Distribution box potential safety hazard detection method and system

Abstract

The invention provides a distribution box potential safety hazard detection method and system, and relates to the technical field of distribution box safety detection, and the method comprises the steps: obtaining original image data, a temperature monitoring signal and a vibration monitoring signal in a target distribution box; performing image enhancement on the original image data to obtain a preprocessed image; respectively carrying out signal denoising on the temperature monitoring signal and the vibration monitoring signal to obtain a preprocessed signal; performing feature extraction on the preprocessed image and the preprocessed signal to obtain an image hidden danger feature and a signal hidden danger feature; carrying out feature fusion on the image hidden danger features and the signal hidden danger features to obtain fusion features; and inputting the fusion feature into a preset potential safety hazard detection model to obtain a detection result. According to the method, through image enhancement and signal denoising, the quality of input data can be effectively improved, and the model can be identified by using clearer and more reliable features. Therefore, false alarms and missing alarms can be reduced, and the overall accuracy of hidden danger detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution box safety detection, and in particular to a method and system for detecting potential safety hazards in a power distribution box. Background Art

[0002] With the continuous development of industrial automation and intelligence, the safety of distribution boxes, as an important part of the power system, is directly related to the stable operation of the entire power system and the safety of life and property of personnel. Traditional distribution box hidden danger detection mainly relies on manual inspection, which is not only time-consuming and labor-intensive, but also easily affected by human factors, resulting in inaccurate and in-time detection results. In recent years, with the rapid development of deep learning technology, automated hidden danger detection methods based on image recognition have gradually become a research hotspot.

[0003] The invention patent with application number 202411072074.6 discloses a method, device, equipment and storage medium for identifying hidden dangers in distribution boxes. Model training is performed based on the original image data inside multiple distribution boxes with different safety levels to obtain a trained deep learning model; the trained deep learning model is used to identify in real time whether there are hidden dangers in the image stream of the distribution box taken from the industrial site, and the hidden danger identification result of the distribution box is obtained; the hidden danger identification result of the distribution box is displayed, and the hidden danger identification result of the distribution box includes hidden danger features, position marks of hidden danger features in the corresponding image, and corresponding alarm information of hidden danger features. In this patent, the distribution box is in a complex environment of the industrial site, and it is difficult to collect original image data, resulting in poor image quality, which affects the recognition rate of the model, such as uneven lighting, occlusion, or angle problems. And only using image data for detection, the model has false alarms (misjudging normal situations as hidden dangers) or missed alarms (failure to identify actual hidden dangers) in an environment with high safety requirements. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method and system for detecting safety hazards in distribution boxes, which has significant beneficial effects in improving detection accuracy, real-time monitoring, comprehensive utilization of multimodal data, and enhancing safety monitoring capabilities, and provides effective technical support for the safety management of electrical facilities.

[0005] To achieve the above object, the present invention provides the following solutions: A method for detecting potential safety hazards in a distribution box, comprising: Acquire original image data inside the target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; Performing image enhancement on the original image data to obtain a preprocessed image; Performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain preprocessed signals; Performing feature extraction on the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; Fusing the image hidden danger feature and the signal hidden danger feature to obtain a fused feature; The fusion features are input into a preset safety hazard detection model to obtain a detection result.

[0006] Preferably, performing image enhancement on the original image data to obtain a preprocessed image includes: Take a square window with each pixel of the original image data as the center, and detect the change of each pixel under the square window; When the change amount of the pixel point is greater than the preset threshold, the original image data in the corresponding square window is smoothed using the median value of the pixel point in the square window to obtain the smoothed pixel point; Continuously moving the square window until the smoothing process of the entire original image data is completed to obtain smoothed original image data; Constructing a pixel enhancement function based on the convolution value of the smoothed original image data; The pixel enhancement function is used to perform image enhancement processing on the original image data to obtain the preprocessed image.

[0007] Preferably, taking a square window with each pixel of the original image data as the center and detecting the change amount of each pixel under the square window includes: Using the formula: Detect the change of each pixel under the square window; where, Represents the gradient value of the original image data in the x direction, Represents the gradient value of the original image data in the y direction, Indicates that the original image data is in Gray value at position, r represents pixel point The amount of change.

[0008] Preferably, when the change amount of the pixel point is greater than a preset threshold, the original image data in the corresponding square window is smoothed using the median value of the pixel points in the square window to obtain the smoothed pixel points, including: Sort each pixel in the square window to form a pixel sequence; The median of the pixel sequence is taken as the output of the pixel point after the square window is smoothed; the output formula of the smoothed pixel point is: In the formula, is the smoothed pixel point, is the side length of the square window, is the pixel point on the original image data The value at position.

[0009] Preferably, constructing a pixel enhancement function based on the convolution value of the smoothed original image data comprises: Perform convolution processing on the smoothed original image data to obtain a convolved image; Determine the enhancement coefficient based on the pixel value of the convolved image; A pixel enhancement function is constructed based on the enhancement coefficient; wherein the pixel enhancement function is: in, represents the preprocessed image, represents the enhancement factor, Indicates that the image after convolution is The pixel value at represents the original image data after smoothing, Represents the mean square error between the smoothed original image data and the convolved image.

[0010] Preferably, the temperature monitoring signal and the vibration monitoring signal are respectively subjected to signal denoising to obtain preprocessed signals, including: Performing translation on the temperature monitoring signal and the vibration monitoring signal respectively to obtain translated signals; Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients; Determine the filtering threshold according to the decomposition scale and length of the signal; Constructing a denoising function according to the filtering threshold; The denoising function is used to remove the translated signal to obtain the preprocessed signal.

[0011] Preferably, the image hidden danger features include structural features, texture features, color features, shape features and space features; the signal hidden danger features include: temperature features, frequency domain features, time domain features and transient features.

[0012] Preferably, the image hidden danger feature and the signal hidden danger feature are subjected to feature fusion to obtain a fusion feature, including: Standardizing the image hidden danger features and the signal hidden danger features to obtain standardized image features and standardized signal features; The standardized image features and the standardized signal features are aligned, and the aligned features are weighted fused using an attention mechanism to obtain the fused features.

[0013] Preferably, the potential safety hazard detection model is a CNN-LSTM network architecture.

[0014] A distribution box safety hazard detection system, comprising: A data acquisition unit, used to acquire original image data inside a target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; An image processing unit, used for performing image enhancement on the original image data to obtain a preprocessed image; A signal processing unit, used for performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain a preprocessed signal; A feature extraction unit, used to extract features from the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; A feature fusion unit, used for fusing the image hidden danger feature and the signal hidden danger feature to obtain a fusion feature; The detection unit is used to input the fusion feature into a preset safety hazard detection model to obtain a detection result.

[0015] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: The present invention provides a method and system for detecting safety hazards in a distribution box, the method comprising: obtaining original image data inside a target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; performing image enhancement on the original image data to obtain a preprocessed image; performing signal denoising on the temperature monitoring signal and the vibration monitoring signal to obtain a preprocessed signal; performing feature extraction on the preprocessed image and the preprocessed signal to obtain image hidden danger features and signal hidden danger features; performing feature fusion on the image hidden danger features and the signal hidden danger features to obtain fusion features; inputting the fusion features into a preset safety hidden danger detection model to obtain a detection result. The present invention can effectively improve the quality of input data through image enhancement and signal denoising, so that the model can use clearer and more reliable features for identification. This helps to reduce false positives and false negatives, and improve the overall accuracy of hidden danger detection; and at the same time, combining image data and multimodal information such as temperature and vibration signals, it can fully reflect the operating status of the distribution box. This diverse data source provides rich contextual information, making hidden danger detection more comprehensive and in-depth. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0017] Figure 1 A flow chart of a method provided by an embodiment of the present invention; Figure 2 A schematic diagram of the system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The purpose of the present invention is to provide a method and system for detecting safety hazards in distribution boxes, which can effectively improve the quality of input data through image enhancement and signal denoising, so that the model can use clearer and more reliable features for identification.

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a method for detecting potential safety hazards in a distribution box, comprising: Step 100: Acquire original image data inside a target distribution box, and temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; Step 200: performing image enhancement on the original image data to obtain a preprocessed image; Step 300: performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain a preprocessed signal; Step 400: extracting features from the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; Step 500: Fusing the image hidden danger feature and the signal hidden danger feature to obtain a fused feature; Step 600: Input the fusion features into a preset safety hazard detection model to obtain a detection result.

[0022] Specifically, step 100 of this embodiment includes: 1. Equipment preparation Use a high-resolution industrial camera or thermal imaging camera that can capture clear images under different lighting conditions. Make sure the camera has the appropriate lens and focal length to clearly capture the details inside the distribution box. Install temperature sensors (such as thermocouples or thermistors) and vibration sensors (such as accelerometers) at key monitoring points in the distribution box. Make sure the sensor selection can meet the measurement range and accuracy requirements.

[0023] 2. When the distribution box is in normal operation, start the image acquisition device and take the original image of the inside of the distribution box regularly or on demand. In this embodiment, trigger conditions (such as when the temperature exceeds a certain threshold) can be set to automatically take images in order to capture potential hidden dangers. And the temperature changes inside the distribution box are monitored in real time through the temperature sensor. Connect the temperature sensor to the data acquisition system, read the temperature value regularly and record the timestamp to form a temperature monitoring signal. Use a vibration sensor to monitor the vibration conditions inside the distribution box, especially the vibration of key components (such as transformers, switches, etc.). Connect the vibration sensor to the data acquisition system to record the vibration signal in real time, including parameters such as acceleration and frequency.

[0024] 3. Data transmission and storage The collected image data is transmitted to the central processing unit (such as a computer or edge computing device) via USB, Wi-Fi, or Ethernet. The temperature and vibration signals are transmitted to the data acquisition system via serial port, I2C or other communication protocols. In the central processing unit, a database or file system is set up to store the collected image data, temperature monitoring signals, and vibration monitoring signals. Ensure the security and integrity of data storage for subsequent processing and analysis.

[0025] 4. Data preprocessing The acquired raw image data is preliminarily processed, such as denoising, contrast enhancement, etc., to improve the effect of subsequent analysis. The temperature and vibration signals are denoised to eliminate environmental interference and ensure the accuracy of the signal.

[0026] 5. Data synchronization and tagging In this embodiment, the timestamps of the image data, temperature monitoring signals, and vibration monitoring signals are ensured to be consistent so that information from different data sources can be associated in subsequent analysis. Data is marked according to the collected timestamps and events (such as device status changes, abnormal alarms, etc.) to facilitate subsequent feature extraction and analysis.

[0027] Through the above steps, this embodiment can effectively obtain the original image data, temperature monitoring signals and vibration monitoring signals inside the target distribution box, providing a reliable data basis for subsequent hidden danger detection.

[0028] Optionally, the selection of monitoring points is a key link in the detection of safety hazards in the distribution box, and the operating characteristics of the equipment, potential hazard areas and environmental factors need to be comprehensively considered. In this embodiment, first, key components inside the distribution box are selected as monitoring points, such as transformers, switches, wiring terminals and fuses, etc. These components are prone to overheating, vibration or poor contact during operation. Secondly, temperature monitoring points should be set at locations with large current or poor heat dissipation, such as the outlet terminal of the transformer and the contact point of the switch, so as to capture abnormal temperature rise in time. At the same time, vibration monitoring points should be selected near mechanical moving parts, such as fans, relays, etc., to monitor their operating status and potential faults. Specific locations can include: the upper left corner (near the transformer), the lower right corner (at the wiring terminal), and the middle position (near the switch and fuse) of the distribution box to ensure that all key components are covered and an effective monitoring network is formed. Through reasonable selection of monitoring points, comprehensive monitoring of the internal state of the distribution box can be achieved, and potential safety hazards can be discovered and handled in time.

[0029] Preferably, performing image enhancement on the original image data to obtain a preprocessed image includes: Take a square window with each pixel of the original image data as the center, and detect the change of each pixel under the square window; When the change amount of the pixel point is greater than the preset threshold, the original image data in the corresponding square window is smoothed using the median value of the pixel point in the square window to obtain the smoothed pixel point; Continuously moving the square window until the smoothing process of the entire original image data is completed to obtain smoothed original image data; Constructing a pixel enhancement function based on the convolution value of the smoothed original image data; The pixel enhancement function is used to perform image enhancement processing on the original image data to obtain the preprocessed image.

[0030] Preferably, taking a square window with each pixel of the original image data as the center and detecting the change amount of each pixel under the square window includes: Using the formula: Detect the change of each pixel under the square window; where, Represents the gradient value of the original image data in the x direction, Represents the gradient value of the original image data in the y direction, Indicates that the original image data is in Gray value at position, r represents pixel point The amount of change.

[0031] Preferably, when the change amount of the pixel point is greater than a preset threshold, the original image data in the corresponding square window is smoothed using the median value of the pixel points in the square window to obtain the smoothed pixel points, including: Sort each pixel in the square window to form a pixel sequence; The median of the pixel sequence is taken as the output of the pixel point after the square window is smoothed; the output formula of the smoothed pixel point is: In the formula, is the smoothed pixel point, is the side length of the square window, is the pixel point on the original image data The value at position.

[0032] Specifically, in practical applications, since noise is unavoidable during the image acquisition process, and the noise is generally very different from the surrounding pixels, the present application can find the noise on the image by detecting the change in each pixel under the square window, and at the same time use the median of the pixel sequence under the square window as the smoothed pixel point, which can better maintain the detail information of the image while removing the noise.

[0033] Preferably, constructing a pixel enhancement function based on the convolution value of the smoothed original image data comprises: Perform convolution processing on the smoothed original image data to obtain a convolved image; Determine the enhancement coefficient based on the pixel value of the convolved image; A pixel enhancement function is constructed based on the enhancement coefficient; wherein the pixel enhancement function is: in, represents the preprocessed image, represents the enhancement factor, Indicates that the image after convolution is The pixel value at represents the original image data after smoothing, Represents the mean square error between the smoothed original image data and the convolved image.

[0034] Optionally, the present invention can quantify the degree of difference between the two by calculating the mean square error. When the mean square error is large, it means that the original image and the image after convolution are very different and need to be enhanced; when the mean square error is small, it means that the original image and the image after convolution are small and no enhancement can be considered. In image enhancement, whether the image needs to be enhanced can be determined based on the size of the mean square error, so that the image can achieve a better visual effect.

[0035] Preferably, the temperature monitoring signal and the vibration monitoring signal are respectively subjected to signal denoising to obtain preprocessed signals, including: Performing translation on the temperature monitoring signal and the vibration monitoring signal respectively to obtain translated signals; Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients; Determine the filtering threshold according to the decomposition scale and length of the signal; Constructing a denoising function according to the filtering threshold; The denoising function is used to remove the translated signal to obtain the preprocessed signal.

[0036] Preferably, the image hidden danger features include structural features, texture features, color features, shape features and space features; the signal hidden danger features include: temperature features, frequency domain features, time domain features and transient features.

[0037] Specifically, in this embodiment, the process of extracting image hidden danger features and signal hidden danger features in the detection of safety hidden dangers in the distribution box is as follows: 1. Image Hidden Danger Feature Extraction 1. Structural feature extraction: Use edge detection algorithms (such as Canny edge detection) to extract edge information of components inside the distribution box. The process is as follows: Grayscale the preprocessed image.

[0038] Apply Gaussian filtering to reduce image noise.

[0039] Use the Canny algorithm to detect edges and obtain edge images.

[0040] The contour extraction algorithm (such as findContours in OpenCV) is used to identify the contours of each component and extract structural features.

[0041] 2. Texture feature extraction, using gray level co-occurrence matrix (GLCM) and local binary pattern (LBP) to extract texture features. The process is as follows: Calculate the gray-level co-occurrence matrix of the image and analyze the spatial relationship of pixel gray values.

[0042] Extract features such as contrast, correlation, energy, and homogeneity from the co-occurrence matrix.

[0043] The LBP algorithm is used to describe the local texture of the image and generate a texture feature vector.

[0044] 3. Color feature extraction: Use color histogram and color space conversion to extract color features. The process is as follows: Convert an image from RGB color space to HSV or Lab color space to better capture color information.

[0045] Calculate the color histogram and count the pixel distribution of different color channels.

[0046] Extract the statistical features of the histogram (such as mean, variance, etc.) as color features.

[0047] 4. Shape feature extraction: Use shape descriptors (such as Hu moments) to extract shape features. The process is as follows: The features of the extracted contours are calculated and the shape is described using invariant features such as Hu moments.

[0048] Calculate the shape's features such as perimeter, area, and shape ratio to form a shape feature vector.

[0049] 5. Spatial feature extraction: Use spatial relationship analysis to extract spatial features. The process is as follows: Calculate the distance and relative position relationship between components.

[0050] Extract the spatial distribution characteristics of components, such as density, distribution uniformity, etc.

[0051] 2. Signal Hidden Danger Feature Extraction 1. Temperature feature extraction: use statistical analysis to extract temperature features. The process is as follows: Collect time series data of temperature monitoring signals.

[0052] Calculate the mean, maximum, minimum, standard deviation and other statistical characteristics of temperature.

[0053] Analyze temperature change trends and extract dynamic features such as temperature change rate.

[0054] 2. Frequency domain feature extraction: Use Fourier transform to extract frequency domain features. The process is as follows: Fast Fourier transform (FFT) is performed on the temperature and vibration signals to convert the time domain signals into frequency domain signals.

[0055] Calculate spectrum characteristics, such as dominant frequency, spectrum energy, frequency distribution, etc.

[0056] The statistics of frequency domain features (such as mean, variance, etc.) are extracted as frequency domain features.

[0057] 3. Time domain feature extraction: Use time domain analysis to extract time domain features. The process is as follows: Collect time domain data of temperature and vibration signals.

[0058] Calculate the time domain characteristics of the signal such as mean, variance, peak, crest and trough.

[0059] Analyze the autocorrelation of the signal and extract the correlation index of the time domain features.

[0060] 4. Transient feature extraction: Use wavelet transform to extract transient features. The process is as follows: Wavelet transform is performed on temperature and vibration signals to capture the changes of transient signals.

[0061] Extract wavelet coefficients and analyze the characteristics of transient signals, such as the amplitude and duration of sudden changes.

[0062] Calculate the statistics of transient features to form a transient feature vector.

[0063] Through the above scheme, this embodiment can effectively extract hidden danger features from images and signals, and provide reliable data support for subsequent safety hidden danger detection.

[0064] Preferably, the image hidden danger feature and the signal hidden danger feature are subjected to feature fusion to obtain a fusion feature, including: Standardizing the image hidden danger features and the signal hidden danger features to obtain standardized image features and standardized signal features; The standardized image features and the standardized signal features are aligned, and the aligned features are weighted fused using an attention mechanism to obtain the fused features.

[0065] Specifically, in the feature fusion process, this embodiment first standardizes the extracted image hidden danger features and signal hidden danger features to eliminate the dimensional differences between different features and ensure that they are compared on the same scale. Specifically, using the Z-score standardization method, each feature is subtracted from its mean and divided by the standard deviation to obtain standardized image features and standardized signal features. Next, feature alignment is performed, and an autoencoder or other alignment algorithm is used to map the standardized image features and standardized signal features to the same feature space for subsequent fusion processing. After alignment, the attention mechanism is used to calculate the correlation between the image features and the signal features, and the attention weight is generated. By applying the attention weight to the aligned features and performing weighted summation, the fused features are obtained. This process not only retains the important information of each feature, but also dynamically adjusts the contribution of different features through the attention mechanism, thereby improving the expressiveness and effectiveness of the fused features, and providing a more comprehensive and accurate feature representation for subsequent hidden danger detection.

[0066] Specifically, the calculation process of the attention weight in this embodiment is as follows: First set: I is the standardized image feature, and its dimension is .

[0067] is the standardized signal feature, the dimension is .

[0068] and is a learnable weight matrix, which is used for linear transformation of image features and signal features respectively, and its dimension is and .

[0069] is the attention weight, and its dimension is .

[0070] The calculation steps are as follows 1. Feature transformation: linear transformation of image features and signal features:

[0071] 2. Calculate the correlation and calculate the correlation between the transformed image features and signal features:

[0072] in, is the correlation matrix, which represents the similarity between image features and signal features.

[0073] 3. Generate attention weights. Generate attention weights by normalizing the correlation matrix:

[0074] Furthermore, this embodiment enhances the ability to capture nonlinear relationships between features through linear transformation and tanh activation function. The generation of attention weights depends on the relationship between image features and signal features, so that the model can dynamically adjust the importance of features according to different input data. The weight matrix is ​​a learnable parameter that can be optimized through back propagation to improve the performance of the model. Through the above formulas and steps, this embodiment can effectively calculate the attention weights between image features and signal features, providing strong support for feature fusion.

[0075] Preferably, the potential safety hazard detection model is a CNN-LSTM network architecture.

[0076] Optionally, the model structure of the fusion feature input of this embodiment is a CNN-LSTM network architecture, which is intended to effectively process the combination of image features and time series signal features. Specifically, first, the input fusion features are processed by multiple convolutional layers (CNN), which can effectively capture the key patterns of the image and time series signals inside the distribution box by extracting local features and spatial information. The pooling layer after the convolution layer reduces the feature dimension while retaining important information, thereby improving computational efficiency. Next, the extracted features are converted into a one-dimensional array through a flattening layer (Flatten) and input into a long short-term memory network (LSTM) layer to capture the time series features of the signal data. The recursive structure of LSTM can effectively learn the dynamic pattern of feature changes over time, thereby analyzing long-term and short-term dependencies. In this process, the fusion features are further processed by a fully connected layer (Fully Connected Layer) and an activation function, and finally a classification output is generated to represent the safety status of the distribution box. The output layer usually uses a Softmax activation function to convert the output of the model into a probability distribution to determine the possibility of potential safety hazards. The entire model is trained through the back-propagation algorithm, and the parameters are optimized according to the real labels (such as whether there are safety hazards) to improve the accuracy and robustness of the detection results. Through this structure, the model can comprehensively analyze the image and time series features to achieve effective detection of potential safety hazards in distribution boxes.

[0077] Corresponding to the above method, this embodiment also provides a distribution box safety hazard detection system, such as Figure 2 As shown, including: A data acquisition unit, used to acquire original image data inside a target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; An image processing unit, used for performing image enhancement on the original image data to obtain a preprocessed image; A signal processing unit, used for performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain a preprocessed signal; A feature extraction unit, used to extract features from the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; A feature fusion unit, used for fusing the image hidden danger feature and the signal hidden danger feature to obtain a fusion feature; The detection unit is used to input the fusion feature into a preset safety hazard detection model to obtain a detection result.

[0078] The beneficial effects of the present invention are as follows: (1) The present invention can effectively improve the quality of input data through image enhancement and signal denoising, enabling the model to use clearer and more reliable features for identification. This helps reduce false positives and false negatives and improves the overall accuracy of hidden danger detection.

[0079] (2) The present invention combines image data with multimodal information such as temperature and vibration signals to fully reflect the operating status of the distribution box. This diverse data source provides rich contextual information, making hidden danger detection more comprehensive and in-depth.

[0080] (3) The present invention can monitor and analyze the internal status of the distribution box in real time and quickly identify potential hidden dangers. This is particularly important in power systems, as timely measures can be taken to prevent accidents, thereby enhancing the safety of equipment and personnel.

[0081] (4) The present invention reduces the reliance on manual inspection and the impact of human factors on inspection results through automated hidden danger detection. This can improve work efficiency and reduce the burden of manual inspection and analysis.

[0082] (5) The present invention uses a deep learning model for hidden danger detection, which has good scalability. The model can be continuously updated and optimized according to actual needs to adapt to different application scenarios and hidden danger types.

[0083] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0084] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for detecting safety hazards of a distribution box, characterized in that: include: Acquire original image data inside the target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; Performing image enhancement on the original image data to obtain a preprocessed image; Performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain preprocessed signals; Performing feature extraction on the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; Fusing the image hidden danger feature and the signal hidden danger feature to obtain a fused feature; The fusion features are input into a preset safety hazard detection model to obtain a detection result.

2. The method for detecting potential safety hazards of a distribution box according to claim 1, characterized in that: Performing image enhancement on the original image data to obtain a preprocessed image includes: Take a square window with each pixel of the original image data as the center, and detect the change of each pixel under the square window; When the change amount of the pixel point is greater than the preset threshold, the original image data in the corresponding square window is smoothed using the median value of the pixel point in the square window to obtain the smoothed pixel point; Continuously moving the square window until the smoothing process of the entire original image data is completed to obtain smoothed original image data; Constructing a pixel enhancement function based on the convolution value of the smoothed original image data; The pixel enhancement function is used to perform image enhancement processing on the original image data to obtain the preprocessed image.

3. The method for detecting potential safety hazards of a distribution box according to claim 2, characterized in that: Take a square window with each pixel of the original image data as the center, and detect the change of each pixel under the square window, including: Using the formula: Detect the change of each pixel under the square window; where, Represents the gradient value of the original image data in the x direction, Represents the gradient value of the original image data in the y direction, Indicates that the original image data is in Gray value at position, r represents pixel point The amount of change.

4. The method for detecting potential safety hazards of a distribution box according to claim 3, characterized in that: When the change in pixel points is greater than a preset threshold, the original image data in the corresponding square window is smoothed using the median of the pixel points in the square window to obtain smoothed pixel points, including: Sort each pixel in the square window to form a pixel sequence; The median of the pixel sequence is taken as the output of the pixel point after the square window is smoothed; the output formula of the smoothed pixel point is: In the formula, is the smoothed pixel point, is the side length of the square window, is the pixel point on the original image data The value at position.

5. The method for detecting potential safety hazards of a distribution box according to claim 4, characterized in that: A pixel enhancement function is constructed based on the convolution value of the smoothed original image data, including: Perform convolution processing on the smoothed original image data to obtain a convolved image; Determine the enhancement coefficient based on the pixel value of the convolved image; A pixel enhancement function is constructed based on the enhancement coefficient; wherein the pixel enhancement function is: in, represents the preprocessed image, represents the enhancement factor, Indicates that the image after convolution is The pixel value at represents the original image data after smoothing, Represents the mean square error between the smoothed original image data and the convolved image.

6. The method for detecting potential safety hazards of a distribution box according to claim 1, characterized in that: The temperature monitoring signal and the vibration monitoring signal are respectively subjected to signal denoising to obtain preprocessed signals, including: Performing translation on the temperature monitoring signal and the vibration monitoring signal respectively to obtain translated signals; Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients; Determine the filtering threshold according to the decomposition scale and length of the signal; Constructing a denoising function according to the filtering threshold; The denoising function is used to remove the translated signal to obtain the preprocessed signal.

7. The method for detecting potential safety hazards in a distribution box according to claim 1, characterized in that: The image hidden danger features include structural features, texture features, color features, shape features and spatial features; The signal hidden danger characteristics include: temperature characteristics, frequency domain characteristics, time domain characteristics and transient characteristics.

8. The method for detecting safety hazards of distribution boxes according to claim 1, characterized in that: The image hidden danger feature and the signal hidden danger feature are subjected to feature fusion to obtain a fusion feature, including: Standardizing the image hidden danger features and the signal hidden danger features to obtain standardized image features and standardized signal features; The standardized image features and the standardized signal features are aligned, and the aligned features are weighted fused using an attention mechanism to obtain the fused features.

9. The method for detecting safety hazards of distribution boxes according to claim 1, characterized in that: The potential safety hazard detection model is a CNN-LSTM network architecture.

10. A distribution box safety hazard detection system, comprising: A data acquisition unit, used to acquire original image data inside a target distribution box, temperature monitoring signals and vibration monitoring signals at monitoring points inside the target distribution box; An image processing unit, used for performing image enhancement on the original image data to obtain a preprocessed image; A signal processing unit, used for performing signal denoising on the temperature monitoring signal and the vibration monitoring signal respectively to obtain a preprocessed signal; A feature extraction unit, used to extract features from the preprocessed image and the preprocessed signal respectively to obtain image hidden danger features and signal hidden danger features; A feature fusion unit, used for fusing the image hidden danger feature and the signal hidden danger feature to obtain a fusion feature; The detection unit is used to input the fusion feature into a preset safety hazard detection model to obtain a detection result.

Images