A method and system for analyzing environmental protection equipment operation and maintenance behavior based on image analysis

Abstract

The present invention discloses a method and system for analyzing the operation and maintenance behavior of environmental protection equipment based on image analysis, and specifically relates to the technical field of equipment operation and maintenance behavior analysis. By deploying light sensors and camera IMU sensors at different positions of wind turbines, light intensity and camera vibration data are obtained, and through feature extraction, abnormal jitter characteristics of light rate changes and camera vibrations are extracted, and these features are input into a machine learning model to judge the interference degree of the image and classify it as a high-quality or low-quality image. For low-quality images, an optimization method is used to improve their quality, and the optimization effect is evaluated for detection accuracy to ensure that the image meets predetermined accuracy requirements. The operation and maintenance behavior of environmental protection equipment is analyzed through the optimized images, which can accurately identify whether the equipment has abnormal conditions, provide scientific decision support for subsequent repairs and maintenance, and thus effectively improve the operation and maintenance efficiency and reliability of wind turbines.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment operation and maintenance behavior analysis, and in particular to an environmental protection equipment operation and maintenance behavior analysis method and system based on image analysis. Background Art

[0002] Image-based analysis of environmental protection equipment operation and maintenance behavior uses image recognition and analysis technology to monitor and evaluate the operation and maintenance (O&M) of environmental protection equipment. By acquiring real-time images or video data from the equipment, the system can analyze the equipment's status, abnormalities during operation, and even track maintenance and repair records. This approach, leveraging computer vision technology, can automatically identify whether equipment is operating properly and whether maintenance or repair is required, thereby improving O&M efficiency and reducing the risk of human oversight.

[0003] Wind turbines are environmentally friendly devices that convert wind energy into electricity, making them a clean energy source. Unlike traditional fossil fuel power generation, wind power generation does not emit harmful gases or pollutants, thus reducing air pollution, greenhouse gas emissions, and promoting sustainable development. Existing technologies utilize image analysis technology for wind turbine operation and maintenance monitoring. Using images captured by high-definition cameras or drones mounted on the wind turbines, the system can detect wear, cracks, or other damage on wind turbine blades in real time. Combined with machine learning algorithms, the image analysis system can automatically identify problems on the blade surface and provide early warnings, helping operators and maintenance personnel perform maintenance in advance and avoiding the tedious and inefficient manual inspections. This image analysis-based technology significantly improves operation and maintenance efficiency, reduces the incidence of equipment failures, ensures stable equipment operation, and reduces maintenance costs.

[0004] The existing technology has the following shortcomings:

[0005] Wind turbines are typically located in open environments, where lighting and weather variations (such as strong winds, haze, rain, and snow) can severely impact image quality. In particular, strong sunlight or backlighting can blur image contrast and detail, making it difficult to accurately identify subtle damage like cracks and wear. Furthermore, wind turbines are often located in remote areas, where adverse weather conditions can affect the stability and quality of drones and cameras, further impacting data acquisition accuracy. Summary of the Invention

[0006] The purpose of the present invention is to provide an environmental protection equipment operation and maintenance behavior analysis method and system based on image analysis to address the shortcomings of the background technology.

[0007] In order to achieve the above-mentioned object, the present invention provides the following technical solution: a method for analyzing the operation and maintenance behavior of environmental protection equipment based on image analysis, comprising the following steps:

[0008] S1: Obtain light intensity data and camera vibration data in different time periods;

[0009] S2: Perform feature extraction on the light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data;

[0010] S3: Input the extracted light rate change features and abnormal jitter features into the machine learning model for calculation, calculate the interference level of each image, and classify the images into high-quality images and low-quality images based on the calculation results;

[0011] S4: Optimize the low-quality image and evaluate the detection accuracy of the optimized image. After the evaluation result reaches the predetermined detection accuracy, continue image analysis.

[0012] S5: Use the optimized images to analyze the operation and maintenance behavior of environmental protection equipment.

[0013] Preferably, in S2, the spectrum of the illumination rate is analyzed to generate an illumination rate fluctuation index, and the illumination rate fluctuation index is obtained by:

[0014] Obtain the light intensity data sequence and its corresponding time label, set the sequence of light intensity changes over time, perform fast Fourier transform on the light rate change data to obtain complex output in the frequency domain, calculate the amplitude of each frequency component, and the fluctuation of the light rate is represented by the energy distribution in the spectrum, that is, the square of the amplitude of the frequency component. Calculate the ratio of the sum of the energy of the high-frequency components to the sum of the energy of the low-frequency components, that is, calculate the light rate fluctuation index.

[0015] Preferably, in S2, the camera vibration periodicity is analyzed and identified to generate a camera shake frequency anomaly index, and the camera shake frequency anomaly index is obtained by:

[0016] The vibration signal of the camera is obtained and the discrete wavelet transform is used to perform multi-scale analysis on the vibration signal to obtain wavelet coefficients at multiple scales and displacements. The coefficients at each scale represent the signal characteristics in different frequency ranges. The coefficients at each scale are analyzed and the energy of the coefficients at different scales is calculated. Based on the extracted energy value, the frequency where the wavelet coefficient energy exceeds the set threshold is regarded as the abnormal frequency. The abnormal frequency components at all scales are weighted averaged and summed to generate the camera jitter frequency anomaly index.

[0017] Preferably, in S3, the extracted light rate change features and abnormal jitter features are input into a machine learning model for calculation. The machine learning model calculates the interference degree of each image based on the input features, and divides the image quality into high-quality images and low-quality images. Specifically, the light rate fluctuation index and the camera jitter frequency abnormality index are converted into a comprehensive feature vector, the comprehensive feature vector is used as the input of the machine learning model, the machine learning model is trained, and the interference degree value of each image is determined according to the model output results.

[0018] Preferably, the interference level value of each acquired image is compared with an interference level reference threshold value pre-set based on historical data. If the interference level value of the image is greater than or equal to the interference level reference threshold value, it indicates that the interference level of the image is high, and the image quality is classified as a low-quality image; if the interference level value of the image is less than the interference level reference threshold value, it indicates that the interference level of the image is low, and the image quality is classified as a high-quality image.

[0019] Preferably, in S4, the low-quality image is optimized and the detection accuracy of the optimized image is evaluated. Specifically, for the low-quality image, that is, the interference level value generated within a fixed time period is greater than or equal to the interference level reference threshold, the interference level values greater than or equal to the interference level reference threshold generated by the optimized low-quality image within a subsequent fixed time period are collected, and a data set is established, the mean and standard deviation of the data set are calculated, and after analyzing them, the detection accuracy of the optimized image is evaluated according to the analysis results.

[0020] Preferably, if the mean of the interference level values in the data set is greater than or equal to the reference threshold of the interference level mean value, and the standard deviation of the interference level values is less than the reference threshold of the interference level standard deviation value, it indicates that the optimized image is stable and of good quality, and the detection accuracy is high;

[0021] If the mean value of the interference level is greater than or equal to the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it indicates that the image quality is unstable and needs further optimization;

[0022] If the mean value of the interference level is less than the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it means that the image quality is poor and has large fluctuations, and more stringent optimization is required;

[0023] If the mean of the interference level values is less than the reference threshold of the interference level value mean, and the standard deviation of the interference level values is less than the reference threshold of the interference level value standard deviation, the image quality is poor and the stability is good, but further improvement is still needed to improve the detection accuracy.

[0024] The present invention also provides an environmental protection equipment operation and maintenance behavior analysis system based on image analysis, which includes a data acquisition module, a feature extraction module, a machine learning calculation module, an image optimization module and an operation and maintenance analysis module;

[0025] Data acquisition module: Light sensors deployed at different locations on the wind turbine and the camera's IMU sensor are used to obtain light intensity data and camera vibration data in different time periods.

[0026] Feature extraction module: performs feature extraction on the acquired light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data respectively;

[0027] Machine learning calculation module: This module inputs the extracted light rate change features and abnormal jitter features into a machine learning model for calculation. The machine learning model calculates the interference level of each image based on the input features and classifies the image quality into high-quality and low-quality images.

[0028] Image optimization module: Optimizes low-quality images and evaluates the detection accuracy of the optimized images. Once the evaluation results reach the predetermined detection accuracy, image analysis continues.

[0029] Operation and maintenance analysis module: Use optimized images to analyze the operation and maintenance behavior of environmental protection equipment, analyze whether there are any abnormal conditions in the equipment, and provide decision support for subsequent repairs and maintenance.

[0030] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0031] 1. This method deploys light sensors and camera IMU sensors on wind turbines, combining light intensity data with camera vibration data to assess and optimize image quality. By analyzing light rate fluctuations and camera vibration characteristics, it generates a light rate fluctuation index and a camera jitter frequency anomaly index. These are then input into a machine learning model to automatically classify and optimize image quality. This method effectively identifies and optimizes low-quality images, thereby improving image detection accuracy and providing more accurate analysis results for subsequent environmental protection equipment operation and maintenance.

[0032] 2. This invention addresses the issue of image quality interference in harsh environmental conditions, particularly under extreme lighting and weather conditions. By accurately quantifying the degree of image interference, it optimizes and assesses the quality of low-quality images. This method can efficiently identify the operating status of wind turbines, promptly detect potential anomalies, and provide accurate decision support for equipment maintenance and repair. This not only improves operational efficiency, but also effectively reduces the risk of equipment failure and extends equipment life. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0034] Figure 1 Flow chart of the method of the present invention.

[0035] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0037] Example 1, please refer to Figure 1 As shown, the environmental protection equipment operation and maintenance behavior analysis method based on image analysis described in this embodiment includes the following steps:

[0038] S1: Obtain light intensity data and camera vibration data in different time periods;

[0039] S2: Perform feature extraction on the light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data;

[0040] S3: Input the extracted light rate change features and abnormal jitter features into the machine learning model for calculation, calculate the interference level of each image, and classify the images into high-quality images and low-quality images based on the calculation results;

[0041] S4: Optimize the low-quality image and evaluate the detection accuracy of the optimized image. After the evaluation result reaches the predetermined detection accuracy, continue image analysis.

[0042] S5: Use the optimized images to analyze the operation and maintenance behavior of environmental protection equipment.

[0043] Light sensors typically use sensors such as light-dependent resistors (LDRs) or photodiodes, which can measure ambient light intensity with high precision. Light sensors should be installed at key locations on the wind turbine (such as in front of the blades, at the top of the blades, and at the top of the tower). This ensures that data on light intensity changes can be captured at different times of day (morning, evening, and on cloudy days). This data from multiple locations can be used to comprehensively assess the impact of lighting conditions on the image.

[0044] Light sensors collect real-time light intensity data at various locations, typically recording data every second, every minute, or at intervals tailored to actual needs. This data represents the change in light intensity per unit time, expressed numerically (e.g., in lux), and stored locally or on a cloud server. By acquiring light data from multiple locations, it is possible to identify changes in light intensity around wind turbines, such as shifts in the sun's position, cloud cover, and changes in weather. This provides accurate light data for subsequent image quality assessment.

[0045] IMU (Inertial Measurement Unit) sensors, typically composed of accelerometers, gyroscopes, and magnetometers, are used to monitor the device's motion (such as vibration, tilt, and rotation) in real time. IMU sensors are often integrated into cameras or drone systems. They can be installed in various locations on the wind turbine, particularly on the camera or drone platform, to monitor motion caused by wind speed fluctuations or vibration. To better capture the camera's dynamics, consider installing IMU sensors in front of, behind, or on the left and right sides of the camera system.

[0046] The IMU sensor collects real-time vibration data from the camera, recording information such as acceleration, angular change, and vibration frequency. This data can be used to determine whether the camera is vibrating or shaking in high winds, which could affect image clarity. Vibration data is typically expressed in acceleration (m / s²) and angular velocity (° / s), and the amplitude, frequency, and direction of vibration can be calculated. This sensor provides real-time feedback on device status, particularly in strong winds or during device operation.

[0047] When acquiring data, it is necessary to divide it according to time periods. Usually, the following time periods can be selected: daytime (strong light conditions), evening (low light conditions), nighttime (no light or extremely low light conditions), and different meteorological conditions (such as foggy weather, cloudy weather, snowy weather, etc.).

[0048] Light intensity and vibration data corresponding to each time period can be collected through light sensors and IMU sensors. This data can reflect the impact of changes in light and wind speed on device status under different environmental conditions.

[0049] The light intensity data collected by the light sensor at different time periods can be used to determine the lighting conditions of the environment in which the wind turbine is located. For example, in strong light or backlight conditions, image quality may be poor, making it difficult to discern details.

[0050] The IMU sensor data can be used to analyze the impact of increased wind speed or equipment vibration. If the wind speed is high, it may cause the camera to vibrate violently, affecting image clarity and causing blur or distortion.

[0051] All collected light intensity and camera vibration data can be transmitted to the cloud or local data center via wireless transmission technologies (such as Wi-Fi, LoRa, and 5G). This data can be stored in real time and synchronized with image data for subsequent image analysis systems to conduct environmental impact assessments.

[0052] S2: performing feature extraction on the acquired light intensity data and camera vibration data, respectively extracting light rate change features from the light intensity data and abnormal jitter features from the camera vibration data.

[0053] The light rate of change refers to the rate at which light intensity changes per unit time. Calculating the rate of change of light intensity over time can reflect the dynamics of the lighting environment. By analyzing the light rate of change over consecutive time periods, we can identify trends in light intensity and whether there are rapid or stable changes. For example, rapid light changes may be related to sudden weather changes (such as cloud cover or sunlight penetration) or the influence of objects around the device (such as flocks of birds or flying objects).

[0054] For time series data of light rate changes, several statistical features can be calculated: The mean reflects the overall rate of change. The variance reflects the degree of fluctuation in light changes; a large variance indicates unstable light changes. The maximum and minimum values can identify moments of extreme changes (for example, large fluctuations in light).

[0055] By analyzing the time domain of illumination rate changes, we can identify moments of sudden changes and periods of stability. Sudden changes may indicate that the lighting environment has a significant impact on image quality. Frequency domain analysis methods such as Fourier transforms can be used to analyze the spectrum of the illumination rate. High-frequency components may represent frequent fluctuations in illumination, while low-frequency components represent a more stable lighting environment.

[0056] After analyzing the spectrum of the illumination rate, the illumination rate fluctuation index is generated. The method for obtaining the illumination rate fluctuation index is as follows:

[0057] Get the light intensity data sequence and its corresponding time tag, set the sequence of light intensity changing with time L(t), and calculate its rate change LICR(t), that is: ; where Δt is the time interval, usually seconds or minutes, depending on the frequency of data acquisition.

[0058] Perform fast Fourier transform on the illumination rate change data LICR(t) to convert the time domain signal into a frequency domain signal to reveal its frequency components. For a signal x(t) with a length of N, the Fourier transform result X(f) is expressed as: ;in, is the kernel function of Fourier transform, x(n) is the value of the signal at the nth sampling point, N is the total number of samples of the signal, and through FFT, the complex output in the frequency domain is obtained , where each Corresponding to the amplitude and phase of a frequency component, calculate the amplitude of each frequency component: ;in, are the real and imaginary parts of the complex output, respectively.

[0059] Fluctuations in the rate of illumination are reflected by the energy distribution in the spectrum It is expressed as the square of the amplitude of the frequency component: ; This energy spectrum reflects the intensity of each frequency component. Higher energy means that the frequency component occupies a larger proportion in the change of light rate.

[0060] Calculate the light rate fluctuation index, the expression is: Where, represents the total energy of high-frequency components, represents the sum of the energy of the low-frequency components, is the frequency threshold, used to separate the low-frequency part from the high-frequency part, and QAS is the light rate fluctuation index.

[0061] Vibration amplitude represents the maximum acceleration of the camera over a specific period of time. This can be measured using the accelerometer in the IMU sensor. Vibration frequency reflects the periodicity of camera vibration. A fast Fourier transform (FFT) of the acceleration signal can be performed to obtain a spectrum and extract the frequency components. Frequency analysis can help identify high-frequency jitter (such as rapid vibration caused by wind speed) or low-frequency vibration (such as slow vibration caused by equipment failure).

[0062] By analyzing changes in vibration amplitude and frequency, we can determine whether there is abnormal jitter. For example, when wind speeds are high, the camera may be severely affected by the wind, causing significant jitter. This jitter needs to be treated as an abnormal condition. By calculating the standard deviation, mean, and maximum value of the vibration amplitude and frequency, we can determine whether there are abnormal fluctuations. For example, if the vibration amplitude suddenly increases and the frequency changes significantly, this is usually caused by extreme weather conditions or equipment instability. Frequency analysis identifies the periodicity of camera vibration. If the vibration frequency fluctuates within a specific range and exceeds the normal threshold, it indicates possible abnormal jitter.

[0063] After analyzing and identifying the periodicity of camera vibration, a camera shake frequency anomaly index is generated. The method for obtaining the camera shake frequency anomaly index is as follows:

[0064] Obtain the camera's vibration signal V(t). This is usually time-domain data captured by an accelerometer or IMU sensor installed on the camera. Signal V(t) represents vibration data in the time domain.

[0065] The discrete wavelet transform (DWT) is used to perform multi-scale analysis on the vibration signal, decomposing the time domain signal into different frequency bands. The wavelet transform can extract the high-frequency and low-frequency components in the signal, helping to identify the characteristics of periodic jitter. The expression is: ;in, is the wavelet basis function, j and k represent the scale and displacement parameters respectively.

[0066] Select an appropriate wavelet basis function. Common wavelet basis functions include Haar wavelets, Daubechies wavelets, and Symlet wavelets. The criteria for selecting a wavelet basis function usually depends on the characteristics of the signal. Haar wavelet: For simple changes, Haar wavelet may be a good choice. Daubechies wavelet: For complex signals and periodic features, Daubechies wavelet can provide better decomposition results.

[0067] After wavelet transform, wavelet coefficients W(j,k) are obtained at multiple scales and displacements. The coefficients at each scale represent signal characteristics within a different frequency range. By analyzing these coefficients, possible periodic and sudden jitter characteristics can be identified. The coefficients at each scale are analyzed and the energy of the coefficients at different scales is calculated: ;in, is the wavelet coefficient energy at scale j, and M is the number of sample points at that scale.

[0068] According to the extracted energy value, it is judged whether there is abnormal frequency in the vibration signal. If the frequency component exceeds a set threshold (usually derived from statistical analysis of historical data), it is considered abnormal. The camera shake frequency anomaly index is generated by taking the weighted average of the abnormal frequency components at all scales and calculating the summation.

[0069] S3, converts the illumination rate fluctuation index and the camera jitter frequency anomaly index into a comprehensive feature vector, and uses the comprehensive feature vector as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the interference degree value label of each image as the prediction target, and takes minimizing the sum of the prediction errors of the interference degree value labels of all images as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The interference degree value of each image is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0070] The method for obtaining the interference degree value of each image is to obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: Where, is the output function of the model, QAS is the light rate fluctuation index, GH is the camera jitter frequency anomaly index, is the interference degree value of each image.

[0071] The interference level value of each acquired image is compared with the interference level reference threshold pre-set based on historical data. If the interference level value of the image is greater than or equal to the interference level reference threshold, it means that the interference level of the image is high, and the image quality is classified as a low-quality image; if the interference level value of the image is less than the interference level reference threshold, it means that the interference level of the image is low, and the image quality is classified as a high-quality image.

[0072] S4. For optimizing low-quality images, image enhancement techniques can be used to improve image quality, especially in low-light conditions or when jitter is present. Common methods include histogram equalization, local contrast enhancement, and denoising. Increasing image contrast and brightness can make details clearer, improving the accuracy of subsequent image analysis. Furthermore, denoising techniques, such as Gaussian filtering or median filtering, can effectively remove image noise caused by environmental interference, reducing its impact on image analysis models.

[0073] Secondly, image stabilization techniques can be used to optimize image blurring caused by vibration and jitter. For example, image alignment methods based on feature point matching or optical flow methods can be used to restore image clarity by correcting for translation and rotation. These image optimization steps help improve image quality, enabling image analysis systems to more accurately identify cracks, wear, or other damage in images, ensuring that subsequent machine learning models can make more precise judgments.

[0074] For low-quality images, that is, the interference level value generated within a fixed time period is greater than or equal to the interference level reference threshold, the interference level values generated by the optimized low-quality images within a subsequent fixed time period that are greater than or equal to the interference level reference threshold are collected, and a data set is established. The mean and standard deviation of the data set are calculated, and after analysis, the detection accuracy of the optimized image is evaluated based on the analysis results.

[0075] If the mean value of the interference degree in the data set is greater than or equal to the reference threshold of the mean value of the interference degree, and the standard deviation of the interference degree is less than the reference threshold of the standard deviation of the interference degree, it indicates that the optimized image is stable and of good quality, and the detection accuracy is high;

[0076] If the mean value of the interference level is greater than or equal to the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it indicates that the image quality is unstable and needs further optimization;

[0077] If the mean value of the interference level is less than the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it means that the image quality is poor and has large fluctuations, and more stringent optimization is required;

[0078] If the mean of the interference level values is less than the reference threshold of the interference level value mean, and the standard deviation of the interference level values is less than the reference threshold of the interference level value standard deviation, the image quality is poor and the stability is good, but further improvement is still needed to improve the detection accuracy.

[0079] Once the optimized image passes the evaluation and meets the predetermined detection accuracy (for example, the error rate is less than a certain threshold, and the recall and precision rates meet the requirements), the subsequent image analysis process can proceed. This includes further processing of the optimized image, such as object detection and defect identification, to provide more accurate data support for subsequent operation and maintenance analysis.

[0080] S5, optimized images provide important support for analyzing the operation and maintenance behavior of environmental protection equipment. Through high-precision image analysis, the system can detect potential equipment anomalies in real time, such as cracks, wear, and corrosion on the equipment surface. These problems are often precursors to equipment failure. Through precise image analysis, the system can provide early warning before equipment anomalies occur, helping operators identify and resolve problems in advance, thereby reducing downtime and maintenance costs.

[0081] During the analysis process, the image processing system, combined with machine learning algorithms, can identify abnormal equipment behavior at various stages based on the equipment's operating status and environmental conditions. For example, by monitoring the surface condition of wind turbine blades, it can analyze whether abnormal vibrations caused by wind speed fluctuations or external damage are present, and further infer whether the blades need repair or replacement. Utilizing historical data and image analysis results, the system can track the health of the equipment in real time, creating an equipment health profile and providing an accurate basis for decision support.

[0082] For subsequent repairs and maintenance, the system's analysis results not only provide detailed reports on equipment issues but also support decision-making for maintenance personnel. Based on the equipment's damage severity, failure frequency, and maintenance history, the system automatically generates a maintenance plan, suggesting the most appropriate repair interval or replacement. Through data-driven analysis, operations and maintenance personnel can develop more precise maintenance strategies, avoid unnecessary repairs, optimize resource utilization, and extend equipment lifespan.

[0083] In this embodiment, light sensors and camera IMUs are deployed at different locations on the wind turbine to capture light intensity data and camera vibration data, respectively. Based on this, the system extracts characteristics of light rate changes and abnormal camera vibration jitter. These characteristics are then fed into a machine learning model to calculate the interference level of each image and classify the images as high or low quality. After optimizing low-quality images, detection accuracy is assessed to ensure that the images meet the required accuracy before further analysis. Finally, the optimized images are used to analyze the operation and maintenance behavior of environmental protection equipment, detect any equipment anomalies, provide decision support, and help develop effective repair and maintenance plans to ensure efficient equipment operation.

[0084] Example 2, please refer to Figure 2As shown, the environmental protection equipment operation and maintenance behavior analysis system based on image analysis described in this embodiment includes a data acquisition module, a feature extraction module, a machine learning calculation module, an image optimization module and an operation and maintenance analysis module;

[0085] Data acquisition module: Light sensors deployed at different locations on the wind turbine and the camera's IMU sensor are used to obtain light intensity data and camera vibration data in different time periods.

[0086] Feature extraction module: performs feature extraction on the acquired light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data respectively;

[0087] Machine learning calculation module: This module inputs the extracted light rate change features and abnormal jitter features into a machine learning model for calculation. The machine learning model calculates the interference level of each image based on the input features and classifies the image quality into high-quality and low-quality images.

[0088] Image optimization module: Optimizes low-quality images and evaluates the detection accuracy of the optimized images. Once the evaluation results reach the predetermined detection accuracy, image analysis continues.

[0089] Operation and maintenance analysis module: Use optimized images to analyze the operation and maintenance behavior of environmental protection equipment, analyze whether there are any abnormal conditions in the equipment, and provide decision support for subsequent repairs and maintenance.

[0090] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0091] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0092] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0093] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for analyzing the operation and maintenance behavior of environmental protection equipment based on image analysis, characterized by: The following steps are involved: S1: Obtain light intensity data and camera vibration data in different time periods; The camera vibration data is: the IMU sensor collects the camera vibration data in real time, including the camera's acceleration, angle change and vibration frequency; S2: Perform feature extraction on the light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data; After analyzing the spectrum of the illumination rate, an illumination rate fluctuation index is generated. The illumination rate fluctuation index is obtained by obtaining a sequence of illumination intensity data and its corresponding time tags, setting a sequence of illumination intensity changes over time, performing a fast Fourier transform on the illumination rate change data to obtain a complex output in the frequency domain, and calculating the amplitude of each frequency component. The fluctuation of the illumination rate is represented by the energy distribution in the spectrum, that is, the square of the amplitude of the frequency component. The ratio of the sum of the energy of the high-frequency components to the sum of the energy of the low-frequency components is calculated to obtain the illumination rate fluctuation index. After analyzing and identifying the periodicity of camera vibration, a camera shake frequency anomaly index is generated. The camera shake frequency anomaly index is obtained by obtaining the camera vibration signal, performing a multi-scale analysis on the vibration signal using discrete wavelet transform to obtain wavelet coefficients at multiple scales and displacements. The coefficients at each scale represent signal characteristics within a different frequency range. The coefficients at each scale are analyzed, and the energy of the coefficients at different scales is calculated. Based on the extracted energy value, the frequency at which the wavelet coefficient energy exceeds the set threshold is regarded as an abnormal frequency. The abnormal frequency components at all scales are weighted averaged and summed to generate the camera shake frequency anomaly index. S3: Input the extracted light rate change features and abnormal jitter features into the machine learning model for calculation, calculate the interference level of each image, and classify the images into high-quality images and low-quality images based on the calculation results; S4: Optimize the low-quality image and evaluate the detection accuracy of the optimized image. After the evaluation result reaches the predetermined detection accuracy, continue image analysis. S5: Use the optimized images to analyze the operation and maintenance behavior of environmental protection equipment.

2. The method for analyzing environmental protection equipment operation and maintenance behavior based on image analysis according to claim 1, characterized in that: In S3, the extracted light rate change features and abnormal jitter features are input into the machine learning model for calculation. The machine learning model calculates the interference degree of each image based on the input features, and divides the image quality into high-quality images and low-quality images. Specifically, the light rate fluctuation index and the camera jitter frequency abnormality index are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model is trained, and the interference degree value of each image is determined according to the model output results.

3. The method for analyzing environmental protection equipment operation and maintenance behavior based on image analysis according to claim 2, characterized in that: The interference level value of each acquired image is compared with the interference level reference threshold pre-set based on historical data. If the interference level value of the image is greater than or equal to the interference level reference threshold, it means that the interference level of the image is high, and the image quality is classified as a low-quality image; if the interference level value of the image is less than the interference level reference threshold, it means that the interference level of the image is low, and the image quality is classified as a high-quality image.

4. The method for analyzing environmental protection equipment operation and maintenance behavior based on image analysis according to claim 3, characterized in that: In S4, the low-quality image is optimized and the detection accuracy of the optimized image is evaluated. Specifically, for the low-quality image, that is, the interference level value generated within a fixed time period is greater than or equal to the interference level reference threshold, the interference level values greater than or equal to the interference level reference threshold generated by the optimized low-quality image within a subsequent fixed time period are collected, and a data set is established. The mean and standard deviation of the data set are calculated, and after analyzing them, the detection accuracy of the optimized image is evaluated according to the analysis results.

5. The method for analyzing environmental protection equipment operation and maintenance behavior based on image analysis according to claim 4, characterized in that: If the mean value of the interference degree in the data set is greater than or equal to the reference threshold of the mean value of the interference degree, and the standard deviation of the interference degree is less than the reference threshold of the standard deviation of the interference degree, it indicates that the optimized image is stable and of good quality, and the detection accuracy is high; If the mean value of the interference level is greater than or equal to the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it indicates that the image quality is unstable and needs further optimization; If the mean value of the interference level is less than the reference threshold of the mean value of the interference level, and the standard deviation of the interference level is greater than or equal to the reference threshold of the standard deviation of the interference level, it means that the image quality is poor and has large fluctuations, and more stringent optimization is required; If the mean of the interference level values is less than the reference threshold of the interference level value mean, and the standard deviation of the interference level values is less than the reference threshold of the interference level value standard deviation, the image quality is poor and the stability is good, but further improvement is still needed to improve the detection accuracy.

6. An environmental protection equipment operation and maintenance behavior analysis system based on image analysis, used to implement the environmental protection equipment operation and maintenance behavior analysis method based on image analysis according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, feature extraction module, machine learning calculation module, image optimization module and operation and maintenance analysis module; Data acquisition module: Light sensors deployed at different locations on the wind turbine and the camera's IMU sensor are used to obtain light intensity data and camera vibration data in different time periods. Feature extraction module: performs feature extraction on the acquired light intensity data and camera vibration data, extracting the light rate change feature in the light intensity data and the abnormal jitter feature in the camera vibration data respectively; Machine learning calculation module: This module inputs the extracted light rate change features and abnormal jitter features into a machine learning model for calculation. The machine learning model calculates the interference level of each image based on the input features and classifies the image quality into high-quality and low-quality images. Image optimization module: Optimizes low-quality images and evaluates the detection accuracy of the optimized images. Once the evaluation results reach the predetermined detection accuracy, image analysis continues. Operation and maintenance analysis module: Use optimized images to analyze the operation and maintenance behavior of environmental protection equipment, analyze whether there are any abnormal conditions in the equipment, and provide decision support for subsequent repairs and maintenance.

Images