Concrete dam underwater chipping and crack detection method based on image sonar

Abstract

The invention discloses a concrete dam underwater chipping and crack detection method based on an image sonar, and the method comprises the steps: dynamically adjusting the transmitting frequency and scanning angle of sonar equipment according to the conditions of a dam body and water flow, and carrying out the omnibearing scanning and collecting high-resolution sonar image data; sequentially carrying out adaptive median filtering noise reduction and histogram equalization enhancement processing on the image, accurately extracting the area, the shape and the edge roughness of a falling block, the length, the width, the direction and the curvature of a crack and other characteristics, and then identifying diseases by using a support vector machine algorithm trained by a large number of samples; diseases are positioned by means of the triangulation principle in combination with sonar equipment position information, the severity of the diseases is evaluated according to hydraulic engineering detection standards, and a visual detection report covering various information of the diseases is finally generated. The method effectively improves the precision and reliability of underwater chipping and crack detection of the concrete dam, provides a scientific basis for dam body maintenance, and is suitable for popularization and application. And safe operation is ensured.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering inspection, and in particular to a method for detecting underwater blockages and cracks in concrete dams based on image sonar. Background Technology

[0002] Concrete dams, operating in long-term underwater environments, are susceptible to damage from various factors such as water flow erosion, water pressure, and temperature changes, leading to issues like spalling and cracking. If these defects are not detected and addressed promptly, they can seriously threaten the safe operation of the dam. Traditional underwater inspection methods, such as direct inspection by divers, are inefficient and highly dangerous; conventional non-destructive testing techniques struggle to guarantee accuracy and reliability in the complex underwater environment. While image-based sonar technology has been applied to underwater inspection, existing methods suffer from insufficient accuracy and poor ability to identify minute cracks and spalling, failing to meet the demands for precise underwater inspection of concrete dams. Summary of the Invention

[0003] (I) Technical Issues

[0004] This paper presents a method for detecting underwater spalling and cracks in concrete dams based on image sonar, in order to improve the accuracy and reliability of underwater spalling and crack detection in concrete dams and achieve rapid and accurate identification of defects.

[0005] (II) Technical Solution

[0006] The technical solution of this invention is: a method for detecting underwater rockfalls and cracks in concrete dams based on image sonar, comprising the following steps:

[0007] S1. Data Acquisition: An image sonar device is used to perform a 360-degree scan of the underwater portion of the concrete dam. The transmission frequency f and scanning angle θ of the sonar device are dynamically adjusted according to the dam structure and water flow conditions to acquire high-resolution sonar image data. The transmission frequency f satisfies the following relationship with the detection distance d and the speed of sound in water c:

[0008] S2. Image Preprocessing: The acquired sonar images are sequentially subjected to adaptive median filtering for noise reduction and histogram equalization for enhancement. The adaptive median filtering dynamically adjusts the filter window size based on local image features. The histogram equalization formula is:

[0009]

[0010] S3. Feature Extraction: Extract area feature A and shape feature C (circularity) for the dropped block. P represents the perimeter of the region and the edge roughness feature R; for cracks, features such as length L, width W, direction α, and curvature K are extracted.

[0011] S4. Disease identification: The Support Vector Machine (SVM) algorithm is used to train the SVM model using labeled sonar image samples. The extracted feature vectors are input into the trained SVM model to identify shards, cracks, and normal areas.

[0012] S5. Disease Location: Based on the location information from the sonar equipment and the characteristic location of the disease in the image, the location coordinates (x, y) of the disease in the plane coordinate system are calculated using the following formula based on the principle of triangulation:

[0013]

[0014] The position of the sonar device in three-dimensional space is determined by combining depth information, where α1 and α2 are the angles corresponding to the disease feature points in the images obtained by the sonar device at different positions, and b is the distance between the two positions of the sonar device.

[0015] S6. Disease Assessment: Based on the identified area and depth of the fallen pieces, as well as the length, width, and depth of the cracks, assess the severity of the disease according to the water conservancy engineering testing standards.

[0016] S7. Generate Detection Report: Generate a detection report in a visual manner, including sonar images, feature parameters, and evaluation results, based on the detected disease information, including disease type, location, and severity.

[0017] Furthermore, in the data acquisition step, the scanning angle θ is adjusted according to the complexity of the dam surface and the expected detection area to ensure coverage of the entire detection area.

[0018] Furthermore, in the image preprocessing step, the adaptive median filter uses a larger window for filtering areas with more noise and a smaller window for areas with rich details.

[0019] Furthermore, in the feature extraction step, the edge information of the crack is obtained through an edge detection algorithm, and the length L, width W, direction α, and curvature K parameters of the crack are calculated using a curve fitting method.

[0020] Furthermore, in the disease identification step, during SVM model training, cross-validation is used to optimize model parameters and improve the model's generalization ability.

[0021] Furthermore, in the disease assessment step, for slabs, the hazard level of the slabs is determined by comparing the area and depth of the slabs with standard thresholds; for cracks, the degree of impact on the stability of the dam structure is comprehensively assessed based on the length, width and depth parameters of the cracks.

[0022] Furthermore, in the step of generating the test report, the test report is output in PDF or XML format for easy storage and transmission.

[0023] (III) Beneficial Effects:

[0024] This invention achieves clearer and more complete underwater sonar images by dynamically adjusting sonar equipment parameters, thus improving the quality of detection data. Adaptive median filtering and histogram equalization are employed for image preprocessing, effectively enhancing image features and improving the accuracy of subsequent feature extraction and defect identification. Extracting multiple feature parameters and utilizing support vector machines for defect identification improves the accuracy of identifying spalling and cracks, especially micro-cracks and spalling. A defect location method based on triangulation principles accurately determines the location of defects on the concrete dam, providing precise location information for subsequent maintenance. Defect assessment based on relevant standards and the generation of detailed inspection reports provide water conservancy project managers with intuitive and comprehensive defect information, facilitating the development of scientific and reasonable maintenance plans and ensuring the safe operation of the concrete dam. Attached Figure Description

[0025] Figure 1 This is a flowchart of the underwater block and crack detection method for concrete dams based on image sonar according to the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] like Figure 1 As shown, this invention discloses a method for detecting underwater rockfalls and cracks in concrete dams based on image sonar, comprising the following steps:

[0028] S1. Data Acquisition: An image sonar device is used to perform a 360-degree scan of the underwater portion of the concrete dam. The transmission frequency f and scanning angle θ of the sonar device are dynamically adjusted according to the dam structure and water flow conditions to acquire high-resolution sonar image data. The transmission frequency f satisfies the following relationship with the detection distance d and the speed of sound in water c: The scanning angle θ is adjusted according to the complexity of the dam surface and the expected detection area to ensure coverage of the entire area to be detected.

[0029] S2. Image Preprocessing: The acquired sonar images are sequentially subjected to adaptive median filtering for noise reduction and histogram equalization for enhancement. The adaptive median filtering dynamically adjusts the filter window size based on local image features, using a larger window for noisy areas and a smaller window for detailed areas. The histogram equalization formula is:

[0030]

[0031] S3. Feature Extraction: Extract area feature A and shape feature C (circularity) for the dropped block. P represents the perimeter of the region and the edge roughness feature R; the edge information of the crack is obtained through the edge detection algorithm, and the length L, width W, direction α and curvature K parameters of the crack are calculated using the curve fitting method.

[0032] S4. Disease Identification: The Support Vector Machine (SVM) algorithm is used to train the SVM model using labeled sonar image samples. The extracted feature vectors are input into the trained SVM model to identify shards, cracks, and normal areas. During SVM model training, cross-validation is used to optimize model parameters and improve the model's generalization ability.

[0033] S5. Disease Location: Based on the location information from the sonar equipment and the characteristic location of the disease in the image, the location coordinates (x, y) of the disease in the plane coordinate system are calculated using the following formula based on the principle of triangulation:

[0034]

[0035] The position of the sonar device in three-dimensional space is determined by combining depth information, where α1 and α2 are the angles corresponding to the disease feature points in the images obtained by the sonar device at different positions, and b is the distance between the two positions of the sonar device.

[0036] S6. Disease Assessment: Based on the identified area and depth of the spalled blocks, as well as the length, width, and depth of the cracks, the severity of the disease is assessed according to the water conservancy engineering testing standards. For spalled blocks, the hazard level is determined by comparing the area and depth of the spalled blocks with the standard threshold. For cracks, the impact on the stability of the dam structure is comprehensively assessed based on the length, width, and depth parameters of the cracks.

[0037] S7. Generate Detection Report: Generate a detection report in a visual format, including sonar images, feature parameters, and evaluation results, based on the detected disease information, including disease type, location, and severity. The detection report is output in PDF or XML format for easy storage and transmission.

[0038] The specific detailed operation process of this invention is as follows:

[0039] I. Data Collection

[0040] Equipment Preparation: Select a high-resolution, wide-beam imaging sonar device with adjustable transmission frequency and scanning angle. A multi-beam imaging sonar (such as the Kongsberg HISAS series) is suitable. It should combine high resolution and wide beam performance. Before use, perform a comprehensive calibration and debugging of the sonar device to ensure all parameters are accurate. Connect the sonar device to the data acquisition terminal to ensure stable data transmission.

[0041] On-site inspection: Before testing, a detailed inspection of the concrete dam's structure, water flow velocity, water depth, and other conditions is conducted. Based on historical data and the actual site conditions, key inspection areas are identified. For example, areas on the dam's upstream side that are easily eroded by the water flow, and the connection between the dam and its foundation, are marked as critical inspection areas.

[0042] Parameter Adjustment: Determine the transmission frequency of the sonar equipment based on the water depth of the detection area and the expected detection distance. Select an appropriate integer value as the transmission frequency. Simultaneously, adjust the scanning angle according to the complexity of the dam surface and the shape of the detection area. For relatively flat areas, set the scanning angle to 60°-90°; for areas with complex structures or potential defects, increase the scanning angle to 120°-180° to ensure the sonar scan covers the entire area to be detected.

[0043] Scanning Operation: The sonar equipment is mounted on an underwater mobile platform, such as an unmanned underwater vehicle (AUV) or a remotely operated underwater vehicle (ROV). The platform is remotely controlled to slowly and uniformly scan the underwater portion of the concrete dam along a predetermined scanning path. During the scan, the sonar images and equipment status are monitored in real time to ensure the continuity and stability of data acquisition. After each scan, the acquired raw sonar image data is stored in a high-capacity storage device on the data acquisition terminal for subsequent processing.

[0044] II. Image Preprocessing

[0045] Adaptive Median Filtering: The acquired raw sonar image is read into image processing software. First, the image is analyzed pixel-by-pixel, calculating the statistical characteristics of the grayscale values ​​of each pixel and its neighboring pixels. An initial filtering window size is set, for example, 3×3. For each pixel, it is determined whether its grayscale value belongs to a noise point within its neighborhood. If the grayscale value of a pixel differs significantly from the grayscale values ​​of most pixels in its neighborhood, it is identified as a noise point. For noise points, the filtering window is increased, for example, to 5×5 or 7×7, and the median value of the pixels within the window is recalculated and used as the new grayscale value for the current pixel. For non-noise points, the original filtering window size is maintained. In this way, noise is removed while preserving the image's detail information to the maximum extent.

[0046] Histogram equalization: After adaptive median filtering, histogram equalization is performed on the image. The number of pixels n at each gray level in the image is counted. j Calculate the total number of pixels, n, in the image. According to the formula... Calculate the grayscale value S after equalization for each grayscale level in sequence. k The grayscale value of each pixel in the original image is replaced according to the calculated equalization mapping relationship, thereby making the grayscale distribution of the image more uniform, enhancing the image contrast and visual effect, and facilitating subsequent feature extraction and analysis.

[0047] III. Block Drop Feature Extraction

[0048] Area Feature: The preprocessed sonar image is binarized to separate potentially missing regions from the background. Image segmentation algorithms, such as thresholding, are used to determine a suitable threshold based on the image's grayscale characteristics. Pixels with grayscale values ​​greater than the threshold are designated as foreground (missing regions), while those less than the threshold are designated as background. Then, pixel counting is used to calculate the number of pixels within the missing regions, and this number is converted into an actual area feature A based on the image resolution.

[0049] Shape characteristics: For the segmented missing area, calculate its perimeter P. According to the formula... Calculate the roundness C, which reflects how closely the shape of the chipped area resembles a circle. Simultaneously, calculate the curvature variation at points along the boundary of the chipped area to statistically determine the boundary roughness characteristic R. For example, use chain code to encode the boundary of the chipped area, and then calculate the angular variation between adjacent chain codes to measure the boundary roughness.

[0050] IV. Crack Feature Extraction

[0051] Edge detection: Edge detection algorithms are used to perform edge detection on the preprocessed image. During the detection process, the parameters of the Gaussian filter are appropriately selected based on the noise level of the image and the characteristics of the cracks to smooth the image and reduce the impact of noise on edge detection.

[0052] Parameter Calculation: For the extracted crack edges, a curve fitting method, such as the least squares method, is used. The mathematical expression of the crack is obtained through fitting, and then parameters such as the crack length L, width W, direction α, and curvature K are calculated. For example, for a straight crack segment, the length can be obtained by calculating the distance between the two endpoints, and the width can be calculated by the pixel distribution perpendicular to the crack direction; for a curved crack, the corresponding parameters are calculated based on the parameters of the fitted curve.

[0053] V. Disease Identification

[0054] Sample Collection and Labeling: A large number of sonar image samples of underwater spalling and cracks in concrete dams of different types and degrees were collected, including image samples of normal areas. Experts in the field of hydraulic engineering were invited to label these samples in detail, clarifying whether spalling or cracks exist in each sample, as well as information such as the type, location, and approximate size of the defects.

[0055] Model Training: The labeled samples are divided into training and test sets, typically in a 7:3 or 8:2 ratio. A disease identification model is constructed using the Support Vector Machine (SVM) algorithm. During training, a suitable kernel function, such as the Radial Basis Function (RBF), is selected. By adjusting the kernel function parameters and penalty factor, the SVM model is repeatedly trained and optimized using the training set. Simultaneously, cross-validation methods, such as 10-fold cross-validation, are employed. The training set is divided into ten subsets, with nine subsets used for training and one subset used for validation each time, repeated ten times to improve the model's generalization ability and accuracy.

[0056] Disease identification: The extracted feature vectors of the image to be detected are input into a trained SVM model. Based on the classification decision boundary obtained during training, the model determines the category to which the feature vector belongs, i.e., whether it is a chipped area, crack, or normal region. The identification results are output and ranked according to the model's confidence level. Results with low confidence levels can be manually reviewed or further analyzed.

[0057] VI. Disease Location

[0058] Sonar equipment positioning: During data acquisition, underwater positioning systems, such as Ultra-Short Baseline (USBL) or Long Baseline (LBL) systems, are used to acquire the real-time underwater location information of the sonar equipment, including latitude and longitude coordinates and depth information. This location information is then associated with and stored in conjunction with the acquired sonar images for subsequent lesion localization.

[0059] Feature point matching: In sonar images acquired at different locations, feature point matching algorithms, such as Scale Invariant Feature Transform (SIFT) or Speed-Up Robust Feature Transform (SURF), are used to find the correspondence between disease feature points in different images. The angles α1 and α2 of the disease feature points in different images, as well as the distance b between the two sonar device positions, are determined.

[0060] Location calculation: Based on the principles of triangulation, using the formula:

[0061]

[0062] Calculate the location coordinates (x, y) of the defect in a two-dimensional coordinate system. Combine this with depth information from images acquired by sonar equipment to determine the defect's specific location in three-dimensional space. Mark the calculated location information on an electronic map or dam model to visually display the defect's location.

[0063] VII. Disease Assessment

[0064] Standards are established based on relevant national and industry standards for water conservancy engineering testing, such as the "Code for Design of Hydraulic Concrete Structures" and the "Regulations for Quality Inspection and Evaluation of Water Conservancy and Hydropower Engineering Construction." Various assessment indicators and thresholds for spalling and cracking are determined. For example, for spalling, its hazard level is classified as mild, moderate, or severe based on the area and depth of the spalling; for cracks, the degree of impact on the stability of the dam structure is determined based on the length, width, and depth of the crack, combined with the structural characteristics and design requirements of the dam.

[0065] Parameter Calculation and Comparison: Based on the parameters obtained from feature extraction and hazard localization, such as the area and depth of the spalled area and the length, width, and depth of the cracks, these parameters are compared with thresholds in the standard. For spalled areas, if the area is less than a certain threshold and the depth is shallow, it is judged as a minor hazard; if the area and depth exceed the corresponding thresholds, its hazard level is further assessed. For cracks, considering all parameters of the crack, a crack severity index is calculated, and the degree of hazard is determined by comparing it with the severity levels in the standard.

[0066] Assessment Report Writing: Based on the assessment results, a detailed disease assessment report will be written. The report will include the type, location, parameters, severity level, and impact analysis on the stability of the dam structure. Corresponding treatment recommendations will also be proposed, such as regular monitoring for mild diseases and timely remedial measures for moderate and severe diseases.

[0067] 8. Generate test report

[0068] Report Template Design: Design a standardized test report template in PDF or XML format for easy storage and transmission. The template includes a cover, table of contents, overview of the test items, test methods, test results, disease assessment, treatment recommendations, and appendices.

[0069] Data Encapsulation: Fill in the sonar images collected during the detection process, extracted feature parameters, disease identification results, location information, and assessment reports according to the report template requirements. In the report, provide appropriate annotations and explanations for the sonar images to visually demonstrate the disease situation. Present feature parameters and assessment results in tables and charts for easy reading and analysis.

[0070] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for detecting underwater rockfalls and cracks in concrete dams based on image sonar, characterized in that, Includes the following steps: S1. Data Acquisition: An image sonar device is used to perform a 360-degree scan of the underwater portion of the concrete dam. The transmission frequency f and scanning angle θ of the sonar device are dynamically adjusted according to the dam structure and water flow conditions to acquire high-resolution sonar image data. The transmission frequency f satisfies the following relationship with the detection distance d and the speed of sound in water c: S2. Image Preprocessing: The acquired sonar images are sequentially subjected to adaptive median filtering for noise reduction and histogram equalization for enhancement. The adaptive median filtering dynamically adjusts the filter window size based on local image features. The histogram equalization formula is: S3. Feature Extraction: Extract area feature A and shape feature C (circularity) for the dropped block. P represents the perimeter of the region and the edge roughness feature R; for cracks, features such as length L, width W, direction α, and curvature K are extracted. S4. Disease identification: The Support Vector Machine (SVM) algorithm is used to train the SVM model using labeled sonar image samples. The extracted feature vectors are input into the trained SVM model to identify shards, cracks, and normal areas. S5. Disease Location: Based on the location information from the sonar equipment and the characteristic location of the disease in the image, the location coordinates (x, y) of the disease in the plane coordinate system are calculated using the following formula based on the principle of triangulation: The position of the sonar device in three-dimensional space is determined by combining depth information, where α1 and α2 are the angles corresponding to the disease feature points in the images obtained by the sonar device at different positions, and b is the distance between the two positions of the sonar device. S6. Disease Assessment: Based on the identified area and depth of the fallen pieces, as well as the length, width, and depth of the cracks, assess the severity of the disease according to the water conservancy engineering testing standards. S7. Generate Detection Report: Generate a detection report in a visual manner, including sonar images, feature parameters, and evaluation results, based on the detected disease information, including disease type, location, and severity.

2. The method for detecting underwater blockages and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the data acquisition step, the scanning angle θ is adjusted according to the complexity of the dam surface and the expected detection area to ensure coverage of the entire detection area.

3. The method for detecting underwater spalling and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the image preprocessing step, the adaptive median filter uses a larger window for filtering areas with more noise and a smaller window for areas with rich details.

4. The method for detecting underwater spalling and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the feature extraction step, the edge information of the crack is obtained by the edge detection algorithm, and the length L, width W, direction α and curvature K parameters of the crack are calculated by the curve fitting method.

5. The method for detecting underwater spalling and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the disease identification step, during SVM model training, cross-validation is used to optimize model parameters and improve the model's generalization ability.

6. The method for detecting underwater blockfalls and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the aforementioned disease assessment steps, for slabs, the hazard level of the slabs is determined by comparing their area and depth with standard thresholds; for cracks, the impact on the stability of the dam structure is comprehensively assessed based on the length, width, and depth parameters of the cracks.

7. The method for detecting underwater spalling and cracks in concrete dams based on image sonar according to claim 1, characterized in that, In the step of generating the test report, the test report is output in PDF or XML format for easy storage and transmission.

Images