Algorithm for unmanned aerial vehicle inspection intelligent analysis of hydraulic power plant dam

Abstract

The invention relates to the technical field of hydraulic engineering detection, and discloses a hydraulic power plant dam unmanned aerial vehicle inspection intelligent analysis algorithm, which comprises the following steps: S1, obtaining hydraulic power plant dam image data based on an unmanned aerial vehicle, uploading and preprocessing the obtained image data, and then storing the image data; s2, key feature recognition and extraction are performed on the dam image data based on a deep learning algorithm, and texture features and deformation features are extracted; and S3, calculating the texture contrast and entropy parameters of the crack region by adopting a gray-level co-occurrence matrix, and understanding the internal development trend of the dam crack to obtain a crack value. According to the method, the potential safety hazards of the dam can be quickly and effectively found, once the safety evaluation value exceeds the safety threshold value, measures can be taken immediately, so that safety accidents are avoided, the safety of the dam is guaranteed, inconvenience and missing of manual data acquisition are avoided in a manner of acquiring data through the unmanned aerial vehicle, and the careless omission condition is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy engineering detection, and in particular to an algorithm for intelligent analysis of unmanned aerial vehicle inspection of a hydropower plant dam. Background Art

[0002] As the importance of hydropower energy in the global energy structure becomes increasingly prominent, the safe operation of hydropower plants has become a vital task. As a key facility of the water conservancy hub, the structural integrity and safety of the hydropower plant dam are directly related to the safety of life and property of the people in the downstream area and the stability of the power supply. The traditional dam inspection method mainly relies on manual inspection and some fixed monitoring equipment. Although manual inspection can observe the surface of the dam on the spot, it has many limitations. On the one hand, manual inspection is inefficient. For large hydropower plant dams, inspectors need to spend a lot of time and energy to complete a comprehensive inspection, and it is difficult to conduct frequent high-frequency inspections. On the other hand, manual inspection is greatly affected by subjective factors. The experience, skill level and physical condition of the inspectors may lead to deviations or omissions in the inspection results. When checking the cracks in the dam body, some fine cracks may not be discovered due to the negligence of the inspectors, or the measurement of the length, width and depth parameters of the cracks is not accurate enough, resulting in omissions and serious consequences. In this regard, we propose an algorithm for intelligent analysis of drone inspection of hydropower plant dams. Summary of the invention

[0003] In order to solve the above technical problems, an algorithm for intelligent analysis of drone inspection of hydropower plant dams is provided. This technical solution solves the problem that the subjective factors of manual inspection are greatly influenced, the measurement is not accurate enough, and omissions are caused.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: an algorithm for intelligent analysis of hydropower plant dam drone inspection, the algorithm steps are:

[0005] S1. Obtain image data of the hydropower plant dam based on the drone, upload the acquired image data, and store it after preprocessing;

[0006] S2, based on the deep learning algorithm, key feature recognition and extraction of dam image data are performed to extract texture features and deformation features;

[0007] S3, using the gray-level co-occurrence matrix to calculate the texture contrast and entropy parameters of the crack area, understand the internal development trend of the dam cracks, and obtain the crack value;

[0008] S4. Match the feature points of the images taken by drones at different times based on the SIFT algorithm, calculate the displacement of the feature points, understand the deformation of the dam body, and obtain the deformation value;

[0009] S5. Based on the crack value and deformation value, the safety assessment value of the dam is comprehensively calculated;

[0010] S6. Establish a safety threshold for dam safety, compare the safety assessment value with the safety threshold, and determine whether there are safety hazards in the current dam;

[0011] S7. Display the algorithm process and results in a visual way for staff to understand and provide feedback, and store the calculation results to provide a basis for subsequent analysis.

[0012] Preferably, in step S1, the drone performs real-time shooting based on a camera, shoots the dam from multiple angles and different heights, uploads the captured images, and performs image preprocessing to first check the integrity of the image data, and uses a filtering algorithm to denoise the image; then, the image is color corrected and enhanced, and the key features in the image are highlighted by adjusting the brightness, contrast and saturation parameters of the image; the image is acquired multiple times at different times, and preprocessed, and the processed data is classified and stored.

[0013] Preferably, in step S2, a multi-layer convolution layer, a pooling layer and a fully connected layer are constructed to automatically learn the texture pattern in the dam image; in step S3, the image is set to I (x, y), the gray level is L, and the gray level co-occurrence matrix is ​​calculated, and the calculation formula is:

[0014]

[0015] in is the gray-level co-occurrence matrix expression, i and j represent two different gray values, d is the distance between pixel pairs, is the direction, M and N represent the number of rows and columns of the image respectively.

[0016] Preferably, the contrast calculation formula in step S3 is:

[0017]

[0018] Where C is the calculation formula for contrast, L is the grayscale, and contrast measures the magnitude of local grayscale changes in the image. The greater the contrast, the clearer the image texture and the more obvious the boundary. The entropy value calculation formula is:

[0019]

[0020] Where H is the entropy value. Entropy represents the randomness of the image texture. The larger the entropy value, the more complex and irregular the texture. The calculation formula for the crack value is:

[0021]

[0022] Where V is the crack value obtained by comprehensive calculation, and w1 and w2 are weight coefficients.

[0023] Preferably, the feature point matching formula in step S4 is:

[0024]

[0025] Where A(b1, b2) is the Euclidean distance value, E1 and E2 are the extracted SIFT feature points and descriptors, b1 and b2 are feature points, and a threshold F is set. When A(b1, b2) is less than F, b1 and b2 are considered to be matching feature points.

[0026] Preferably, in step S4, for the successfully matched feature point (b1, b2), assuming that the coordinates of b1 in image I1 are (x1, y1), and the coordinates of b2 in image I2 are (x2, y2), then the calculation formula for the displacement of the feature point pair is: △x=x2-x1, △y=y2-y1.

[0027] Preferably, the deformation value calculation formula in step S4 is:

[0028]

[0029] Where S is the deformation value, △x and △y are the displacements of the feature points; the safety assessment value calculation formula in step S5 is:

[0030] U=w3V+w4S

[0031] Where U is the safety assessment value, and w3 and w4 are weight coefficients.

[0032] Preferably, the method for establishing the safety threshold in step S6 is obtained through historical data analysis. By organizing and analyzing the monitoring data during the long-term operation of the dam, including crack development data, deformation data and seepage data in different periods, the performance of the dam under various working conditions is understood, thereby determining the deformation safety threshold.

[0033] Preferably, in step S6, the calculated safety assessment value is compared with the threshold value. If the safety assessment value is lower than the safety threshold value, it indicates that the dam is currently in a relatively safe state and all indicators are within an acceptable range. On the contrary, if the safety threshold value is exceeded, it indicates that the dam has safety hazards, and the emergency plan is activated, and professional personnel are organized to conduct a comprehensive and in-depth inspection and assessment of the dam to identify hidden dangers.

[0034] Preferably, in step S7, an intuitive user interface is constructed to present the data collection process in the form of a dynamic flow chart on the interface, the results obtained after calculation and analysis of the collected data are presented in the form of charts for viewing by staff, and the data analyzed this time is stored.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention calculates the safety assessment value by comprehensively combining the crack value and the deformation value, and comprehensively evaluates the safety status of the dam from multiple key angles. By comparing the safety assessment value with the safety threshold, the potential safety hazards of the dam can be quickly and effectively discovered. Once the safety assessment value exceeds the safety threshold, measures can be taken immediately to avoid the occurrence of safety accidents and ensure the safety of the dam. The method of collecting data by drone avoids the inconvenience and omissions of manual data collection, is not affected by subjective factors, can discover some minor cracks, and avoid omissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The figure is a flowchart of the algorithm steps of the present invention. DETAILED DESCRIPTION

[0038] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0039] Reference Figure 1 As shown in the figure, an algorithm for intelligent analysis of drone inspection of hydropower plant dams is shown in the figure. The algorithm steps are as follows:

[0040] S1. Obtain image data of the hydropower plant dam based on the drone, upload the acquired image data, and store it after preprocessing;

[0041] S2, based on the deep learning algorithm, key feature recognition and extraction of dam image data are performed to extract texture features and deformation features;

[0042] S3, using the gray-level co-occurrence matrix to calculate the texture contrast and entropy parameters of the crack area, understand the internal development trend of the dam cracks, and obtain the crack value;

[0043] S4. Match the feature points of the images taken by drones at different times based on the SIFT algorithm, calculate the displacement of the feature points, understand the deformation of the dam body, and obtain the deformation value;

[0044] S5. Based on the crack value and deformation value, the safety assessment value of the dam is comprehensively calculated;

[0045] S6. Establish a safety threshold for dam safety, compare the safety assessment value with the safety threshold, and determine whether there are safety hazards in the current dam;

[0046] S7. Display the algorithm process and results in a visual way for staff to understand and provide feedback, and store the calculation results to provide a basis for subsequent analysis.

[0047] This application uses drones to obtain dam image data, which greatly improves the efficiency and scope of data collection. Drones can easily reach every corner of the dam, including some areas that are difficult to reach or inconvenient to observe manually, such as steep slopes and high positions of the dam, thereby obtaining comprehensive and rich image information and avoiding data omissions and monitoring blind spots;

[0048] The texture contrast and entropy parameters calculated by the gray-level co-occurrence matrix can convert the texture characteristics of cracks into numerical indicators, namely crack values, which makes the evaluation of cracks more objective and accurate. Different cracks can be quantitatively compared, which is convenient for determining the severity and development priority of cracks.

[0049] The SIFT algorithm is invariant to the scale, rotation and illumination changes of the image. It can accurately match feature points at different times and under different shooting conditions, and then accurately calculate the displacement to obtain the deformation value. This is extremely critical for monitoring the long-term micro-deformation of the dam body. It can timely discover the potential structural instability of the dam body and provide reliable data support to ensure the safe operation of the dam. It can conduct detailed deformation analysis of the dam body as a whole and in local areas. It can not only determine the displacement trend of the dam body as a whole, but also locate local weak links or abnormal deformation areas, which is helpful to formulate targeted maintenance and reinforcement plans, improve resource utilization efficiency, and reduce maintenance costs.

[0050] The safety assessment value is calculated by combining the crack value and the deformation value, taking into account the two key factors of dam safety and avoiding the one-sidedness of single factor assessment. This comprehensive consideration can more comprehensively and accurately reflect the actual safety status of the dam, because cracks and deformation are often interrelated and affect each other. The combined assessment of the two can better predict the stability and potential risks of the dam.

[0051] By comparing with the safety assessment value, dam safety hazards can be discovered in time and early warnings can be issued. When the assessment value exceeds the threshold, the corresponding emergency plan can be immediately activated, such as increasing the monitoring frequency, organizing experts to conduct on-site inspections, and formulating emergency maintenance plans to effectively prevent the occurrence of safety accidents and protect the lives and property of people in downstream areas.

[0052] In step S1, the drone takes real-time photos based on the camera, photographs the dam from multiple angles and different heights, and uploads the captured images. The image preprocessing first checks the integrity of the image data and uses a filtering algorithm to denoise the image. Then, the image is color corrected and enhanced by adjusting the brightness, contrast and saturation parameters of the image to highlight the key features in the image. The image is acquired multiple times at different times and preprocessed, and the processed data is classified and stored.

[0053] This application shoots the dam from multiple angles and different heights to ensure that all parts of the dam can be clearly captured. Whether it is the dam top, dam slope, dam foundation, or the side and back areas of the dam, there will be no blind spots in monitoring. Shooting from a low angle can better observe the connection between the dam foundation and the surrounding ground. Shooting from a high angle helps to grasp the overall shape of the dam and the relative position relationship between different areas, thereby providing a rich source of data for a comprehensive and accurate assessment of the structural integrity and safety of the dam.

[0054] This shooting method helps to construct a three-dimensional image model of the dam, allowing analysts to study the dam from different perspectives. For some complex structural features or potential safety hazards, such as the direction of cracks on the dam body and the three-dimensional shape of the deformation area, images shot from multiple angles can provide more comprehensive information, facilitate in-depth analysis, and accurately determine the degree of impact on the overall stability of the dam.

[0055] In step S2, multiple convolutional layers, pooling layers and fully connected layers are constructed to automatically learn the texture pattern in the dam image; in step S3, the image is set to I (x, y), the gray level is L, and the gray level co-occurrence matrix is ​​calculated. The calculation formula is:

[0056]

[0057] in is the gray-level co-occurrence matrix expression, i and j represent two different gray values, d is the distance between pixel pairs, is the direction, M and N represent the number of rows and columns of the image respectively.

[0058] The grayscale co-occurrence matrix of the present application is an important tool for analyzing image texture features. It describes the texture information of an image by counting the frequency of occurrence of pixel grayscale pairs at a certain distance and direction in the image. The characteristics of the texture in the image can be reflected by the grayscale relationship between pixels. In image areas with regular textures, such as the uniform concrete surface on the dam body, pixel pairs with certain grayscale value combinations will appear frequently; while in areas with cracks or other anomalies, the distribution of grayscale pairs will be different.

[0059] The contrast calculation formula in step S3 is:

[0060]

[0061] Where C is the calculation formula for contrast, L is the grayscale, and contrast measures the magnitude of local grayscale changes in the image. The greater the contrast, the clearer the image texture and the more obvious the boundary. The entropy value calculation formula is:

[0062]

[0063] Where H is the entropy value. Entropy represents the randomness of the image texture. The larger the entropy value, the more complex and irregular the texture. The calculation formula for the crack value is:

[0064]

[0065] Where V is the crack value obtained by comprehensive calculation, and w1 and w2 are weight coefficients.

[0066] The feature point matching formula in step S4 is:

[0067]

[0068] Where A(b1, b2) is the Euclidean distance value, E1 and E2 are the extracted SIFT feature points and descriptors, b1 and b2 are feature points, and a threshold F is set. When A(b1, b2) is less than F, b1 and b2 are considered to be matching feature points.

[0069] In step S4, for the successfully matched feature point (b1, b2), assuming that the coordinates of b1 in image I1 are (x1, y1), and the coordinates of b2 in image I2 are (x2, y2), then the formula for calculating the displacement of the feature point pair is: △x=x2-x1, △y=y2-y1.

[0070] The deformation value calculation formula in step S4 is:

[0071]

[0072] Where S is the deformation value, △x and △y are the displacements of the feature points; the safety assessment value calculation formula in step S5 is:

[0073] U=w3V+w4S

[0074] Where U is the safety assessment value, and w3 and w4 are weight coefficients.

[0075] The method for establishing the safety threshold in step S6 is obtained through historical data analysis. By organizing and analyzing the monitoring data during the long-term operation of the dam, including crack development data, deformation data and seepage data in different periods, the performance of the dam under various working conditions is understood, thereby determining the deformation safety threshold.

[0076] This application deeply analyzes various monitoring data during the long-term operation of the dam, such as crack development, deformation and seepage data, and the deformation safety threshold determined by it can closely fit the actual characteristics and operating conditions of the dam. These data reflect the actual performance of the dam under different environmental conditions, water level changes, seasonal alternations and geological activities. This makes the safety threshold highly targeted and accurate. For a dam located in an earthquake-active zone that has experienced multiple strong earthquakes, its deformation safety threshold will fully consider the instantaneous and long-term impacts of earthquakes on the dam structure, so as to more effectively judge whether the deformation caused by the earthquake exceeds the safety range in subsequent monitoring, avoiding misjudgment or omission that may result from the use of a general threshold.

[0077] In step S6, the calculated safety assessment value is compared with the threshold. If the safety assessment value is lower than the safety threshold, it indicates that the dam is currently in a relatively safe state and all indicators are within an acceptable range.

[0078] On the contrary, if the safety threshold is exceeded, it indicates that the dam has safety hazards. The emergency plan will be activated and professional personnel will be organized to conduct a comprehensive and in-depth inspection and assessment of the dam to identify potential hazards.

[0079] This application can accurately determine the current safety status of the dam by comparing the safety assessment value with the threshold. This quantitative comparison method avoids subjective assumptions and vague judgments, and provides a clear basis for the safety management of the dam. Whether it is daily routine monitoring or when facing special working conditions, it can quickly and accurately understand whether the dam is safe, making decision-making more scientific and reasonable. When continuous rainfall in the rainy season causes the reservoir water level to rise, timely comparison of the safety assessment value with the threshold can quickly determine whether the dam poses a safety risk due to water level pressure, so as to make preparations in advance or adjust the operation strategy.

[0080] In step S7, an intuitive user interface is constructed to present the data collection process in the form of a dynamic flow chart on the interface. The results obtained after calculation and analysis of the collected data are presented in the form of charts for staff to view, and the data analyzed this time is stored.

[0081] This application presents the data collection process in the form of a dynamic flowchart, allowing staff to clearly understand the execution order and logical relationship of a series of links from drone takeoff, image capture, data transmission to preprocessing. This helps staff to quickly grasp the operation status of the entire data acquisition process and promptly discover possible problems or abnormal links. If a step in the data collection process is delayed or erroneous, staff can intuitively locate the link through the dynamic flowchart and quickly take corresponding solutions to ensure smooth data collection.

[0082] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention, and these changes and improvements fall within the scope of the present invention claimed.

Claims

1. An algorithm for intelligent analysis of hydropower plant dam drone inspection, characterized in that: The algorithm steps are: S1. Obtain image data of the hydropower plant dam based on the drone, upload the acquired image data, and store it after preprocessing; S2, based on the deep learning algorithm, key feature recognition and extraction of dam image data are performed to extract texture features and deformation features; S3, using the gray-level co-occurrence matrix to calculate the texture contrast and entropy parameters of the crack area, understand the internal development trend of the dam cracks, and obtain the crack value; S4. Match the feature points of the images taken by drones at different times based on the SIFT algorithm, calculate the displacement of the feature points, understand the deformation of the dam body, and obtain the deformation value; S5. Based on the crack value and deformation value, the safety assessment value of the dam is comprehensively calculated; S6. Establish a safety threshold for dam safety, compare the safety assessment value with the safety threshold, and determine whether there are safety hazards in the current dam; S7. Display the algorithm process and results in a visual way for staff to understand and provide feedback, and store the calculation results to provide a basis for subsequent analysis.

2. According to claim 1, the algorithm for intelligent analysis of hydropower plant dam drone inspection is characterized in that: In step S1, the drone takes real-time photos based on the camera, photographs the dam from multiple angles and different heights, and uploads the captured images. The image preprocessing first checks the integrity of the image data, and uses a filtering algorithm to denoise the image. Then, the image is color corrected and enhanced, and the key features in the image are highlighted by adjusting the brightness, contrast and saturation parameters of the image. The images are acquired multiple times at different times, and preprocessed, and the processed data are classified and stored.

3. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: In step S2, multiple convolutional layers, pooling layers and fully connected layers are constructed to automatically learn the texture pattern in the dam image; in step S3, the image is set to I (x, y), the gray level is L, and the gray level co-occurrence matrix is ​​calculated. The calculation formula is: ； in is the gray-level co-occurrence matrix expression, i and j represent two different gray values, d is the distance between pixel pairs, is the direction, M and N represent the number of rows and columns of the image respectively.

4. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: The contrast calculation formula in step S3 is: ； Where C is the calculation formula for contrast, L is the grayscale, and contrast measures the magnitude of local grayscale changes in the image. The greater the contrast, the clearer the image texture and the more obvious the boundary. The entropy value calculation formula is: ； Where H is the entropy value. Entropy represents the randomness of the image texture. The larger the entropy value, the more complex and irregular the texture. The calculation formula for the crack value is: ； Where V is the crack value obtained by comprehensive calculation, and w1 and w2 are weight coefficients.

5. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: The feature point matching formula in step S4 is: ； Where A(b1, b2) is the Euclidean distance value, E1 and E2 are the extracted SIFT feature points and descriptors, b1 and b2 are feature points, and a threshold F is set. When A(b1, b2) is less than F, b1 and b2 are considered to be matching feature points.

6. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: In step S4, for the successfully matched feature point (b1, b2), assuming that the coordinates of b1 in image I1 are (x1, y1), and the coordinates of b2 in image I2 are (x2, y2), then the formula for calculating the displacement of the feature point pair is: △x=x2-x1, △y=y2-y1.

7. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: The deformation value calculation formula in step S4 is: ； Where S is the deformation value, △x and △y are the displacements of the feature points; the safety assessment value calculation formula in step S5 is: U=w3V+w4S Where U is the safety assessment value, and w3 and w4 are weight coefficients.

8. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: The method for establishing the safety threshold in step S6 is obtained through historical data analysis. By organizing and analyzing the monitoring data during the long-term operation of the dam, including crack development data, deformation data and seepage data in different periods, the performance of the dam under various working conditions is understood, thereby determining the deformation safety threshold.

9. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: In step S6, the calculated safety assessment value is compared with the threshold. If the safety assessment value is lower than the safety threshold, it indicates that the dam is currently in a relatively safe state and all indicators are within an acceptable range. On the contrary, if the safety threshold is exceeded, it indicates that the dam has safety hazards. The emergency plan will be activated and professional personnel will be organized to conduct a comprehensive and in-depth inspection and assessment of the dam to identify potential hazards.

10. The algorithm for intelligent analysis of hydropower plant dam inspection by drone according to claim 1 is characterized in that: In step S7, an intuitive user interface is constructed to present the data collection process in the form of a dynamic flow chart on the interface. The results obtained after calculation and analysis of the collected data are presented in the form of charts for staff to view, and the data analyzed this time is stored.

Images