A method and system for intelligent detection of building damage

Abstract

The present invention relates to the field of data processing technology, and specifically to a method and system for intelligent detection of building damage, including: obtaining historical echo signals and their corresponding defect categories, and training a classification model; wherein obtaining historical echo signals specifically includes: based on multiple acquisition frequencies, transmitting ultrasonic signals to different positions of a wall, and obtaining multiple historical echo signals at each acquisition frequency; using an optimal acquisition frequency to transmit ultrasonic signals to different positions of a wall, and obtaining real-time echo signals, wherein the optimal acquisition frequency is the acquisition frequency with the highest stability; calculating the acquisition frequency with the highest stability based on the similarity between two historical echo signals; inputting the real-time echo signal into a trained classification model, and outputting defect categories at different positions of the wall. The present invention solves the problem of low accuracy in damage detection of buildings.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a method and system for intelligently detecting building damage. Background Art

[0002] As the service life of buildings gradually increases, the safety and stability of building structures have become increasingly prominent and have become a core concern in the engineering field. Traditional building damage detection methods usually rely on manual inspection or data collection of a single device, such as simple visual observation, knock detection, or a single infrared thermal imaging technology. However, these methods not only have low detection efficiency, but are also easily affected by human experience and subjective judgment, resulting in large errors in the results. At the same time, for complex building structures or minor early damage, traditional methods are often difficult to accurately identify, and are prone to missed detection or false detection. In addition, when faced with complex damage patterns, such as cracks, cavities, or structural deformations, a single detection device is limited by detection capabilities and data analysis methods, and cannot fully reflect the actual damage of the building.

[0003] The patent application document with publication number CN117351013A discloses a system and method for intelligent detection of building damage. The patent application document generates a high-frequency enhanced thermal image sequence by numbering multiple areas of the building to be detected and collecting infrared thermal signals; extracts the pixel features of the thermal image using a local feature algorithm to generate a feature image, and extracts the contour of the feature image using a gradient vector algorithm to form a regional contour set; classifies the regional contour set into a damaged contour subset and a non-damaged contour subset, and further calculates the entropy value of the thermal area image corresponding to the damaged contour subset, thereby identifying the extreme damage of the area to be detected.

[0004] However, although the above technical solution realizes intelligent detection of building damage, it does not fully consider the impact of the difference in echo signals at different acquisition frequencies on the detection results, resulting in low accuracy in building damage detection. Summary of the invention

[0005] In order to solve the problem of low accuracy in damage detection of buildings raised in the above background technology, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for intelligent detection of building damage, comprising:

[0007] Obtain historical echo signals and their corresponding defect categories, and train the classification model; wherein, obtaining historical echo signals specifically includes: based on multiple acquisition frequencies, transmitting ultrasonic signals to different positions of the wall to obtain multiple historical echo signals at each acquisition frequency; using the optimal acquisition frequency to transmit ultrasonic signals to different positions of the wall to obtain real-time echo signals, and the optimal acquisition frequency is the acquisition frequency with the highest stability; wherein the acquisition frequency is Stability for:

[0008] ;

[0009] In the formula, Indicates the frequency The sum of the differences between the historical echo signals of different defect categories at the time Indicates the frequency The sum of the differences between the historical echo signals of the same defect category at the same time; the difference represents the similarity between the historical echo signals; the real-time echo signal is input into the trained classification model to output the defect categories at different positions of the wall.

[0010] The above technical solution improves the accuracy and robustness of wall defect detection by combining multi-frequency acquisition and optimal frequency selection strategies. By acquiring historical echo signals at multiple acquisition frequencies and calculating the stability index of each frequency, the system can select the frequency with the highest stability for real-time echo signal acquisition, thereby ensuring the reliability of the signal and sensitivity to defects, significantly improving the efficiency and accuracy of wall damage detection, while reducing interference caused by signal noise and environmental changes, and ensuring the stable operation of the detection system under complex conditions.

[0011] Furthermore, the difference between the two historical echo signals is:

[0012] ;

[0013] In the formula, Indicates The historical echo signal and the The difference between the historical echo signals, Indicates The historical echo signal and the The covariance of the historical echo signals, Indicates The standard deviation of the historical echo signal, Indicates The standard deviation of the historical echo signal.

[0014] The above technical solution uses the ratio of covariance and standard deviation to quantify the similarity between echo signals, which can accurately evaluate the distribution of different echo signals in the feature space. Covariance reflects the linear relationship between echo signals, while standard deviation measures the fluctuation range of echo signals. Through this standardized difference calculation, it can eliminate the interference caused by different signal amplitudes and focus on changes in signal morphology. This difference measurement method improves the sensitivity of echo signals to different defect types, allowing the classification model to more accurately distinguish different types of defects during training, thereby improving the accuracy and robustness of defect detection.

[0015] Furthermore, the differences between the historical echo signals are specifically:

[0016] Perform Fourier transform on each historical echo signal, extract the corresponding period and amplitude features, generate a feature sequence corresponding to each historical echo signal based on the period and amplitude features, and calculate the difference between each pair of historical echo signals based on the feature sequence:

[0017] ;

[0018] In the formula, Indicates The historical echo signal and the The difference of the historical echo signals, Indicates The characteristic sequence corresponding to the historical echo signal, Indicates The characteristic sequence corresponding to the historical echo signal, Represents the dynamic time warping distance between two feature sequences.

[0019] The above technical solution can accurately capture the timing characteristics and change patterns of the signal. The Fourier transform converts the echo signal into the frequency domain, highlighting the periodicity and amplitude information, so that the key features of the signal can be fully extracted. The dynamic time warping algorithm can handle the timing deformation and nonlinear alignment problems between signals, avoiding the errors that may be caused by the misalignment of the time axis in traditional methods. Through this difference calculation method that combines time domain and frequency domain features, the similarity measure between echo signals can be significantly improved, so that the defect classification model can better handle complex and changing signals, thereby improving the accuracy, stability and robustness of defect detection.

[0020] Furthermore, the historical echo signal includes bandwidth, signal-to-noise ratio, radial velocity and Doppler shift.

[0021] Furthermore, it also includes data cleaning and data standardization processing of the historical echo signal.

[0022] The above technical solution significantly improves the accuracy and consistency of signal analysis by performing data cleaning and standardization on historical echo signals. Data cleaning removes outliers and unnecessary fluctuations caused by noise, sensor errors or environmental interference, thereby ensuring the quality of input data and avoiding the interference of these irrelevant factors on subsequent analysis results. Data standardization converts the characteristic values ​​of different signals to the same scale range, eliminating the effects of different acquisition conditions, signal strength or frequency differences, making subsequent analysis methods more stable and universal, and significantly improving the reliability and accuracy of analysis results.

[0023] Furthermore, the classification model is a long short-term memory network model.

[0024] Furthermore, the training classification model is specifically as follows: inputting the training set into a pre-built classification model for training, and during the training process, calculating the loss between the output prediction value and the label; adjusting the model parameters using the gradient descent method to minimize the prediction error; iteratively adjusting the parameters of the classification model until the loss is less than a certain value or reaches a set number of training times, and finally obtaining a trained classification model.

[0025] The above technical solution uses the gradient descent method to iteratively optimize the classification model, effectively improving the learning ability and classification accuracy of the model. It not only enhances the classification model's ability to recognize different defect types, but also improves the model's generalization ability, enabling it to maintain a high accuracy on new data that has not been seen.

[0026] In a second aspect, the present invention provides a building damage intelligent detection system, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, any one of the above-mentioned building damage intelligent detection methods is implemented.

[0027] The beneficial effects of the present invention are:

[0028] The present invention realizes efficient and accurate intelligent detection of building damage by combining technical means such as multi-frequency acquisition, signal processing, classification model training and data cleaning. It ensures the acquisition of high-quality historical echo signals by selecting the acquisition frequency with the highest stability, and effectively improves the ability to distinguish different defect categories by performing Fourier transform, feature extraction and difference calculation on the echo signals. Through the training of long short-term memory networks, the system can automatically learn complex signal patterns and defect characteristics, and can still maintain high classification accuracy and stability when facing complex environments, and realize intelligent recognition and classification of defects in different positions of the wall. The present invention not only improves the detection accuracy and robustness, but also has strong adaptability and scalability, and can provide efficient and automated building damage detection solutions in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0030] Figure 1 is a flow chart schematically illustrating a method for intelligently detecting building damage according to an embodiment of the present invention;

[0031] Figure 2 It is a structural block diagram schematically showing a building damage intelligent detection system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0033] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0034] An embodiment of a method for intelligently detecting building damage.

[0035] like Figure 1 As shown, a flow chart of a building damage intelligent detection method according to an embodiment of the present invention includes the following steps:

[0036] S1: Obtain historical echo signals and their corresponding defect categories, and train the classification model.

[0037] In one embodiment, obtaining the historical echo signal specifically includes: based on multiple acquisition frequencies, transmitting ultrasonic signals to different positions of the wall to obtain multiple historical echo signals at each acquisition frequency.

[0038] Specifically, by changing the acquisition frequency, the echo signals of the wall at different frequencies can be obtained. These signals contain multi-dimensional information about the internal structure of the wall. Echo signals of different frequencies have different response characteristics to different types of defects, thus providing more comprehensive defect detection data. By comprehensively analyzing the echo signals at each acquisition frequency, the accuracy and robustness of defect detection can be improved, especially for complex or variable wall structures, which can effectively reduce the errors and limitations caused by single frequency acquisition.

[0039] In one embodiment, the historical echo signal includes bandwidth, signal-to-noise ratio, radial velocity and Doppler shift, and the defect categories are divided into no defect, void and crack.

[0040] Among them, when all parameters are within the normal range, that is, the bandwidth is small, the signal-to-noise ratio is high, the radial velocity is close to zero, and the Doppler shift is small, it is considered that there is no defect; when the echo signal shows a moderate degree of disturbance, such as an increase in bandwidth, a moderate signal-to-noise ratio, and slight changes in radial velocity and Doppler shift, it is marked as a crack; if the echo signal shows obvious abnormalities, such as a large bandwidth, a low signal-to-noise ratio, and radial velocity and Doppler shift deviate from the normal range, it is marked as a hole;

[0041] Exemplarily: thresholds can be set for bandwidth, signal-to-noise ratio, radial velocity and Doppler shift. In this embodiment, the defect-free bandwidth is less than or equal to 5 Hz, the crack bandwidth is 5-10 Hz, and the void bandwidth is greater than 10 Hz; the echo signals are classified based on the set thresholds; in actual applications, the thresholds need to be continuously adjusted to improve the classification accuracy.

[0042] In one embodiment, the historical echo signal needs to be cleaned and standardized before classification to ensure the accuracy and reliability of the analysis results. Data cleaning first eliminates the influence caused by sensor errors, environmental factors or instability in the signal acquisition process by removing outliers and noise interference, thereby ensuring the quality of the echo signal. Then, through data standardization, the characteristic values ​​of the echo signals at different acquisition frequencies and different positions are converted to a unified scale range, eliminating the influence caused by differences in signal amplitude, frequency, etc., so that subsequent analysis and classification algorithms can be effectively compared under the same standard. Data cleaning and standardization not only improve the accuracy of defect classification, but also enhance the robustness of the system, enabling it to better adapt to different detection environments and signal fluctuations, and ultimately improve the recognition ability and classification accuracy of wall defects.

[0043] In one embodiment, the classification model is trained by:

[0044] The training of the classification model is to iteratively optimize the training set into the pre-built model to achieve high-precision classification. The training set is the historical echo signal and its corresponding defect category. Specifically, the input features of the training samples are matched with the corresponding labels during the training process. After the model generates the predicted value, the error between the predicted value and the true label is calculated through the loss function, such as mean square error or cross entropy. In order to minimize the error, the gradient descent algorithm is used to update the model parameters. The optimization process includes calculating the gradient of the loss function with respect to the model parameters and adjusting the parameter values ​​to gradually reduce the error. This process is carried out in an iterative manner until the loss function value drops to the set threshold or reaches the preset training rounds, ensuring the convergence and stability of the model. Through this optimization strategy, the classification model can learn the complex relationship between the input features and the target category, and finally achieve high-precision prediction of unknown data. The optimized model has good generalization ability and robustness, and can provide more reliable classification results in practical applications.

[0045] S2: Using the optimal acquisition frequency to transmit ultrasonic signals to different positions of the wall to obtain real-time echo signals, the optimal acquisition frequency is the acquisition frequency with the highest stability.

[0046] In one embodiment, the acquisition frequency is Stability for:

[0047] ;

[0048] In the formula, Indicates the frequency The sum of the differences between the historical echo signals of different defect categories at the time Indicates the frequency The sum of the differences between the historical echo signals of the same defect category at the same time;

[0049] The greater the difference between the echo signals of different defect categories, the higher the distinction between the defect categories. The smaller the signal difference within the same defect category, the more stable the signal. In this way, the most suitable acquisition frequency can be effectively evaluated and selected to improve the stability and reliability of the echo signal, thereby achieving more accurate defect identification and classification. This solution can not only improve the system's sensitivity to different defect types, but also reduce the selection error of the acquisition frequency, and improve the accuracy and robustness of the detection results.

[0050] The difference between the two historical echo signals is calculated as:

[0051] ;

[0052] In the formula, Indicates The historical echo signal and the The difference between the historical echo signals, Indicates The historical echo signal and the The covariance of the historical echo signals, Indicates The standard deviation of the historical echo signal, Indicates The standard deviation of the historical echo signal.

[0053] In another embodiment, the difference between two historical echo signals is calculated, specifically:

[0054] Perform Fourier transform on each historical echo signal, extract the corresponding period and amplitude features, generate a feature sequence corresponding to each historical echo signal based on the period and amplitude features, and calculate the difference between each pair of historical echo signals based on the feature sequence:

[0055] ;

[0056] In the formula, Indicates The historical echo signal and the The difference of the historical echo signals, Indicates The characteristic sequence corresponding to the historical echo signal, Indicates The characteristic sequence corresponding to the historical echo signal, Represents the dynamic time warping distance between two feature sequences.

[0057] By combining Fourier transform with dynamic time warping to calculate the difference between two historical echo signals, the periodicity and amplitude characteristics of the signal can be captured more accurately. Fourier transform converts each echo signal into frequency domain features and extracts the period and amplitude information of the signal, which can reflect the key patterns and changes in the signal. By generating feature sequences based on these period and amplitude features and then using the dynamic time warping algorithm to calculate the difference between two feature sequences, it is possible to effectively handle time shifts and nonlinear deformations in time series, compare the similarities or differences between signals, and thus avoid the limitations of traditional distance measurement methods.

[0058] S3: Input the real-time echo signal into a trained classification model to output defect categories at different positions of the wall.

[0059] In one embodiment, the classification model is a long short-term memory network model.

[0060] Exemplarily, after preprocessing and feature extraction, the real-time echo signal is input into the trained long short-term memory network model for defect identification. By taking the echo signal of each acquisition position as input, the model can accurately predict the defect type at different positions of the wall according to the mapping relationship between the features and categories learned during the training process. The fully trained long short-term memory network model can efficiently classify complex echo signals, identify different types of defects such as cracks and cavities, and has high robustness and accuracy. This process not only realizes automated defect detection, but also responds quickly in real-time detection, improving the efficiency and accuracy of detection. In addition, defect classification based on the training model can significantly reduce the error of manual judgment and improve the reliability and stability of the entire detection system in practical applications, thereby providing more accurate decision support for wall health detection.

[0061] The solution of the present invention proposes an efficient and accurate intelligent detection method and system for building damage by combining ultrasonic signal acquisition, feature extraction, classification model training and optimization. By acquiring historical echo signals based on multiple acquisition frequencies and selecting the optimal acquisition frequency with the highest stability, the quality and accuracy of the signal are effectively improved, and the influence of interference factors is reduced. The Fourier transform and dynamic time warping algorithms are used to accurately calculate the differences between historical echo signals, so that the similarities and differences between signals are analyzed more finely, thereby enhancing the model's ability to distinguish. It not only improves the accuracy and robustness of building damage detection, but also can classify and identify wall defects in different locations in real time and efficiently, with significant practical value and technical advantages.

[0062] An embodiment of a building damage intelligent detection system:

[0063] like Figure 2 As shown, a structural block diagram of an intelligent building damage detection system according to an embodiment of the present invention includes a processor and a memory.

[0064] The present invention also provides a building damage intelligent detection system. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent building damage detection method described above according to the present invention is implemented.

[0065] The intelligent building damage detection system also includes other components familiar to those skilled in the art, such as a communication interface, and the configuration and functions of the components are known in the art, so they will not be described in detail here.

[0066] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0067] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0068] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for intelligent detection of building damage, characterized in that: include: Obtain historical echo signals and their corresponding defect categories, and train classification models; The acquisition of historical echo signals specifically includes: based on multiple acquisition frequencies, transmitting ultrasonic signals to different positions of the wall to obtain multiple historical echo signals at each acquisition frequency; Ultrasonic signals are emitted to different positions of the wall using the optimal acquisition frequency to obtain real-time echo signals. The optimal acquisition frequency is the acquisition frequency with the highest stability. Stability for: ； In the formula, Indicates the frequency The sum of the differences between the historical echo signals of different defect categories at the time Indicates the frequency The sum of the differences between the historical echo signals of the same defect category at the time; the difference represents the similarity between the historical echo signals; The real-time echo signal is input into a trained classification model to output defect categories at different positions of the wall.

2. A building damage intelligent detection method according to claim 1, characterized in that: The difference between the two historical echo signals is: ； In the formula, Indicates The historical echo signal and The difference between the historical echo signals, Indicates The historical echo signal and The covariance of the historical echo signals, Indicates The standard deviation of the historical echo signal, Indicates The standard deviation of the historical echo signal.

3. A building damage intelligent detection method according to claim 1, characterized in that: The differences between the historical echo signals are specifically: Perform Fourier transform on each historical echo signal, extract the corresponding period and amplitude features, generate a feature sequence corresponding to each historical echo signal based on the period and amplitude features, and calculate the difference between each pair of historical echo signals based on the feature sequence: ； In the formula, Indicates The historical echo signal and the The difference of the historical echo signals, Indicates The characteristic sequence corresponding to the historical echo signal, Indicates The characteristic sequence corresponding to the historical echo signal, Represents the dynamic time warping distance between two feature sequences.

4. A building damage intelligent detection method according to claim 1, characterized in that: The historical echo signal includes bandwidth, signal-to-noise ratio, radial velocity and Doppler frequency shift.

5. The intelligent building damage detection method according to claim 1, characterized in that: The method also includes performing data cleaning and data standardization processing on the historical echo signal.

6. A building damage intelligent detection method according to claim 1, characterized in that: The classification model is a long short-term memory network model.

7. A building damage intelligent detection method according to claim 1, characterized in that: The training classification model is specifically: The training set is input into the pre-built classification model for training. During the training process, the loss between the output prediction value and the label is calculated; Use gradient descent to adjust model parameters to minimize prediction error; Iteratively adjust the parameters of the classification model until the loss is less than a certain value or the set number of training times is reached, and finally a trained classification model is obtained.

8. An intelligent building damage detection system, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent building damage detection method as claimed in any one of claims 1 to 7 is implemented.

Images