of the present application, the particle swarm optimization algorithm is used to search for the optimal preprocessed parameter combination, including the type, size and threshold of the filter, which can more effectively remove noise in the image, enhance the edges and details of the image, and thus significantly improve the quality of the video and image data. Since the particle swarm optimization algorithm is a population-based random search algorithm, it can adaptively adjust the search strategy and does not depend on the specific form of the problem, so it can show good adaptability and robustness for different types of video and image data and data collected in different environments. Traditional preprocessed parameter setting often relies on experienced professionals to manually adjust, which not only consumes time and effort, but also cannot guarantee the best processing effect each time. The particle swarm optimization algorithm can automatically find the optimal preprocessed parameter combination, greatly reducing the need for manual intervention and improving processing efficiency. By automatically adjusting the preprocessed parameters through the algorithm, image information loss or redundant calculation caused by inappropriate processing can be avoided, so that computing resources can be more reasonably utilized to achieve efficient video and image processing. The particle swarm optimization algorithm can be combined with other image processing techniques to form a more complex preprocessing process, and at the same time, the processing effect has the potential to be further improved as the algorithm is continuously iterated and optimized.

[0089] In a preferred embodiment of the present application, a machine learning algorithm is used to identify differences in preprocessed images, and by comparing consecutive image frames, abnormal values in the environment around the power transmission line are detected, including:

[0090] Continuous image frames are extracted from the video stream at a set frame rate, specifically including: opening the video stream using a video processing library (such as OpenCV), setting a frame rate, for example, extracting 5 image frames per second; traversing the video stream, extracting image frames at the set frame rate interval, and saving as image files or directly processing in memory.

[0091] Feature extraction is performed on each image frame to obtain a feature vector representing the image content.

[0092] The number of initial frames used to establish the Gaussian mixture model is determined, for example, the first 100 frames, and the feature vectors of the initial frames are input into the Gaussian mixture model as training data, and the distribution of the feature vectors is learned through the Gaussian mixture model algorithm to establish a Gaussian mixture model representing the "normal" state.

[0093] The parameters of the Gaussian mixture model are set, including the number of Gaussian distributions and the type of covariance matrix, specifically including: according to the application requirements and experimental results, the number of Gaussian distributions in the Gaussian mixture model is set, and the type of covariance matrix is selected, such as a diagonal matrix, these parameters will affect the complexity of the model and the fitting ability to the data.

[0094] For each new image in the video stream, its feature vector is extracted, and the feature vector of the new image is input into the established Gaussian mixture model, and the difference value of the new image feature vector and each Gaussian distribution in the Gaussian mixture model is obtained by calculating the Mahalanobis distance, specifically including: a new frame of image is extracted from the video stream, and the feature vector representing the content of the image has been obtained by the feature extraction method (such as SIFT), which will be used as the basis for analyzing whether the frame image is consistent with the "normal" state; Next, the feature vector is input into the Gaussian mixture model (GMM) that has been trained before, the GMM is a probability model composed of multiple Gaussian distributions, each Gaussian distribution has its mean and covariance matrix, these parameters have been estimated by algorithm in the training stage. For the input feature vector, calculate the Mahalanobis distance of each Gaussian distribution in the GMM, the Mahalanobis distance is a distance measurement method considering the covariance structure of the data, the Mahalanobis distance considers the correlation between dimensions, by calculating the Mahalanobis distance of the input feature vector and each Gaussian distribution, a set of distance values can be obtained, these distance values reflect the similarity between the input feature vector and each Gaussian distribution in the GMM; After obtaining the Mahalanobis distance, further process the distance values to obtain a comprehensive difference value, the difference value can be the average of all Mahalanobis distances, the size of the difference value reflects the deviation of the input feature vector from the "normal" state model. Finally, a threshold is set to determine whether the input feature vector is abnormal, the threshold is usually determined by experiment or experience, which represents the boundary between "normal" state and "abnormal" state. If the calculated difference value exceeds the threshold, then the image frame is considered abnormal, because it is significantly different from the "normal" state represented by the training data.

[0095] In the embodiments of the present application, by setting the frame rate to continuously extract image frames from the video stream, the system can monitor the surrounding environment of the power transmission line in real time. Once an anomaly is detected, an alarm can be triggered immediately or other measures can be taken, thereby greatly reducing potential safety risks. Feature extraction is performed on each image frame to obtain a feature vector representing the image content, which helps to compress data and focus on key information in the image. The use of feature vectors improves the efficiency and accuracy of subsequent anomaly detection processes. By learning the distribution of feature vectors through Gaussian Mixture Model (GMM), the system can establish a model representing the "normal" state, which is adaptive and can be updated according to changes in the actual environment, thereby maintaining an accurate description of the "normal" state. Users can set parameters of the Gaussian Mixture Model, such as the number of Gaussian distributions and the type of covariance matrix, to adjust the complexity and sensitivity of the model according to specific application scenarios. This flexibility and configurability enables the system to be widely applicable to different monitoring environments and requirements. By calculating the Mahalanobis distance between the feature vector of the new image and each Gaussian distribution in the Gaussian Mixture Model, the system can accurately identify outliers that deviate significantly from the "normal" state. This method is robust to interference factors such as light changes, shadows, and occlusions, reducing false positives and false negatives. The difference value and anomaly detection results provided by the system can serve as data-driven decision support to help maintenance personnel more accurately assess the safety of the power transmission line and develop targeted maintenance strategies.

[0096] In a preferred embodiment of the present application, feature extraction is performed on each image frame to obtain a feature vector representing the image content, including:

[0097] The pre-processed single-frame image data is received as input, and the SIFT feature detection algorithm is initialized, including setting the number of layers and scale factors for scale space extreme value detection, specifically including: receiving the pre-processed single-frame image as the input of the algorithm, the preprocessing includes steps such as grayscale, noise reduction, and contrast enhancement to improve the accuracy of feature extraction; initialize the SIFT algorithm and set the algorithm parameters, among which the important parameters include the number of layers and scale factors for scale space extreme value detection, the number of layers determines the depth of the scale space, and the scale factor controls the scale change between adjacent scale layers.

[0098] A scale space of the image is constructed by performing Gaussian blur and downsampling on the input image at different scales to find stable feature points at different scales, specifically including: performing Gaussian blur on the input image at different scales, the purpose of Gaussian blur is to simulate the performance of the image at different scales, which is achieved by convolving the image with Gaussian kernels of different scales, and downsampling the blurred image to generate a series of images at different scales, which constitute the scale space of the image. Each scale layer represents the information of the image at different resolutions.

[0099] In the scale space, local extreme points are found by comparing the pixel values of each point with its neighboring scales and neighboring positions, extreme points are considered as candidate key points, specifically including: traversing each point in the scale space, comparing it with its neighboring scales and neighboring positions, which usually involves comparing the pixel values of the current point with its upper and lower scales and surrounding 8 points, if the current point is extreme (maximum or minimum) in all these comparisons, it is marked as a local extreme point, these extreme points are considered as candidate key points because they exhibit significant features at different scales.

[0100] The candidate key points detected in the previous step are screened and positioned, and a main direction is assigned to each key point by calculating the gradient direction and amplitude of the pixels around the key point, specifically including: further screening the candidate key points detected in the previous step to remove low-contrast points and unstable edge response points, which is usually achieved by calculating the Hessian matrix (or similar methods) at the key point to evaluate its stability and significance, accurately positioning the screened key points, which includes adjusting the position and scale of the key points to make them more accurately correspond to the actual features in the image, assigning a main direction to each key point by calculating the gradient direction and amplitude of the pixels around the key point, which is used to ensure the rotation invariance of SIFT features.

[0101] In the neighborhood around the key point, the gradient information of the pixels is counted according to the assigned main direction to generate a fixed-length descriptor vector, specifically including: in the neighborhood around the key point, the gradient information of the pixels is counted according to the assigned main direction, which usually involves dividing the neighborhood into multiple sub-regions and calculating the statistical information (such as histogram) of the gradient direction and amplitude in each sub-region, integrating these statistical information into a fixed-length descriptor vector, which captures the local image structure information around the key point and has certain stability for image rotation, scale change and illumination change.

[0102] All the descriptor vectors of the key points are arranged into a feature vector set, which represents the content of the entire image, specifically including: arranging all the descriptor vectors of the key points into a feature vector set, which represents the content of the entire image and can be used for subsequent image matching, recognition and other tasks.

[0103] In the embodiments of the present application, the SIFT feature has good stability to the scale change and rotation of the image, which means that even if the objects in the image present different sizes or directions due to the change of shooting distance or angle, SIFT can still accurately identify and extract the features of these objects, which greatly enhances the robustness and scope of application of the feature extraction method. SIFT focuses on local features of the image rather than global features. This enables it to accurately identify objects or regions of interest in a complex background, even if these objects or regions are partially occluded or overlapped with other objects. In addition, local features are also more conducive to subsequent image matching and recognition tasks. By searching for stable feature points at different scales and going through the screening and positioning process, SIFT can extract the most significant and stable features in the image. These features not only have moderate quantity but also high quality, and can effectively represent the content of the image, which improves the stability and reliability of the feature vector. SIFT generates a fixed-length descriptor vector for each key point. This fixed-length representation is not only convenient for storage and management, but also conducive to subsequent machine learning algorithm processing. For example, when training a classifier or building an image retrieval system, these fixed-length feature vectors can be directly used as input. Due to the above advantages of SIFT features, it is widely used in various computer vision tasks such as image matching, target tracking, three-dimensional reconstruction, etc. Therefore, using SIFT for feature extraction can provide rich choices and flexibility for subsequent application development.

[0104] In a preferred embodiment of the present application, the distribution of feature vectors is learned by Gaussian mixture model algorithm to establish a Gaussian mixture model representing the "normal" state, including:

[0105] Collect feature vector data representing the "normal" state, determine the number of Gaussian distributions used, i.e. the number of components of the mixture model; initialize the parameters of each Gaussian distribution, including the mean vector, covariance matrix and mixing coefficient, i.e. the weight of each Gaussian distribution in the mixture model, specifically including: collecting a large amount of data generated under the "normal" state from the actual application scenario (such as monitoring video, sensor data, etc.), preprocessing these data including cleaning, standardization, etc. to eliminate the influence of noise and outliers, extracting the feature vectors of the data, which should be able to effectively represent the essential properties of the data; according to the complexity of the data and the actual needs, select the appropriate number of Gaussian distributions (i.e. the number of components of the mixture model), initialize the parameters of each Gaussian distribution, including the mean vector, covariance matrix and mixing coefficient, these parameters can be initialized by random way or based on the statistical characteristics of the data.

[0106] The Gaussian mixture model parameters are iteratively updated using the expectation-maximization algorithm, the posterior probability of each data point belonging to each Gaussian distribution, i.e., the responsibility, is calculated based on the current Gaussian mixture model parameters, the mean, covariance, and mixing coefficients of each Gaussian distribution are updated based on the calculated posterior probabilities, and the iteration converges until the Gaussian mixture model parameters converge, specifically including: calculating the posterior probability (responsibility) of each data point belonging to each Gaussian distribution based on the current Gaussian mixture model parameters, which involves using the probability density function of the Gaussian distribution to calculate the likelihood of the data point under each distribution, and weighting according to the mixing coefficient; based on the calculated posterior probability, update the mean, covariance and mixing coefficient of each Gaussian distribution, these updates are completed by maximizing the expected likelihood function of the data, where the update of the mean and covariance involves weighted average and weighted covariance calculation, while the update of the mixing coefficient is realized by normalizing the sum of the posterior probabilities, repeat the above steps until the parameters of the Gaussian mixture model converge, the convergence can be judged by monitoring the size of the parameter change or setting the maximum number of iterations.

[0107] The fitting effect of the Gaussian mixture model is evaluated using the silhouette coefficient to obtain an evaluation result, specifically including: using the silhouette coefficient to quantify the fitting effect of the Gaussian mixture model, the silhouette coefficient measures the clustering quality of the model by comparing the distance of the data points with other points in their own category and the distance with points of other categories, calculates the silhouette coefficient, and evaluates the performance of the model according to its value. Higher silhouette coefficient value usually means better fitting effect.

[0108] According to the evaluation result, the corresponding Gaussian mixture model is determined as the Gaussian mixture model representing the "normal" state, specifically including: selecting the Gaussian mixture model with the best performance as the model representing the "normal" state according to the evaluation result, which can be done by comparing the silhouette coefficients of different models (with different number of Gaussian distributions or initialization parameters), saving the selected Gaussian mixture model for subsequent tasks such as anomaly detection, data generation, etc.

[0109] In the embodiments of the present application, the Gaussian mixture model can flexibly fit complex data distribution. Since it uses the weighted sum of multiple Gaussian distributions to represent the data, it can capture the multi-modal characteristics that a single Gaussian distribution cannot describe, which makes the model more accurately represent the true distribution of "normal" state data. By establishing a Gaussian mixture model representing the "normal" state, it is easy to detect abnormal data that deviates from the normal mode, which is crucial for many application scenarios such as fault detection, security monitoring, etc. because timely identification of abnormalities helps to prevent potential problems or risks. The expectation maximization (EM) algorithm is used to iteratively update the parameters of the Gaussian mixture model, which is a robust parameter estimation method. Even in the presence of noise or missing data, the EM algorithm can effectively estimate the model parameters, ensuring the robustness of the model. The profile coefficient and other evaluation indicators are used to quantify the fitting effect of the Gaussian mixture model, which provides an objective basis for model selection and optimization. By comparing the profile coefficients of different models, the model that best represents the "normal" state can be selected, thereby improving the accuracy of subsequent analysis. Each Gaussian distribution in the Gaussian mixture model can correspond to a certain group or mode in the data, which enhances the interpretability of the model. By examining the parameters (such as mean and covariance) of each Gaussian distribution, different characteristics and patterns of the "normal" state can be understood in depth.

[0110] In a preferred embodiment of the present application, the calculation formula of the dynamic factor is:

[0111]

[0112] where A(t) represents the dynamic adjustment factor; α represents the adjustment coefficient; I i (t) and I i (t+1) represent the gray value of the i-th pixel in image I(t) and image I(t+1) respectively; N is the total number of pixels in the image; β represents the adjustment coefficient; v(t) represents the wind speed at time t; κ is a constant, representing the magnification multiple of wind speed on image difference; γ represents the adjustment coefficient; b i (t) is the actual value of the i-th pixel in image b(t); is the denoising estimated value of the i-th pixel in image b(t); i represents the index value.

[0113] In the embodiments of the present application, A(t) can be dynamically adjusted according to actual conditions, which means it can maintain effectiveness in different environments and conditions, for example, in a scene where wind speed changes greatly, A(t) can be adjusted accordingly to reflect such changes, thereby improving the adaptability and robustness of the overall system. By calculating the change of pixel gray value between adjacent image frames, A(t) can accurately capture the dynamic information in the image, which is crucial for application scenarios that require real-time monitoring of image changes (such as video monitoring, motion analysis, etc.), as it provides a direct and quantitative indicator of image content changes. Incorporating wind speed data into the calculation makes A(t) not only dependent on image information, but also considers environmental factors, which enhances the representation ability of A(t), especially in scenarios where environmental factors have a significant impact on image quality (such as outdoor monitoring, remote sensing imaging, etc.). By introducing the difference term between the actual value and the denoising estimated value of the pixel in the image, A(t) can suppress the influence of noise to a certain extent, which helps to improve the stability and accuracy of the system in a noisy environment, especially in image processing and analysis tasks, where noise is usually a non-negligible problem.

[0114] In a preferred embodiment of the present application, the dynamic factor is used to adjust the abnormal value of the environment around the power transmission line to obtain a modified index, comprising:

[0115] The dynamic factor is multiplied by the corresponding difference value to obtain the modified index.

[0116] In the embodiments of the present application, by using the dynamic factor to adjust the abnormal value, the actual situation of the environment around the power transmission line can be more accurately reflected. The dynamic factor considers multiple factors, including image changes, wind speed, etc., so as to more accurately correct the abnormal value in the original data, making the modified index closer to the true situation. In the monitoring of power transmission lines, changes in the surrounding environment may cause data fluctuations, which in turn affect the accurate judgment of the line state. By introducing the dynamic factor to adjust the abnormal value, the fluctuations can be reduced, the stability of the monitoring data can be enhanced, and the reliability of the system for evaluating the state of the power transmission line can be improved. Since the dynamic factor can reflect the changes of environmental factors (such as wind speed, image difference, etc.) in real time, the abnormal value adjustment method based on the dynamic factor can respond more timely to the changes in the surrounding environment. The modified index can provide more accurate data support for the operation and management of the power transmission line, and by dynamically adjusting the abnormal value, the operation and maintenance personnel can make decisions based on more realistic data, thereby optimizing resource allocation and improving operation efficiency. The introduction of the dynamic factor to adjust the abnormal value is part of the intelligent monitoring system, and the application of this method helps to improve the intelligent level of the entire monitoring system, enabling it to more autonomously adapt to environmental changes and improve the accuracy of early warning and response.

[0117] In a preferred embodiment of the present application, the correction index is analyzed to determine the type, location and severity of the hidden danger, including:

[0118] The hidden danger analysis module first receives the correction index, and pre-sets the correction index threshold corresponding to different hidden danger types. These thresholds are based on historical data and are used to distinguish normal changes from potential hidden dangers. Each received correction index is compared with the threshold corresponding to the hidden danger type. If a correction index exceeds the threshold corresponding to the hidden danger type, the index is marked as abnormal and the corresponding hidden danger type is recorded. The image difference recognition module has output the image frame containing the abnormal area or the coordinate information of the abnormal area. If an image frame is output, image processing techniques such as edge detection and color segmentation are used to automatically identify and locate the abnormal area and extract its coordinate information. If the coordinate information of the abnormal area is directly output, the coordinates are directly used.

[0119] The geographic location (such as latitude and longitude) of the video monitoring device and its shooting range are determined. The shooting range may be affected by factors such as camera direction, focal length, lens type, etc. Based on this information, a mapping table can be created to map the pixel coordinates (x, y) in the image to the geographic coordinates (longitude, latitude) in the real world. For example, by placing reference points with known geographic locations within the camera's field of view, the mapping relationship can be calibrated and adjusted. Once the mapping relationship is established, it is used to convert the pixel coordinates of the abnormal area detected in the image to actual geographic coordinates, which usually involves inserting the pixel coordinates into a mapping function or looking up a mapping table to output the corresponding latitude and longitude coordinates. Combined with the converted geographic coordinates and the hidden danger type determined by the correction index, the specific location of the hidden danger in the real world can now be determined. This location is represented as a point. Next, the correction index is used to assess the severity of the hidden danger. The correction index is a quantitative indicator that reflects the degree of abnormality detected by the image difference recognition module. Generally, the higher the correction index, the more obvious the abnormality and the more serious the hidden danger. Different hidden danger levels (such as low, medium and high) can be set according to the size of the correction index to trigger different levels of response or alarms.

[0120] In a preferred embodiment of the present application, according to the type, location and severity of the hidden danger, a warning information is generated and real-time prompts the operation and maintenance personnel through the user interface, including:

[0121] The hidden danger type, location and severity determined in the previous step are obtained. These information are the basis for generating warning information, ensuring that the warning content is accurate and specific. According to the summary of hidden danger information, the system automatically generates warning information. The warning information should contain the following key elements:

[0122] Hidden danger type, indicating which type of hidden danger is detected, such as tree collapse, foreign object intrusion, fire, etc.

[0123] Location description, providing the specific location of the hidden danger, which can include latitude and longitude coordinates, relative position description (such as "500 meters away from a certain transformer station"), or a marked point on the map.

[0124] Severity, indicating the severity of the hidden danger, which can use qualitative description (such as "minor" or "serious") or quantitative index (such as the specific value of the correction index).

[0125] Suggested measures, according to the type and severity of the hidden danger, provide preliminary suggested measures to guide the operation and maintenance personnel to respond quickly.

[0126] The format of the early warning information can be adjusted according to actual needs to ensure that the information is clear, easy to understand, and can quickly convey key information.

[0127] After generating the early warning information, the system needs to prompt the operation and maintenance personnel in real time through the user interface. This can be achieved in the following ways:

[0128] Interface pop-up, a window containing early warning information is popped up in the user interface to ensure that the operation and maintenance personnel can immediately notice when using the system.

[0129] Sound or vibration reminder, if the system supports, it can remind the operation and maintenance personnel to pay attention to new early warning information through sound or vibration.

[0130] Mobile application push, if the system integrates a mobile application, it can send early warning information to the operation and maintenance personnel's mobile devices through the application push function.

[0131] Email or SMS notification, as an alternative, the system can also send early warning information to the operation and maintenance personnel through email or SMS to ensure that they can receive timely notifications even if they are not in front of the system.

[0132] After receiving the early warning information, the operation and maintenance personnel should take appropriate action according to the content of the information and the suggested measures. They may need to go to the scene for further inspection, start the emergency response process, or coordinate with other relevant departments. Through this step, the power transmission line surrounding hidden danger dynamic detection system based on image difference recognition can achieve rapid discovery and effective disposal of potential hidden dangers, thereby improving the safety and stability of the power transmission line.

[0133] As shown in Figure 2 , a power transmission line surrounding hidden danger dynamic detection method based on image difference recognition includes:

[0134] Collect video and image data around the power transmission line, the data is derived from the video monitoring device arranged around the power transmission line;

[0135] Preprocess the collected video and image data to obtain preprocessed video and image data;

[0136] Differential recognition is performed on the preprocessed image by using a machine learning algorithm, the abnormal value of the environment around the power transmission line is detected by comparing the images of consecutive frames, and the abnormal value of the environment around the power transmission line is adjusted according to a dynamic factor to obtain a correction index;

[0137] The correction index is analyzed to determine the type, location and severity of the hidden danger;

[0138] According to the type, location and severity of the hidden danger, a warning information is generated, and an operation and maintenance personnel is prompted in real time through a user interface.

[0139] It should be noted that the system is a system corresponding to the above method, and all implementation manners in the above method embodiment are applicable to this embodiment, and the same technical effects can be achieved.

[0140] Embodiments of the application also provide a computing device, comprising a processor and a memory storing a computer program, when the computer program is run by the processor, the method as described above is executed. All implementation manners in the above method embodiment are applicable to this embodiment, and the same technical effects can be achieved.

[0141] Embodiments of the application also provide a computer readable storage medium, storing instructions, when the instructions are run on a computer, the computer executes the method as described above. All implementation manners in the above method embodiment are applicable to this embodiment, and the same technical effects can be achieved.

[0142] The above is the preferred embodiment of the application, it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A power transmission line periphery hidden danger dynamic detection system based on image difference recognition, characterized in that, The application relates to a power transmission line surrounding environment monitoring method and device. The data acquisition module is used for acquiring video and image data of the power transmission line surrounding environment, and the data is obtained from video monitoring equipment arranged around the power transmission line, and the data acquisition module comprises the following steps: A plurality of grey wolf individuals are randomly initialized in the arrangement space of the power transmission line surrounding environment video monitoring equipment, that is, candidate video monitoring equipment positions; each grey wolf individual represents a solution and comprises the position, orientation and focal length parameters of the video monitoring equipment; A target function is defined, which is used for evaluating the pros and cons of the grey wolf individuals, that is, the data acquisition quality of the video monitoring equipment under the specific position and parameter configuration; the calculation formula of the target function is: ； where F(x) represents the objective function; 、 and are weight coefficients; represents the image sharpness, which is measured by calculating the sum of the gradients of each pixel in the image in the x and y directions; represents the coverage of the monitoring range, where A is the overlapping area of the monitoring device's field of view and the area to be monitored, and T is the total area of the area to be monitored; represents the unweighted target visibility, where P is the number of pixels of the target in the image, S is the total number of pixels in the image, C is the average contrast of the target, M is the maximum contrast, and d is the distance between the monitoring device and the key target, is set as the center position of the key target, is the standard deviation of the Gaussian function; is a summation symbol, indicating summation over all pixel points (i, j) in the image, where i and j represent the horizontal and vertical coordinates of the pixel points, respectively; is the image gradient in the x direction at pixel point (i, j); is the image gradient in the y direction at pixel point (i, j); The grey wolf individuals update their positions according to the current final solution, simulate the hunting behavior of the grey wolves, search for corresponding solutions in the solution space, and gradually approach the current corresponding solution; when the grey wolf individuals approach the final solution, the search is performed to find the global final solution; according to the output result of the grey wolf optimization algorithm, the position and parameter configuration of the video monitoring equipment are adjusted; Actual video and image data acquisition is performed under the updated video monitoring equipment configuration to obtain the video and image data of the power transmission line surrounding environment; The image preprocessing module is used for preprocessing the acquired video and image data to obtain preprocessed video and image data; The image difference recognition module is used for recognizing the differences of the preprocessed images by using a machine learning algorithm, detecting the abnormal values of the power transmission line surrounding environment by comparing the continuous image frames, adjusting the abnormal values of the power transmission line surrounding environment according to dynamic factors to obtain a correction index, and the calculation formula of the dynamic factor is: ； wherein, denotes a dynamic adjustment factor; denotes an adjustment coefficient; and denote the gray value of the i-th pixel in the image and the image respectively; N is the total number of pixels in the image; β denotes an adjustment coefficient; denotes the wind speed at time ; is a constant, denoting the magnification of the wind speed on the image difference; denotes an adjustment coefficient; is the actual value of the i-th pixel in the image ; is the de-noised estimated value of the i-th pixel in the image ; denotes an index value; The hidden danger analysis module is used for analyzing the correction index to determine the type, position and severity of the hidden danger; The early warning prompt module is used for generating early warning information according to the type, position and severity of the hidden danger and real-time prompting the operation and maintenance personnel through a user interface.

2. The image difference recognition-based power transmission line periphery hidden danger dynamic detection system according to claim 1, characterized in that, The acquired video and image data are preprocessed to obtain preprocessed video and image data, which comprises the following steps: A plurality of particles are randomly initialized in the space of the preprocessing parameters, that is, candidate preprocessing parameter combinations; each particle represents a set of preprocessing parameters, including the type, size and threshold value of the filter; A fitness function is defined to evaluate the quality of the preprocessed video and image data; The speed of each particle is updated according to the historical final position of the individual and the historical final position of the group, and the position of each particle, that is, the preprocessing parameter, is adjusted according to the updated speed; The preprocessing operation is performed on each new position, and the preprocessing effect is evaluated by using the fitness function; the speed updating, position updating and evaluation updating steps are repeatedly executed until a preset iteration number is reached, so as to obtain the preprocessing parameter combination corresponding to the historical final position of the group; The preprocessing operation is performed on the acquired video and image data by using the preprocessing parameter combination corresponding to the final position, so as to obtain the preprocessed video and image data.

3. The image difference recognition-based power transmission line periphery hidden danger dynamic detection system according to claim 2, characterized in that, The preprocessed images are recognized by using a machine learning algorithm, and the abnormal values of the power transmission line surrounding environment are detected by comparing the continuous image frames, which comprises the following steps: Continuous image frames are extracted from the video stream according to a set frame rate; extracting features from each frame of image to obtain a feature vector representing the content of the image; determining the number of initial frames for establishing a Gaussian mixture model, inputting the feature vectors of the initial frames into the Gaussian mixture model as training data, and learning the distribution of the feature vectors by the Gaussian mixture model algorithm to establish a Gaussian mixture model representing the normal state; setting parameters of the Gaussian mixture model, including the number of Gaussian distributions and the type of covariance matrix; for each new frame of image in the video stream, extracting its feature vector, inputting the feature vector of the new frame of image into the established Gaussian mixture model, and obtaining a difference value of the new frame of image from the Gaussian mixture model by calculating the Mahalanobis distance between the feature vector of the new frame of image and each Gaussian distribution in the Gaussian mixture model.

4. The image difference recognition-based power transmission line periphery hidden danger dynamic detection system according to claim 3, characterized in that, extracting features from each frame of image to obtain a feature vector representing the content of the image, including: receiving pre-processed single-frame image data as input, initializing the SIFT feature detection algorithm, including setting the number of layers and scale factors for scale space extreme value detection; constructing the scale space of the image by Gaussian blurring and down-sampling the input image at different scales to find stable feature points at different scales; finding local extreme points in the scale space by comparing the pixel values of each point with adjacent scales and adjacent positions, and the extreme points are considered as candidate key points; screening and positioning the candidate key points, and assigning a main direction to each key point by calculating the gradient direction and amplitude of the pixels around the key point; generating a fixed-length descriptor vector according to the assigned main direction by counting the gradient information of the pixels in the neighborhood of the key point; organizing the descriptor vectors of all key points into a feature vector set, and the feature vector set represents the content of the entire image.

5. The image difference recognition-based power transmission line periphery hidden danger dynamic detection system according to claim 4, characterized in that, learning the distribution of the feature vectors by the Gaussian mixture model algorithm to establish a Gaussian mixture model representing the normal state, including: collecting feature vector data representing the normal state, determining the number of Gaussian distributions to be used, i.e. the number of components of the mixture model; initializing parameters for each Gaussian distribution, including mean vector, covariance matrix and mixing coefficient, i.e. the weight of each Gaussian distribution in the mixture model; iteratively updating the parameters of the Gaussian mixture model using the expectation maximization algorithm, calculating the posterior probability of each data point belonging to each Gaussian distribution, i.e. the responsibility, according to the current parameters of the Gaussian mixture model; updating the mean, covariance and mixing coefficient of each Gaussian distribution based on the posterior probability, and iteratively converging until the parameters of the Gaussian mixture model converge; evaluating the fitting effect of the Gaussian mixture model using the profile coefficient to obtain an evaluation result; determining the corresponding Gaussian mixture model as the Gaussian mixture model representing the normal state according to the evaluation result.

6. The image difference recognition-based power transmission line periphery hidden danger dynamic detection system according to claim 5, characterized in that, adjusting the abnormal values of the environment around the transmission line according to the dynamic factor to obtain a correction index, including: multiplying the dynamic factor by the corresponding difference value to obtain the correction index.

7. A power transmission line periphery hidden danger dynamic detection method based on image difference recognition, characterized in that, The method is used to execute the system as claimed in any one of claims 1 to 6, and the method comprises the following steps: collecting video and image data around the transmission line, and the data is derived from video monitoring equipment arranged around the transmission line; Preprocessing the collected video and image data to obtain preprocessed video and image data; Differential recognition is performed on the preprocessed image by using a machine learning algorithm, and by comparing the images of consecutive frames, the abnormal values of the environment around the power transmission line are detected; and the abnormal values of the environment around the power transmission line are adjusted according to a dynamic factor to obtain a correction index; The correction index is analyzed to determine the type, location and severity of the hidden danger; According to the type, location and severity of the hidden danger, a warning information is generated, and an operation and maintenance personnel is prompted in real time through a user interface.

8. A computing device, comprising: Comprise: One or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method as claimed in claim 7.

Images