Concrete crack early warning method and system

Abstract

The invention provides a concrete crack early warning method and system, and belongs to the technical field of information. The method comprises the following steps: acquiring temperature data of a concrete surface by using infrared thermal imaging equipment, decomposing signals through discrete wavelet transform, and respectively extracting high-frequency and low-frequency thermal signal components; in combination with an adaptive frequency division technology, a signal decomposition scale is dynamically adjusted, frequency overlapping of a micro crack and a large crack is avoided, and crack characteristics are accurately identified. Through calculation of signal intensity, temperature change rate and frequency domain analysis, processing of high-frequency and low-frequency signals is optimized, and temperature anomaly features of the crack are extracted. And further adopting a K nearest neighbor algorithm to classify the cracks, and generating a detection report containing crack types, spatial positions, geometric dimensions and detection confidence. The method can effectively overcome the limitation of a traditional crack detection method, is suitable for crack monitoring and early warning in a complex environment, and has high engineering application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, in particular to a concrete crack early warning method and system based on infrared thermal imaging technology. The method is suitable for health monitoring and crack detection of concrete structures, especially in complex environmental conditions, and can accurately identify small cracks and large cracks, providing effective technical support for safety evaluation and maintenance of concrete structures. BACKGROUND

[0002] Concrete structures are widely used in construction, bridges, tunnels and other engineering projects, and their structural safety is directly related to public safety. Cracks are one of the most common diseases of concrete structures, and the occurrence of cracks not only affects the strength and durability of the structure, but also may cause secondary hazards such as water penetration and corrosion, so timely detection and repair of cracks is crucial for structural health monitoring.

[0003] Traditional concrete crack detection methods mainly rely on manual inspection, visual detection and local sensor monitoring, which have the disadvantages of tedious operation, limited detection range, poor accuracy, etc. With the development of technology, infrared thermal imaging technology has gradually become an important tool in crack detection due to its advantages of non-destructive testing, non-contact and high efficiency. Infrared thermal imaging can detect temperature anomalies caused by cracks by detecting the temperature distribution on the surface of the structure, but in complex environments, the overlap of temperature responses, noise interference and the diversity of crack morphology make it difficult for traditional crack detection methods to accurately distinguish between small cracks and large cracks, affecting the accuracy and reliability of the detection results.

[0004] In recent years, with the development of signal processing technology, discrete wavelet transform (DWT) has been applied to crack detection, which effectively extracts the temperature variation characteristics of cracks through high-frequency and low-frequency decomposition. However, the existing method still has a technical problem: the fixed decomposition scale in discrete wavelet transform cannot adapt to the temperature response of different crack sizes (especially small cracks), resulting in overlapping frequency characteristics of different crack types, affecting the detection accuracy.

[0005] To solve the above problems, the present application proposes a concrete crack early warning method based on infrared thermal imaging, which combines adaptive frequency division technology and dynamically adjusts the signal decomposition scale to avoid frequency overlap, thereby achieving the goal of accurately distinguishing between small cracks and large cracks. This method not only improves the accuracy of crack detection, but also enhances the robustness and real-time performance of the system in complex environments, and has important engineering application value. SUMMARY

[0006] The application provides a concrete crack early warning method and system, aiming to solve the technical problems of difficult crack feature extraction under infrared thermal imaging conditions, influence of frequency overlap on identification accuracy, and unstable detection results in dynamic environment. The method is based on the spatiotemporal characteristics of the temperature field, fuses multiple types of signal processing algorithms and intelligent classification algorithms, and constructs a complete crack identification and early warning process, with high precision, high robustness and good real-time performance.

[0007] In a first aspect, the application provides a concrete crack early warning method, which comprises: Step 1: obtaining original temperature data from an infrared thermal imaging device to generate a first temperature distribution matrix; Step 2: performing discrete wavelet transform on the first temperature distribution matrix, using Daubechies wavelet basis function to decompose the signal to obtain high-frequency thermal signal components and low-frequency thermal trend components; Step 3: calculating the signal intensity value based on the high-frequency thermal signal components, and if the signal intensity value is less than a preset threshold, performing adaptive median filtering on the high-frequency thermal signal components to obtain a first processed signal; Step 4: calculating the temperature change rate based on the low-frequency thermal trend components, and if the temperature change rate exceeds a preset range, performing fast Fourier transform and high-pass filtering on the low-frequency thermal trend components to obtain a second processed signal; Step 5: reconstructing the first processed signal and the second processed signal by wavelet inverse transform to generate a second temperature distribution matrix, which contains temperature anomaly features of crack regions; Step 6: performing frequency domain analysis on the second temperature distribution matrix to separate the thermal features of micro cracks and large cracks, and constructing a comprehensive crack feature vector, which contains the width, depth and temperature difference parameters of the cracks; Step 7: using K-nearest neighbor algorithm to match the comprehensive crack feature vector with a preset crack template library to obtain crack classification data; Step 8: using a sliding window algorithm to process time-continuous frame data according to the crack classification data to generate a crack detection report, which contains crack type identification, spatial coordinate position, geometric size value and detection confidence value.

[0008] In a second aspect, the application provides a concrete crack early warning system, which comprises: An infrared thermal imaging device is used to obtain original temperature data from the infrared thermal imaging device to generate a first temperature distribution matrix; A signal decomposition unit is used to perform discrete wavelet transform on the first temperature distribution matrix, using Daubechies wavelet basis function to decompose the signal to obtain high-frequency thermal signal components and low-frequency thermal trend components; An adaptive filtering unit is configured to calculate a signal intensity value based on the high-frequency thermal signal component, and perform adaptive median filtering on the high-frequency thermal signal component to obtain a first processed signal if the signal intensity value is less than a preset threshold value; A Fourier transform unit is configured to calculate a temperature change rate based on the low-frequency thermal trend component, and perform fast Fourier transform and high-pass filtering on the low-frequency thermal trend component to obtain a second processed signal if the temperature change rate exceeds a preset range; A reconstruction unit is configured to reconstruct the first processed signal and the second processed signal by inverse wavelet transform to generate a second temperature distribution matrix, which contains temperature anomaly characteristics of the crack region; A crack feature extraction unit is configured to perform frequency domain analysis on the second temperature distribution matrix to separate thermal characteristics of micro cracks and large cracks, and construct a comprehensive crack feature vector containing width, depth and temperature difference parameters of the cracks; A crack classification unit is configured to match the comprehensive crack feature vector with a preset crack template library using a K-nearest neighbor algorithm to obtain crack classification data; A report generation unit is configured to process time-continuous frame data using a sliding window algorithm based on the crack classification data to generate a crack detection report, which contains crack type identification, spatial coordinate position, geometric size value and detection confidence value.

[0009] Compared with the prior art, the technical scheme of the present application has at least the following advantages: 1. By combining infrared thermal imaging technology and wavelet transform technology, the temperature anomaly characteristics of the crack region can be accurately extracted, which plays a positive role in accurate identification of cracks in complex environments. By decomposing high-frequency and low-frequency signals through wavelet transform, errors caused by noise are effectively eliminated, and the accuracy of crack detection is improved.

[0010] 2. By using the K-nearest neighbor algorithm and frequency domain analysis technology, different types of cracks can be accurately classified. For micro cracks and large cracks, the system can independently extract their different thermal response characteristics, and accurately classify the cracks based on these characteristics, greatly improving the reliability of crack classification.

[0011] 3. By dynamically adjusting the signal decomposition scale and frequency domain analysis, the characteristics of cracks can still be accurately extracted and classified under complex environmental conditions, especially in the case of heavy noise interference, which has high robustness. In addition, the sliding window algorithm is used to realize real-time processing of continuous frame data, which can timely capture the trend of crack changes and issue early warnings, ensuring the real-time and efficiency of the crack monitoring system.

[0012] 4. Without destructive operation on the structure surface, combined with infrared thermal imaging equipment and intelligent algorithm, remote, non-contact and automatic crack monitoring can be realized, which is suitable for tunnels, bridges, building outer walls and other types of concrete structures, supports template library updating and algorithm online adjustment, and has good scalability and engineering adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0014] Figure 1 An embodiment of a concrete crack early warning method in the present application; Figure 2 An embodiment of a concrete crack early warning system in the present application. DETAILED DESCRIPTION

[0015] The present application provides a concrete crack early warning method and system. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "includes" or "has" and any variation thereof is intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0016] For the sake of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of a concrete crack early warning method in the present application includes: Step 1, obtaining original temperature data from an infrared thermal imaging device to generate a first temperature distribution matrix.

[0017] In a specific embodiment, the process of step 1 can specifically include the following steps: (1) The original temperature data is converted into digital temperature signal by analog-digital conversion, and the digital temperature signal is subjected to mean filtering denoising treatment to obtain smooth temperature data; (2) According to the smoothed temperature data, a first temperature distribution matrix is generated according to the pixel position, and each matrix element represents the temperature value of the corresponding pixel position; (3) If there are missing values in the first temperature distribution matrix, the temperature values of the missing pixels are calculated by an interpolation algorithm to obtain a completed temperature distribution matrix as the updated first temperature distribution matrix.

[0018] Specifically, in order to make the data more suitable for subsequent processing, it is necessary to first perform analog-to-digital conversion on the original temperature data to convert the analog signal into a digital temperature signal. The digital signal usually contains noise and irregular fluctuations, which need to be denoised by a mean filtering algorithm. Mean filtering reduces the influence of high-frequency noise and smooths the temperature data, thereby obtaining more stable and accurate temperature values. Based on the denoised and smoothed temperature data, a first temperature distribution matrix is generated according to the pixel position, and each element in the matrix represents the temperature value at a specific position. These temperature values come from different pixel points captured by the infrared thermal imaging device. Each pixel position corresponds to a specific temperature value, forming a two-dimensional matrix with spatial distribution characteristics, which can reflect the temperature distribution of the entire region. In the actual temperature data collection process, the infrared thermal imaging device may have some missing values or incomplete temperature data due to device resolution limitations or environmental interference. In order to solve this problem, interpolation algorithms are used to fill in the missing pixels. The interpolation algorithm usually uses linear interpolation or other appropriate methods to ensure that the generated temperature data has high precision and consistency.

[0019] In a preferred embodiment, if the resolution of the first temperature distribution matrix is lower than the preset image accuracy threshold, a thermal image super-resolution reconstruction process is performed on the first temperature distribution matrix, and the super-resolution reconstruction process includes: (1) Constructing a multi-scale image representation model based on the temperature gradient change relationship between pixels, the multi-scale image representation model being used to represent temperature variation characteristics at different scales; (2) Training and generating a temperature image enhancement network using the multi-scale image representation model, and performing reconstruction processing on the first temperature distribution matrix based on the temperature image enhancement network to output a new first temperature distribution matrix.

[0020] Specifically, when the resolution of the first temperature distribution matrix is lower than the preset image accuracy threshold, the image resolution is improved through a thermal image super-resolution reconstruction process. The key of this process is to construct a multi-scale image representation model based on the temperature gradient change relationship between pixels. This model can capture temperature change features at different levels through different scale image representations, so that key temperature features can be preserved at different resolutions, especially in the case of low image resolution, the temperature change of the crack area can still be accurately reflected. By using the multi-scale image representation model, a temperature image enhancement network is further constructed, which can effectively reconstruct the first temperature distribution matrix after training. In this process, the network converts low-resolution information of the image into a higher resolution image, and restores the details and temperature features in the image through iterative optimization. In the reconstruction process, it is not just a simple pixel expansion, but through the network to learn the relationship between temperature gradient and spatial structure, to enhance the details in the image, so as to improve the quality and information accuracy of the reconstructed image.

[0021] After processing by the temperature image enhancement network, the output new first temperature distribution matrix has higher resolution than the original matrix, and can better reflect the temperature anomaly features of the crack area, thereby providing more accurate data support for subsequent crack identification and classification. Through the super-resolution technology, the details lost in the low-resolution image are filled, and the quality of the image is effectively improved, so that the crack detection can still achieve high-precision identification effect under low-resolution input data.

[0022] The present application can effectively solve the problems of image blur and detail loss caused by low resolution of original temperature data, improve the accuracy of crack detection, and avoid missed detection or false detection caused by insufficient image resolution when processing large-scale monitoring areas.

[0023] In another preferred embodiment, the super-resolution reconstruction process further comprises: (1) constructing a physical model based on the relationship between the physical law of crack growth and temperature change; (2) combining the physical model with the temperature image enhancement network, introducing the physical law as a regularization term in the network training process to optimize the learning process of the network, so that the output high-resolution temperature image conforms to the temperature change relationship of crack growth; (3) in the reconstruction process, the hypothesis of temperature change is constrained by using the physical model to ensure that the generated high-resolution temperature image not only conforms to the statistical characteristics of the image, but also can reflect the temperature distribution in the crack growth.

[0024] Although the super-resolution reconstruction technology can improve the resolution of the image, it is still based on the existing low-resolution image to infer the detail information, therefore, there may be missing or error in some details, especially when the details of the crack in the low-resolution image are very blurred, the reconstructed high-resolution image may not completely restore these details. Based on this, the reconstruction is carried out on the basis of the physical law of temperature change and crack growth, and the accuracy of crack feature reconstruction is improved by combining the physical model with the data-driven super-resolution method.

[0025] Specifically, the crack growth will affect the thermal conductivity characteristics and temperature distribution of the material, and with the expansion of the crack, the temperature gradient of the surrounding area will change significantly. The physical model constructed based on these physical laws can describe the temperature change trend in the crack growth process and provide a quantitative relationship between the crack and the temperature. The above-mentioned physical model simulates the change law of temperature distribution in the crack growth process, and can accurately capture the mutual relationship between the temperature evolution of the crack and its growth state.

[0026] According to the influence of crack growth on temperature gradient, the physical model considers the temperature distribution change caused by the crack, and can accurately model the temperature evolution of the crack area through the heat conduction equation. The heat conduction equation is: ; Where T(x,t) is the temperature of the crack area, a is the thermal diffusivity, ∇ 2 T is the temperature gradient, and Q(x,t) is the heat source term caused by the crack. Through this model, the dynamic change of temperature in the crack growth process can be accurately described.

[0027] In order to effectively integrate the physical model into the temperature image enhancement network, a method combining physical model and deep learning is adopted. In the training process, the physical law is introduced as a regularization term in the loss function of the temperature image enhancement network, and the learning process of the network is optimized. The regularization term constrains the high-resolution temperature image output by the network to satisfy the temperature change relationship defined in the physical model, so that the network not only learns the statistical characteristics of the image, but also follows the physical law of temperature change in the crack growth process. In this way, the output of the network not only has higher resolution, but also can accurately reflect the temperature distribution characteristics of the crack, thereby solving the problem that the physical law may be ignored in the temperature image enhancement process in the traditional method.

[0028] By adding the constraint term of the physical model in the loss function of the network, the high-resolution temperature image output by the network not only conforms to the statistical characteristics of the image. The loss function L total includes the image reconstruction error and the regularization term of the physical model, and the specific form is: ; wherein, is a weight coefficient, L reconstruction is a reconstruction error, usually the mean square error (MSE): ; L physicalL is a regularization term of physical law, which ensures that the reconstructed image satisfies the physical law of crack growth. This term can constrain the generated image temperature gradient by introducing the heat conduction equation of crack growth, expressed as: ; wherein, ∇ 2 T SR represents the temperature gradient of the super-resolution image generated by the network, and Q(x,t) is the heat source term caused by the crack. By optimizing this loss function, the network can generate high-resolution images while following the physical law of temperature change during crack growth.

[0029] In the image reconstruction process, the physical model constrains the assumption of temperature change, ensuring that the generated high-resolution temperature image not only meets the statistical characteristics of the image, but also reflects the temperature distribution characteristics in the crack growth. In this way, the generated high-resolution temperature image can accurately reflect the physical characteristics of the crack area while recovering details, avoiding artifacts or unreasonable temperature distribution that may be caused by relying only on image features.

[0030] By combining the physical model with the deep learning network, not only the resolution of the temperature image is effectively improved, but also the consistency of the reconstructed image in physical rationality is ensured, thereby improving the precision and robustness of the crack detection system in practical application. This method is especially suitable for crack monitoring in complex environments, and can accurately reflect the influence of crack growth on temperature distribution, providing more accurate support for structural health monitoring.

[0031] Step 2, perform discrete wavelet transform on the first temperature distribution matrix, and use Daubechies wavelet basis function to decompose the signal to obtain high-frequency thermal signal components and low-frequency thermal trend components.

[0032] In a specific embodiment, during signal decomposition, different frequency decomposition scales are dynamically selected according to the temperature change characteristics of the crack area, including: For small cracks, based on their high-frequency temperature change characteristics, the decomposition scale is automatically adjusted to the first frequency range to accurately extract the thermal response characteristics of small cracks; For large cracks, based on their low-frequency temperature change characteristics, the decomposition scale is automatically adjusted to the second frequency range to accurately extract the temperature trend characteristics of large cracks, wherein the first frequency range is higher than the second frequency range; By analyzing the thermal response amplitude and temperature change rate of the crack in real time, whether the high-frequency characteristic change of the micro crack and the low-frequency trend characteristic of the large crack have significant fluctuations is judged respectively. Based on the above analysis results, the frequency scale range of signal decomposition is dynamically adjusted to optimize the frequency domain separation effect of micro crack and large crack characteristics.

[0033] Specifically, through wavelet transform, the original temperature data can be decomposed into components of different frequencies, so that the high-frequency component can capture the subtle changes of the temperature change in the crack area, and the low-frequency component can reveal the overall trend of the temperature change. For the crack area, the temperature change of the micro crack usually shows a relatively sharp high-frequency component, while the large crack shows a relatively flat low-frequency change.

[0034] The high-frequency and low-frequency components after discrete wavelet transform are decomposed based on fixed scales, however, different sizes of cracks (especially micro cracks) may have frequency overlap in temperature response, that is, micro cracks and large cracks may have similar frequency characteristics, especially under complex environmental conditions, noise or other factors may cause signal overlap of different crack types. Therefore, during signal decomposition, different frequency decomposition scales are dynamically selected according to the temperature change characteristics of the crack area. For example, for micro cracks, based on their high-frequency temperature change characteristics, by analyzing the thermal response amplitude and temperature change rate of the crack, the decomposition scale is automatically adjusted to a higher frequency range, so that the network can accurately extract the thermal response characteristics of the micro crack. This higher frequency range can accurately capture the thermal disturbance and rapid change characteristics of the micro crack. For large cracks, temperature changes usually show a relatively smooth trend, so the decomposition scale is adjusted to a lower frequency range to accurately extract the temperature trend characteristics of the large crack. This lower frequency range can effectively capture the heat diffusion trend and slow-changing temperature characteristics in the crack propagation process.

[0035] During real-time monitoring of the crack, the frequency decomposition scale is dynamically adjusted according to the thermal response amplitude and temperature change rate of the crack to effectively distinguish the characteristics of micro cracks and large cracks. Specifically, when the thermal response amplitude of the crack area is significantly changed, or the temperature change rate of the crack is suddenly changed, the system will adjust the decomposition scale accordingly to optimize the extraction effect of high-frequency and low-frequency components. In this way, different characteristics of temperature change can be accurately extracted when dealing with different types of cracks, and reliable basis can be provided for subsequent crack classification and identification.

[0036] According to the different characteristics of the cracks, the decomposition scale is dynamically adjusted, effectively improving the frequency domain separation effect of micro cracks and large cracks. Through this adaptive signal decomposition method, the different frequency components of the temperature image can be fully mined, thereby improving the extraction accuracy of the crack feature of the crack detection system. In practical application, especially when dealing with a large monitoring area, the thermal response characteristics of the cracks can be more accurately analyzed, and the misjudgment or omission problem caused by signal frequency overlap can be avoided. This method can also adapt to changes in different environments and crack types, providing a more robust and efficient crack monitoring means to meet the high-precision requirements in complex structure monitoring.

[0037] Step 3, calculating the signal intensity value based on the high-frequency thermal signal component, if the signal intensity value is less than the preset threshold, performing adaptive median filtering on the high-frequency thermal signal component to obtain a first processed signal.

[0038] Specifically, by analyzing the amplitude of the high-frequency thermal signal component, the intensity value of the signal can be obtained. If the signal intensity value is less than the preset threshold, it indicates that this part of the signal may be affected by noise or interference and cannot provide effective crack feature information, then adaptive median filtering is performed on the high-frequency signal component. Median filtering is a technique commonly used for noise removal, which can effectively remove isolated noise points in the image without affecting the overall structure of the image, and is especially suitable for processing signals containing salt and pepper noise. In the adaptive median filtering process, the filter dynamically adjusts the window size according to the local characteristics of the signal to ensure the optimization of the filtering effect.

[0039] By performing adaptive median filtering on the high-frequency signal component, the system obtains a first processed signal, which can remove noise while retaining effective crack thermal response characteristics. By accurately extracting the high-frequency characteristics of the cracks, the accuracy and reliability of crack detection are further improved, especially in the case of small cracks or weak temperature changes, valuable crack information can still be effectively identified and extracted, enhancing the stability and robustness of the system in complex environments.

[0040] Step 4, calculating the temperature change rate based on the low-frequency thermal trend component, if the temperature change rate exceeds the preset range, performing fast Fourier transform and high-pass filtering on the low-frequency thermal trend component to obtain a second processed signal.

[0041] In a specific embodiment, after performing fast Fourier transform and high-pass filtering on the low-frequency thermal trend component, Gaussian smoothing filtering is performed on the second processed signal to calculate its corresponding temperature change rate, if the temperature change rate still exceeds the preset range, the standard deviation parameter of the Gaussian filter is dynamically adjusted based on the current change rate, the smoothing process is re-performed, and the iteration is updated until the temperature change rate meets the preset range, generating a new second processed signal.

[0042] Specifically, the low-frequency thermal trend component represents the overall change trend of temperature, which is usually used to capture the temperature change in the crack propagation process. When the crack propagates, the change trend of temperature may exhibit irregular or excessive fluctuations, which may be caused by environmental noise, sensor errors or interference of other external factors. These abnormal fluctuations make the temperature change rate exceed the preset range, affecting the accuracy of the signal.

[0043] When the temperature change rate exceeds the preset range, the low-frequency and high-frequency components in the signal can be effectively separated through frequency domain conversion by FFT, so as to remove irrelevant low-frequency components and retain meaningful high-frequency information. By high-pass filtering, the main features of the signal can be retained and the fine noise can be removed. In the case of relatively stable temperature change, smoothing filtering can ensure the accuracy of the signal. Based on this, the temperature trend characteristics of the crack region can be extracted, the low-frequency noise component in the temperature signal can be removed, and the long-term trend of temperature change can accurately reflect the thermal diffusion characteristics of the crack.

[0044] In some cases, after smoothing filtering, the temperature change rate may still exceed the preset range, indicating that there is still excessive noise or irregular fluctuations in the signal. Therefore, the system dynamically adjusts the standard deviation parameter of the Gaussian filter based on the current temperature change rate, re-performs smoothing processing, and iteratively updates until the temperature change rate meets the preset range. This process can flexibly adjust the parameters of the filter under different environmental conditions to adapt to the temperature change characteristics of different crack regions, ensuring that the second processed signal generated finally can accurately reflect the temperature change trend of the crack. Preferably, an iteration upper limit number is set, and the iteration is stopped when the iteration upper limit number is reached.

[0045] Through the above technical means, the present application can optimize the signal processing process of the low-frequency thermal trend component in real time, effectively remove noise and irrelevant interference signals, while retaining important temperature characteristics in the crack growth process. This method is particularly suitable for processing in the crack region where the temperature changes slowly. By dynamically adjusting the filter parameters, the accurate capture of temperature change is ensured, avoiding false positives and missed detections, and improving the robustness and accuracy of the crack monitoring system.

[0046] Step 5, reconstructing the first processed signal and the second processed signal by wavelet inverse transform to generate a second temperature distribution matrix, the second temperature distribution matrix containing temperature anomaly characteristics of the crack region.

[0047] Specifically, the core of this process is to restore the signal that has been processed by wavelet transform into the original temperature distribution matrix through wavelet inverse transform, so as to restore the temperature anomaly characteristics of the crack area. In wavelet transform, the signal is decomposed into different frequency components, the first processed signal contains the high-frequency components filtered and processed, and the second processed signal contains the low-frequency temperature change characteristics. Through wavelet inverse transform, the system recombines these processed signals and converts them into a complete temperature distribution image with temperature change characteristics.

[0048] Wavelet inverse transform can recombine the processed high-frequency and low-frequency signals into an integral temperature distribution matrix, ensuring that the reconstructed signal can retain the temperature anomaly characteristics of the crack area to the greatest extent. These anomaly characteristics include temperature changes caused by cracks, such as changes in heat distribution caused by crack propagation, which can effectively reflect the position and size of the crack in the structure. In this way, the reconstructed second temperature distribution matrix not only restores the overall temperature change, but also effectively highlights the temperature anomaly characteristics of the crack area, providing accurate data support for subsequent crack identification and analysis.

[0049] Step 6, performing frequency domain analysis on the second temperature distribution matrix, separating the thermal characteristics of small cracks and large cracks, and constructing a comprehensive crack feature vector containing the width, depth, and temperature difference parameters of the crack.

[0050] In a specific embodiment, the process of performing step 6 can specifically include the following steps: Performing Fourier transform on the second temperature distribution matrix to obtain frequency spectrum data; Extracting high-frequency components and low-frequency components from the frequency spectrum data, wherein the high-frequency components correspond to the thermal response characteristics of small cracks, and the low-frequency components correspond to the temperature trend characteristics of large cracks; Extracting the thermal response amplitude and temperature change rate of small cracks based on the high-frequency components, and constructing a small crack feature vector; Extracting the temperature change slope and fluctuation range of large cracks based on the low-frequency components, and constructing a large crack feature vector; Combining the small crack feature vector and the large crack feature vector to generate a comprehensive crack feature vector containing the width, depth, and temperature difference of the crack.

[0051] Specifically, Fourier transform is performed on the second temperature distribution matrix to obtain frequency spectrum data, and high-frequency and low-frequency components of the signal are extracted from the frequency spectrum data. The high-frequency components generally represent the thermal response characteristics of small cracks, and the low-frequency components represent the temperature change trend characteristics of large cracks.

[0052] When extracting the high-frequency component in the spectral data, first, the feature vector of the micro crack is constructed by analyzing its thermal response amplitude and temperature change rate. Micro cracks usually exhibit a more intense thermal response, and their high-frequency components can effectively reflect the instantaneous temperature change and rapid fluctuations during the crack propagation process. Therefore, by calculating the amplitude change and change rate of the high-frequency component, the thermal response characteristics of the crack can be effectively quantified. When constructing the feature vector of the micro crack, the system calculates the width, depth and temperature difference parameters of the crack based on the characteristics of the high-frequency component.

[0053] For large cracks, the low-frequency component in the spectrum mainly reflects the heat conduction trend during the crack propagation process. By analyzing the low-frequency component, the system extracts the slope and fluctuation range of the temperature change, and then constructs the feature vector of the large crack. The temperature change of the low-frequency component is relatively stable, but its slope and fluctuation range can accurately represent the size and temperature distribution characteristics of the crack, thereby providing effective information about the size and thermal characteristics of the crack.

[0054] Once the feature vectors of the micro crack and the large crack are constructed, the feature vectors of the two are combined to generate a comprehensive crack feature vector. This comprehensive crack feature vector contains the width, depth and temperature difference parameters of the crack, and fully reflects the thermal response and temperature change trend of the crack.

[0055] For example, the feature vector of the micro crack should contain its thermal response amplitude, temperature change rate, and geometric parameters (width, depth, temperature difference) of the crack. Assuming that the thermal response amplitude of the micro crack is A high = 0.05A, the temperature change rate is R high = 0.02R, the width, depth and temperature difference are 2.5 cm, 1.2 cm and 3.0°C respectively, then the feature vector of the micro crack is: V small = [2.5, 1.2, 3.0, 0.05, 0.02]; For the large crack, its feature vector should contain its temperature change slope, fluctuation range and geometric parameters (width, depth, temperature difference) of the crack. Assuming that the temperature change slope of the large crack is S low = 0.1, the fluctuation range is W low = 3.5W. The width, depth and temperature difference are 10 cm, 3.5 cm and 5°C respectively, then the feature vector of the large crack can be expressed as: Vlarge = [10.0, 3.5, 5.0, 0.1, 3.5].

[0056] The comprehensive crack feature vector combines the feature vectors of the micro crack and the large crack into one whole, containing multiple parameters such as the width, depth, temperature difference, thermal response amplitude, temperature change rate, temperature change slope and fluctuation range of the crack. The combined comprehensive crack feature vector is: Vtotal =[2.5,1.2,3.0,0.05,0.02,10.0,3.5,5.0,0.1,3.5]。

[0057] By effectively separating the high-frequency and low-frequency characteristics of the cracks through frequency domain analysis, the thermal characteristics of both micro-cracks and large cracks can be accurately extracted, ensuring that the characteristics of different types of cracks can be accurately identified, and avoiding errors caused by frequency domain overlap in traditional methods.

[0058] Step 7, using the K-Nearest Neighbor algorithm, match the comprehensive crack feature vector with the pre-set crack template library to obtain crack classification data.

[0059] Specifically, the comprehensive crack feature vector contains parameters such as crack width, depth, and temperature difference, which describe the geometric and thermal response characteristics of the crack. The K-Nearest Neighbor algorithm compares the crack features to be classified with known crack templates by calculating the similarity between the comprehensive crack feature vector and each crack template in the template library. Each template represents a different type of crack, and by measuring the distance or similarity, the KNN algorithm selects the K nearest neighbors with the shortest distance, and according to the classification of these neighbors, it determines the type of crack to be classified.

[0060] The system compares the comprehensive feature vector of the crack to be classified with the feature vectors of all crack templates in the template library, calculates the distance between each pair of feature vectors, and the commonly used distance measurement methods include Euclidean distance or Manhattan distance. Select K neighbors with the smallest distance, and according to the classification results of the neighbors, determine the type of crack to be classified. If most of the neighbors belong to a certain type of crack, the system classifies the crack into that type.

[0061] Through this process, the K-Nearest Neighbor algorithm can efficiently and accurately classify cracks and classify them into pre-set crack types.

[0062] Step 8, according to the crack classification data, use the sliding window algorithm to process time-continuous frame data, generate a crack detection report, the report contains crack type identification, spatial coordinate position, geometric size value and detection confidence value.

[0063] Specifically, the sliding window algorithm processes crack information frame by frame by sliding a fixed size window in continuous frame data to ensure that crack features can be fully analyzed in time and space.

[0064] By using the sliding window algorithm, the system can dynamically analyze each region of the crack detection, analyze the crack characteristics of each small region within the window, and determine the type of crack according to the crack classification data obtained in the previous steps. After processing each frame of data through the window, the system extracts the spatial coordinate position and geometric size of the crack, and matches it with the crack characteristics in the crack template library, thereby outputting the classification information of the crack. At the same time, based on the processing of consecutive frame data, the system can analyze the trend of the crack and calculate the geometric size of the crack at each corresponding time, including width, depth and temperature difference.

[0065] In addition to the geometric characteristics of the crack, the system also calculates the detection confidence of each crack feature. This confidence value is quantified according to the algorithm results and error range in the crack classification process, reflecting the accuracy of crack identification. By considering all processing results comprehensively, the system generates a detailed crack detection report, which not only contains the type and spatial coordinates of the crack, but also provides geometric size data and detection confidence of the crack, helping users to make more accurate crack evaluation and subsequent structure repair.

[0066] The sliding window algorithm can effectively process time-continuous frame data, ensuring the dynamics and spatio-temporal consistency of the crack detection process, avoiding information loss due to data changes or different crack expansion speeds. At the same time, by accurately extracting the geometric size and classification data of the crack, combined with the detection confidence, it can provide high-precision crack monitoring results, thereby supporting real-time warning and maintenance decision-making of the structure health monitoring system.

[0067] The above describes a concrete crack warning method in an embodiment of the present application, and the following describes a concrete crack warning system in an embodiment of the present application. Please refer to Figure 2 An embodiment of a concrete crack warning system in an embodiment of the present application includes: An infrared thermal imaging device 10 is configured to obtain original temperature data from the infrared thermal imaging device and generate a first temperature distribution matrix.

[0068] A signal decomposition unit 20 is configured to perform a discrete wavelet transform on the first temperature distribution matrix, decompose the signal using a Daubechies wavelet basis function, and obtain a high-frequency thermal signal component and a low-frequency thermal trend component.

[0069] An adaptive filtering unit 30 is configured to calculate a signal intensity value based on the high-frequency thermal signal component, and perform adaptive median filtering on the high-frequency thermal signal component if the signal intensity value is less than a preset threshold to obtain a first processed signal.

[0070] The Fourier transform unit 40 is configured to calculate a temperature change rate according to the low-frequency thermal trend component, and perform a fast Fourier transform and a high-pass filtering on the low-frequency thermal trend component if the temperature change rate is out of a preset range, to obtain a second processing signal.

[0071] The reconstruction unit 50 is configured to reconstruct the first processing signal and the second processing signal by inverse wavelet transform to generate a second temperature distribution matrix, and the second temperature distribution matrix contains temperature anomaly features of the crack area.

[0072] The crack feature extraction unit 60 is configured to perform a frequency domain analysis on the second temperature distribution matrix to separate thermal features of micro cracks and large cracks, and construct a comprehensive crack feature vector, and the comprehensive crack feature vector contains width, depth and temperature difference parameters of the cracks.

[0073] The crack classification unit 70 is configured to match the comprehensive crack feature vector with a preset crack template library by using a K-nearest neighbor algorithm to obtain crack classification data.

[0074] The report generation unit 80 is configured to process time-continuous frame data by using a sliding window algorithm according to the crack classification data to generate a crack detection report, and the report contains crack type identification, spatial coordinate position, geometric size value and detection confidence value.

[0075] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0076] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that makes contributions to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0077] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for early warning of concrete cracks, characterized in that, The method includes: Step 1: Obtain raw temperature data from the infrared thermal imaging device and generate the first temperature distribution matrix; Step 2: Perform discrete wavelet transform on the first temperature distribution matrix, and decompose the signal using the Daubechies wavelet basis function to obtain the high-frequency thermal signal component and the low-frequency thermal trend component; Step 3: Calculate the signal strength value based on the high-frequency thermal signal component. If the signal strength value is less than a preset threshold, perform adaptive median filtering on the high-frequency thermal signal component to obtain the first processed signal. Step 4: Calculate the temperature change rate based on the low-frequency thermal trend component. If the temperature change rate exceeds a preset range, perform a fast Fourier transform and high-pass filtering on the low-frequency thermal trend component to obtain a second processed signal. Step 5: Reconstruct the first processed signal and the second processed signal by wavelet inverse transform to generate a second temperature distribution matrix, which contains the temperature anomaly features of the crack region. Step 6: Perform frequency domain analysis on the second temperature distribution matrix to separate the thermal characteristics of micro-cracks and large cracks, and construct a comprehensive crack feature vector, which includes crack width, depth and temperature difference parameters. Step 7: Using the K-nearest neighbor algorithm, the comprehensive crack feature vector is matched with a preset crack template library to obtain crack classification data; Step 8: Based on the crack classification data, use the sliding window algorithm to process the time-continuous frame data and generate a crack detection report. The report includes crack type identifier, spatial coordinate location, geometric dimension value and detection confidence value.

2. The method for early warning of concrete cracks according to claim 1, characterized in that, Step 1 includes: The original temperature data is converted from analog to digital to obtain a digital temperature signal, and then subjected to mean filtering and noise reduction processing to obtain smooth temperature data. Based on the smoothed temperature data, a first temperature distribution matrix is ​​generated according to the pixel position, where each matrix element represents the temperature value at the corresponding pixel position. If there are missing values ​​in the first temperature distribution matrix, the temperature value of the missing pixel is calculated by interpolation algorithm to obtain the completed temperature distribution matrix, which is used as the updated first temperature distribution matrix.

3. The method according to claim 1, characterized in that, If the resolution of the first temperature distribution matrix is ​​lower than a preset image accuracy threshold, then thermal image super-resolution reconstruction processing is performed on the first temperature distribution matrix. The super-resolution reconstruction processing includes: A multi-scale image representation model is constructed based on the temperature gradient change relationship between pixels. The multi-scale image representation model is used to represent the temperature change features at different scales. The multi-scale image representation model is used to train and generate a temperature image enhancement network. Based on the temperature image enhancement network, the first temperature distribution matrix is ​​reconstructed, and a new first temperature distribution matrix is ​​output.

4. The method for early warning of concrete cracks according to claim 3, characterized in that, The super-resolution reconstruction process further includes: A physical model was constructed based on the relationship between the physical laws of crack growth and temperature changes. The physical model is combined with the temperature image enhancement network. By introducing the physical laws as regularization terms during the network training process, the learning process of the network is optimized so that the output high-resolution temperature image conforms to the temperature change relationship of crack growth. During the reconstruction process, the physical model is used to constrain the assumptions about temperature changes, ensuring that the generated high-resolution temperature image not only conforms to the statistical characteristics of the image, but also reflects the temperature distribution during crack growth.

5. A method for early warning of concrete cracks according to claim 1, characterized in that, During signal decomposition, different frequency decomposition scales are dynamically selected based on the temperature change characteristics of the crack region, including: For microcracks, based on their high-frequency temperature change characteristics, the decomposition scale is automatically adjusted to the first frequency range in order to accurately extract the thermal response characteristics of microcracks; For large cracks, based on their low-frequency temperature change characteristics, the decomposition scale is automatically adjusted to a second frequency range in order to accurately extract the temperature trend characteristics of large cracks, wherein the first frequency range is higher than the second frequency range. By analyzing the thermal response amplitude and temperature change rate of the cracks in real time, we can determine whether there are significant fluctuations in the high-frequency characteristic changes of micro-cracks and the low-frequency trend characteristics of large cracks. Based on the above analysis results, the frequency scale range of signal decomposition is dynamically adjusted to optimize the frequency domain separation effect of micro-crack and large-crack features.

6. The method according to claim 1, characterized in that, After performing a fast Fourier transform and high-pass filtering on the low-frequency thermal trend component, a Gaussian smoothing filter is applied to the second processed signal to calculate its corresponding temperature change rate. If the temperature change rate still exceeds the preset range, the standard deviation parameter of the Gaussian filter is dynamically adjusted based on the current change rate, and the smoothing process is repeated and iterated until the temperature change rate meets the preset range, generating a new second processed signal.

7. The method according to claim 1, characterized in that, Step 6 includes: Perform a Fourier transform on the second temperature distribution matrix to obtain the spectral data; High-frequency and low-frequency components are extracted from the spectral data, wherein the high-frequency components correspond to the thermal response characteristics of microcracks, and the low-frequency components correspond to the temperature trend characteristics of large cracks; Based on the high-frequency components, the thermal response amplitude and temperature change rate of the microcracks are extracted, and a feature vector of the microcracks is constructed. Based on the low-frequency components, the slope and fluctuation range of temperature change in large cracks are extracted, and a feature vector of large cracks is constructed. The microcrack feature vector is combined with the large crack feature vector to generate the composite crack feature vector, which includes crack width, depth and temperature difference.

8. A concrete crack early warning system, used to implement the method as described in any one of claims 1 to 7, characterized in that, The system includes: An infrared thermal imaging device is used to acquire raw temperature data from an infrared thermal imaging device and generate a first temperature distribution matrix. The signal decomposition unit is used to perform discrete wavelet transform on the first temperature distribution matrix and decompose the signal using the Daubechies wavelet basis function to obtain high-frequency thermal signal components and low-frequency thermal trend components. An adaptive filtering unit is used to calculate the signal strength value based on the high-frequency thermal signal component. If the signal strength value is less than a preset threshold, adaptive median filtering is performed on the high-frequency thermal signal component to obtain a first processed signal. The Fourier transform unit is used to calculate the temperature change rate based on the low-frequency thermal trend component. If the temperature change rate exceeds a preset range, a fast Fourier transform and high-pass filtering are performed on the low-frequency thermal trend component to obtain a second processed signal. The reconstruction unit is used to reconstruct the first processed signal and the second processed signal through wavelet inverse transform to generate a second temperature distribution matrix, the second temperature distribution matrix containing the temperature anomaly features of the crack region; The crack feature extraction unit is used to perform frequency domain analysis on the second temperature distribution matrix, separate the thermal features of micro cracks and large cracks, and construct a comprehensive crack feature vector, which includes crack width, depth and temperature difference parameters. The crack classification unit is used to match the comprehensive crack feature vector with a preset crack template library using the K-nearest neighbor algorithm to obtain crack classification data; The report generation unit is used to process time-continuous frame data using a sliding window algorithm based on the crack classification data to generate a crack detection report. The report includes crack type identifier, spatial coordinate location, geometric dimension value, and detection confidence value.

Images