Building damage detection method based on sequential InSAR coherence images based on generative model

Abstract

The present invention relates to the technical field of remote sensing image anomaly detection, and specifically to a method for detecting building damage using sequential InSAR coherence images based on a generative model, comprising: generating a time series of InSAR coherence images of a building area using a time series of SAR images of the building area; training a WGAN generative model using InSAR coherence images of the building area from time 2 to time T-1 to obtain a trained WGAN generative model; subtracting an InSAR coherence image of the building area at time T+1 before a disaster, generated by the trained WGAN generative model, from an InSAR coherence image of the building area at time T+1 after a disaster to obtain an anomaly image; and obtaining anomaly information based on the anomaly image, and using a set threshold for comparison to screen building damage areas. The present invention provides a more accurate detection method for post-disaster building damage, thereby improving the performance of building damage detection.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image anomaly detection, and in particular to a method for detecting building damage using sequential InSAR coherence images based on a generative model. Background Art

[0002] In recent years, the rapid development of satellite remote sensing technology has enabled the acquisition of high-resolution, wide-bandwidth remote sensing images, enabling detailed and extensive observations of land surfaces. These images are a crucial tool for assessing large-scale building damage following natural disasters. By comparing changes in satellite imagery before and after a disaster, the extent of building damage can be accurately assessed, quickly understanding the scope and severity of the impact, and ultimately formulating effective rescue and reconstruction plans. Because synthetic aperture radar (SAR) can penetrate clouds and fog, it can capture clear images even in adverse weather conditions, providing a crucial basis for assessing building damage caused by disasters such as earthquakes, floods, and landslides. Its all-weather, all-day imaging capabilities further enhance the value of remote sensing technology in disaster emergency response.

[0003] Traditional SAR image-based building damage detection methods typically utilize the intensity values ​​of SAR images. This involves comparing the changes in scattering intensity of building areas in pre- and post-disaster images to detect corresponding damaged areas. However, this backscatter intensity information reflects the electromagnetic properties and roughness of surface objects and is insensitive to subtle changes such as building collapses and surface cracks. This results in low accuracy in reflecting surface disturbances and, in turn, in detecting building damage areas. To address this issue, existing techniques employ interferometric processing of two SAR images taken before and after a disaster. The resulting coherent image can reflect the phase consistency of the land surface before and after the disaster. This image is highly sensitive to subtle changes in the land surface, enabling the detection of subtle deformation and structural changes. Furthermore, compared to backscatter intensity, coherence changes can more accurately reflect surface disturbances, particularly for subtle changes such as building collapses and surface cracks, offering higher detection capabilities. Furthermore, sequential InSAR coherent images can further improve the accuracy of building damage detection.

[0004] Deep learning has become a mainstream technology for time series data analysis. Recurrent neural networks (RNNs) have been used as probabilistic anomaly detectors for detecting anomalous changes in sequential interferometric synthetic aperture radar (InSAR) coherent images. However, to simplify the mathematical model, existing methods assume that the coherent data follows a Gaussian distribution. This assumption, however, cannot be applied to actual coherent data, thus affecting the effectiveness of building damage detection. Summary of the Invention

[0005] To this end, the technical problem to be solved by the present invention is to overcome the problem in the prior art that the processing of sequential InSAR coherent images is not based on the actual distribution of coherent data, resulting in the final abnormal image being of low accuracy and having no reference value, which affects the results of building damage detection.

[0006] To solve the above technical problems, the present invention provides a building damage detection method based on sequential InSAR coherence images using a generative model, comprising:

[0007] Acquire SAR images of the building area from time 1 to time T+1 to form a SAR image time series of the building area; wherein, the SAR images of the building area from time 1 to time T are SAR images of the building area before the disaster occurs; the SAR image of the building area at time T+1 is the SAR image of the building area after the disaster occurs;

[0008] Perform registration and coherence calculation on the SAR image time series of the building area to obtain the InSAR coherent images of the building area from time 2 to time T+1, forming the InSAR coherent image time series of the building area;

[0009] Inputting the InSAR coherent images of the building area from time 2 to time T-1 into the WGAN generative model to generate the InSAR coherent images of the building area from time 3 to time T; wherein the WGAN generative model includes a generator and a discriminator; using the InSAR coherent images of the building area from time 3 to time T generated by the WGAN generative model as generated image samples, and using the InSAR coherent images of the building area from time 3 to time T as real image samples; constructing a loss function for the generator with the goal of minimizing the Wasserstein distance and the absolute error loss between the real image samples and the generated image samples; constructing a loss function for the discriminator with the goal of minimizing the Wasserstein distance between the generated image samples and the real image samples, and training the WGAN generative model;

[0010] The InSAR coherent images of the building area from time 2 to time T are input into the trained WGRN model to generate InSAR coherent images of the building area from time 3 to time T+1 when no disaster occurs. The generated InSAR coherent image of the building area at time T+1 when no disaster occurs is subtracted from the InSAR coherent image of the building area at time T+1 after the disaster occurs to obtain an anomaly image. The building area map is converted into a binary mask image of the building area and multiplied with the anomaly image to extract multiple anomaly information values ​​of the building area. Each anomaly information value of the building area is compared with the set threshold. If the current anomaly information is greater than the threshold, the area corresponding to the current anomaly information is determined to be a building damaged area.

[0011] Preferably, performing a registration operation on the SAR image time series of the building area to obtain the registered SAR image time series of the building area includes:

[0012] The SAR image at the first moment in the SAR image time series of the building area is used as the main image, and the remaining images in the SAR image time series of the building area are used as the SAR image time series to be registered;

[0013] The registration operation includes: a coarse registration stage and a fine registration stage;

[0014] The coarse registration stage includes:

[0015] Using the center point of the master image as the initial control point, an iterative algorithm is used to find the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point. Based on the coordinates of the initial control point and the coordinates of the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point, the coordinate offset between the SAR image to be registered at the current moment and the master image is calculated.

[0016] The fine registration stage includes:

[0017] Select multiple control points in the main image, and determine the same-name points corresponding to the current SAR image to be registered and each control point based on the coordinate offset between the current SAR image to be registered and the main image obtained in the coarse registration stage;

[0018] The coordinates of the points with the same names corresponding to the control points in the SAR image to be registered at the current moment are screened, and the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment is fitted; based on the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment, the coordinate transformation and resampling of the SAR image to be registered at the current moment are performed to obtain the registered SAR image at the current moment;

[0019] The registration operation is performed on the SAR images to be registered at each moment in turn to obtain the registered SAR images at each moment, and a time series of SAR images of the registered building area is constructed.

[0020] Preferably, according to the coherence coefficient calculation formula, the coherence calculation is performed on the time series of the registered SAR images of the building area to obtain the time series of the InSAR coherent images of the building area, which includes:

[0021] The coherence coefficient calculation formula is expressed as:

[0022] ;

[0023] in, 、 They are the first Time, Coordinates in the SAR image at time The complex value at ; Represents the first SAR image at the moment and The pixels of the SAR image at the moment are conjugate multiplied to obtain the Interference images of moments; 、 Respectively represent the number of rows and columns of the data block during coherence calculation; ;

[0024] The InSAR coherent images of the building area from time 2 to time T+1 are obtained in sequence to form a time series of InSAR coherent images of the building area.

[0025] Preferably, the loss function of the construction generator is expressed as:

[0026] ;

[0027] in, represents the loss of the generator; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; Denotes the discriminator's The score of the generated image sample at the moment; The weight hyperparameter representing the loss; Indicates the The absolute error loss between the real image sample and the generated image sample at each moment; represents a real image sample.

[0028] Preferably, the loss function for constructing the discriminator is expressed as:

[0029] ;

[0030] in, represents the loss of the discriminator; represents a real image sample; Represents the distribution of real image samples; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; represents the score of the discriminator for the real image sample; Denotes the discriminator's The score of the generated image sample at time t.

[0031] Preferably, the generator is composed of a GRU model with three hidden layers of size 128 and a fully connected layer.

[0032] Preferably, the update expression of the GRU model is:

[0033] ;

[0034] ;

[0035] ;

[0036] ;

[0037] in, represents the update gate; Represents the Sigmoid activation function, which maps the model input to the interval [0,1]; Indicates the The input sample corresponding to the moment, that is, A true image of the moment; 、 Respectively represent the weight matrices of the update gate input state and hidden state; represents the update gate bias term; Reset gate. 、 Represent the weight matrices of the reset gate input state and hidden state respectively; Represents the reset gate bias term; Indicates the Candidate hidden states at time t; Indicates the The hidden state of the moment; Indicates the The hidden state corresponding to the moment; 、 Represent the input state of the candidate hidden state and the weight matrix of the hidden state respectively; A bias term representing the candidate hidden state; Represents element-wise multiplication;

[0038] No. The update expression of the hidden state at the moment is: ;in, Represents the GRU model.

[0039] Preferably, the expression of the fully connected layer is:

[0040] ;

[0041] in, Indicates the The output sample at the moment is the output of the fully connected layer, that is, the Generate images at the moment; represents a fully connected layer; Indicates the The hidden state at the moment is the input of the fully connected layer.

[0042] Preferably, the discriminator is composed of a one-dimensional CNN model with a convolution kernel of 1×1 and 64 output channels and a fully connected layer.

[0043] Preferably, during the training process of the WGAN generative model, in each training cycle, the discriminator is updated five times in a cycle and then the generator is updated once.

[0044] The above technical solution of the present invention has the following beneficial effects compared with the prior art:

[0045] The present invention discloses a method for detecting building damage using sequential InSAR coherence images based on a generative model. The method performs registration, interference operation, and coherence calculation on sequential SAR images of the land surface of the building area within a certain time period to obtain sequential InSAR coherence images of the land surface within a certain time period, namely, an InSAR coherence image time series. During the model training process, in order to enable the generator to better learn the temporal changes of the InSAR coherence images of the building area before the disaster occurs and to ensure that the images generated by the generator have a higher similarity with the real images, the loss function of the generator is constructed with the goal of minimizing the Wasserstein distance between the real image samples and the generated image samples and minimizing the absolute error loss. In order to allow the discriminator to focus only on the difference between the generated samples and the real samples during the training process, the loss function of the discriminator is constructed with the goal of minimizing the Wasserstein distance between the generated image samples and the real image samples. At the same time, due to the time before the disaster occurs, Since all the sequential InSAR coherent images are free of anomalies, they can be used as training data and the predicted coherent images generated by the generator in the trained WGAN generative model will not be affected by anomalies. In addition, the Wasserstein distance is taken into account in the loss of the WGAB generative model, which enables the WGAB generative model to better capture complex image distributions, thereby generating more realistic and higher-resolution coherent images. Therefore, the coherent images predicted and generated by the WGAB generative model before a disaster are highly accurate, providing more accurate data support for the subsequent assessment of post-disaster building damage. The first predicted coherent image after the disaster is subtracted from the first true coherent image to obtain an anomaly image. The set threshold is used to filter out the post-disaster building damage area, which can more accurately distinguish between changes caused by disasters and changes caused by natural development, further improving the detection accuracy of coherence anomaly changes and the accuracy of the detection results of building damage areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0047] Figure 1 This is a flow chart of a method for detecting building damage using sequential InSAR coherence images based on a generative model provided by the present invention;

[0048] Figure 2 This is an abnormal image of the Italian town of Norcia after the disaster, obtained by using a sequential InSAR coherence image building damage detection method based on a generative model;

[0049] Figure 3This is a diagram showing the statistical distribution of true coherence values ​​and the coherence value statistical distribution generated by WGAN after the disaster in the small town of Norcia, Italy;

[0050] Figure 4 This is a comparison of the precision-recall curves of damage detection in the small town of Norcia, Italy, obtained based on different methods. DETAILED DESCRIPTION

[0051] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0052] Reference Figure 1 As shown, Figure 1 This is a flow chart of a method for building damage detection based on sequential InSAR coherence images using a generative model provided by the present invention; specifically, it includes:

[0053] S1: Acquire SAR images of the building area from time 1 to time T+1 to form a time series of SAR images of the building area; the SAR images of the building area from time 1 to time T are all SAR images of the building area before the disaster; the SAR image of the building area at time T+1 is the SAR image of the building area after the disaster;

[0054] S2: Perform registration and coherence calculation on the SAR image time series of the building area to obtain the InSAR coherent images of the building area from time 2 to time T+1, forming the InSAR coherent image time series of the building area. Among them, the InSAR coherent images of the building area from time 2 to time T are all InSAR coherent images of the building area before the disaster occurs; the InSAR coherent image of the building area at time T+1 is the InSAR coherent image of the building area at time T+1 after the disaster occurs, that is, the first true coherent image of the building area after the disaster occurs, including:

[0055] S21: performing a registration operation on the SAR image time series of the building area to obtain a registered SAR image time series of the building area, including:

[0056] The SAR image at the first moment in the SAR image time series of the building area is used as the main image, and the remaining images in the SAR image time series of the building area are used as the SAR image time series to be registered;

[0057] The registration operation includes: a coarse registration stage and a fine registration stage;

[0058] The coarse registration stage includes:

[0059] Using the center point of the master image as the initial control point, an iterative algorithm is used to find the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point. Based on the coordinates of the initial control point and the coordinates of the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point, the coordinate offset between the SAR image to be registered at the current moment and the master image is calculated.

[0060] The fine registration stage includes:

[0061] Select multiple control points in the main image, and determine the same-name points corresponding to the current SAR image to be registered and each control point based on the coordinate offset between the current SAR image to be registered and the main image obtained in the coarse registration stage;

[0062] The coordinates of the points with the same names corresponding to the control points in the SAR image to be registered at the current moment are screened, and the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment is fitted; based on the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment, the coordinate transformation and resampling of the SAR image to be registered at the current moment are performed to obtain the registered SAR image at the current moment;

[0063] Perform registration operations on the SAR images to be registered at each moment in turn to obtain the registered SAR images at each moment, and form a time series of SAR images of the registered building area;

[0064] S22: According to the coherence coefficient calculation formula, the coherence of the SAR image time series of the building area after registration is calculated to obtain the InSAR coherent image time series of the building area, including:

[0065] The coherence coefficient calculation formula is expressed as:

[0066] ;

[0067] in, 、 They are the first Time, Coordinates in the SAR image at time The complex value at ; Represents the first SAR image at the moment and The pixels of the SAR image at the moment are conjugate multiplied to obtain the Interference images of moments; 、 Respectively represent the number of rows and columns of the data block during coherence calculation; ;

[0068] The InSAR coherent images of the building area from time 2 to time T+1 are obtained in sequence to form a time series of InSAR coherent images of the building area.

[0069] In summary, assuming there are T+1 SAR images, we will get T InSAR coherent images. Interferometric coherence is used to describe the similarity between the regions with the same name in the two interferometric images, and is distributed in the interval [0,1]. A coherence of 0 indicates that the two images are completely unrelated, and a coherence of 1 indicates that they are completely consistent.

[0070] S3: Input the InSAR coherent images of the building area from time 2 to time T-1 into the WGAN generative model to generate the InSAR coherent images of the building area from time 3 to time T; wherein the WGAN generative model includes a generator and a discriminator; the InSAR coherent images of the building area from time 3 to time T generated by the WGAN generative model are used as generated image samples, and the InSAR coherent images of the building area from time 3 to time T are used as real image samples; with the goal of minimizing the Wasserstein distance and the absolute error loss between the real image samples and the generated image samples, a loss function of the generator is constructed; with the goal of minimizing the Wasserstein distance between the generated image samples and the real image samples, a loss function of the discriminator is constructed to train the WGAN generative model;

[0071] Among them, the loss function of the generator is expressed as:

[0072] ;

[0073] in, represents the loss of the generator; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; Denotes the discriminator's The score of the generated image sample at the moment; The weight hyperparameter representing the loss; Represents the absolute error loss between real image samples and generated image samples; represents a real image sample; in a specific embodiment of the present invention, ;

[0074] Among them, the expression of the loss function of the discriminator is:

[0075] ;

[0076] in, represents the loss of the discriminator; represents a real image sample; Represents the distribution of real image samples; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; represents the score of the discriminator for the real image sample; Denotes the discriminator's The score of the generated image sample at the moment;

[0077] Among them, the update expression of the GRU model is:

[0078] ;

[0079] ;

[0080] ;

[0081] ;

[0082] in, represents the update gate, which is used to control how much of the state information at the previous moment should be retained in the current state, and how much new information in the current input should be added to the current state. In this way, the network can flexibly adjust its memory capacity to effectively capture long-term dependencies; Represents the Sigmoid activation function, which maps the model input to the interval [0,1]; Indicates the The input sample corresponding to the moment, that is, A true image of the moment; 、 Respectively represent the weight matrices of the update gate input state and hidden state; represents the update gate bias term; Represents the reset gate, which is used to control the extent to which the hidden state of the previous moment can be ignored. If the reset gate believes that past experience is not relevant to the current task, it will make the network rely more on the current input to make decisions; 、 Represent the weight matrices of the reset gate input state and hidden state respectively; Represents the reset gate bias term; Indicates the Candidate hidden states at time t; Indicates the The hidden state of the moment; Indicates the The hidden state corresponding to the moment; 、 Represent the input state of the candidate hidden state and the weight matrix of the hidden state respectively; A bias term representing the candidate hidden state; Represents element-by-element multiplication. By using the update gate and reset gate, GRU can effectively ignore data with large seasonal differences when predicting InSAR coherence, and focus on sequence patterns with historically similar conditions, thereby improving prediction accuracy.

[0083] No. The update expression of the hidden state at the moment is: ;in, Represents the GRU model;

[0084] Among them, the expression of the fully connected layer of the generator is:

[0085] ;

[0086] in, Indicates the The output sample at the moment is the output of the fully connected layer, that is, the Generate image samples at time instant; represents a fully connected layer; Indicates the The hidden state at the moment is the input of the fully connected layer;

[0087] In summary, assuming that the image before the disaster is , the training is mainly used To predict coherence time series;

[0088] The WGAN network consists of a generator and a discriminator. It mainly uses coherent images without abnormalities to train the generator, learn the time variation of coherence without abnormalities, and the generated coherence map is also without abnormalities. The discriminator is used to represent the Wasserstein distance. When training WGAN, the input is normalized to [-1,1] and the discriminator is used to train the generator. Indicates the The InSAR coherent image at the moment The real image of the moment, Indicates the generated The InSAR coherent image at the moment The generated image at the moment; the calculation formula of Wasserstein distance is:

[0089] ;

[0090] in, Represents the true sample distribution The sample distribution generated by the generator Wasserstein distance between them; express and The set of all possible joint distributions; express and A possible joint distribution of

[0091] The advantage of Wasserstein distance is that it can reflect the distance between two distributions even if they do not overlap. The Wasserstein distance is converted into the following formula:

[0092] ;

[0093] Define the transformed loss as the generator; where express The Lipschitz constant does not exceed K; during the training process, all parameters of the neural network are set Not exceeding a certain range [-c, c], construct a The discriminator network ; Based on the discriminator network , in the parameter Without exceeding the range of [-c,c], To get the maximum possible, The Wasserstein distance between the approximate real image sample distribution and the generated image sample distribution;

[0094] Since the generator needs to minimize the Wasserstein distance between the real image sample distribution and the generated image sample distribution, that is, minimize , and because The first term of has nothing to do with the generator, so the generator is to minimize the following formula, namely: ;

[0095] Therefore, the generator loss and discriminator loss of the WGAN model are:

[0096] ;

[0097] ;

[0098] in, Represents the original generator loss of the WGAN model;

[0099] The generator and discriminator of WGAN are trained using the time series coherent images before the disaster to predict the time series coherent images. Since the coherent images before the disaster are not affected by the anomaly, the generator only generates coherent images that are not affected by the anomaly.

[0100] In the present invention, the loss of the generator is: ;

[0101] in, It is the L1 loss (absolute error loss) of the generated data and the actual data. Adding the L1 loss to the generator loss can improve the quality and stability of the generated images, as well as increase the similarity in content between the generated images and the real images;

[0102] In a specific embodiment of the present invention, the generator is composed of a GRU model with three hidden layers of 128 and a fully connected layer; the discriminator is composed of a one-dimensional CNN model with a convolution kernel of 1×1 and 64 output channels and a fully connected layer; the GRU network can learn long-term dependencies in time series, and the discriminator uses a one-dimensional CNN network with a convolution kernel of 1×1 and a fully connected layer for feature extraction and fusion; the one-dimensional CNN network can optimize input features through channel transformation without changing the time dimension of the feature map, thereby learning richer feature representations; a fully connected layer is used for feature fusion, integrating the diverse features extracted by the CNN network, and the generated data and real data are added to the discriminator to achieve the representation of Wasserstein distance;

[0103] During the training process, the present invention updates the discriminator 5 times in a cycle and then updates the generator once, so that the negative discriminator loss can better represent the Wasserstein distance between the generated sequence and the actual sequence; after multiple iterations, the generator can better generate data until the discriminator loss is close to stable, and then the training is stopped;

[0104] Since the coherent images before the disaster were normal, the trained WGAN network was used to generate the coherent images after the disaster. The resulting predicted coherent images included the effects of temporal decorrelation but were not affected by anomalies. Since the discriminator loss is the negative Wasserstein distance, its absolute value can be used to indicate the training results.

[0105] S4: Input the InSAR coherent images of the building area from time 2 to time T into the trained WGRN model to generate the InSAR coherent images of the building area from time 3 to time T+1 when no disaster occurs; subtract the generated InSAR coherent image of the building area at time T+1 when no disaster occurs from the InSAR coherent image of the building area at time T+1 after the disaster occurs to obtain an anomaly image; convert the building area map into a binary mask image of the building area, multiply it with the anomaly image, and extract multiple anomaly information values ​​of the building area; compare each anomaly information value of the building area with the set threshold. If the current anomaly information is greater than the threshold, the area corresponding to the current anomaly information is determined to be a building damaged area.

[0106] To validate the effectiveness of the proposed building damage detection method, this experiment used data from the central Italy earthquake of August 24, 2016. This event destroyed much of the small town of Norcia, Italy. The Copernicus Emergency Management Service compiled a damage map assessing the extent of damage to each building. This comprehensive damage assessment quantitatively validated the effectiveness of the WGAN method for anomaly detection based on the existing loss levels. The experiment generated 35 coherent images processed using Sentinel-1 VV-polarized SLC images, including 34 pre-disaster coherent images and one post-disaster coherent image.

[0107] Figure 2 The results of building damage detection using the WGAN method are shown. The present invention uses real damage data from the Copernicus Management Service to select the optimal threshold for damage.

[0108] The WGAN method mainly makes predictions by learning the distribution of data, without making any assumptions about the data distribution of the image. Figure 3 As can be seen in Figure 2, the distribution of coherent images does not strictly obey the Gaussian distribution, but the WGAN method can learn the distribution of coherent images more accurately.

[0109] To calculate the optimal loss threshold for each method, the weighted harmonic mean of precision and recall can be used , the formula is:

[0110] ;

[0111] Will change with the change of threshold, set in the experiment , thus calculating all possible thresholds of the three methods score; selected The weighted precision is twice as high as the recall rate because we hope to direct limited emergency resources to the places most likely to be damaged, thereby supporting higher precision at the expense of lower recall rate; When the threshold is 0, the precision of the three methods is higher than 0.8, which means that more than 80% of the damaged points are indeed damaged. However, the recall of the WGAN method is 0.6332, while the recalls of the other two methods are 0.6080 and 0.5611, which is a significant quantitative improvement. The optimal threshold and the corresponding precision and recall of each method are shown in Table 1.

[0112] Table 1 Experimental results obtained using different detection methods:

[0113] method Optimal threshold Optimal accuracy Optimal recall Accuracy AUC CCD 0.3079 0.8505 0.5611 0.8133 RNN 2.1259 0.88018 0.6080 0.8588 WGAN 0.8473 0.8879 0.6332 0.8622

[0114] The area under the precision-recall curve can be used to quantitatively compare the effectiveness of different detection methods; the closer the area under the precision-recall curve is to 1, the better the effect; Figure 4 The following are the precision-recall curves of different methods. The area under the precision-recall curve of the WGAN method is 0.8622, while the areas under the precision-recall curves of the RNN and CDD methods are 0.8588 and 0.8133 respectively, which show some improvement.

[0115] In summary, the sequential InSAR coherent image building damage detection system based on deep generative model proposed in the present invention can achieve better post-disaster building damage detection effect.

[0116] This invention relates to a method for building damage detection using sequential InSAR coherent images based on a deep generative model. The method generates a coherent image time series using SAR image time series; calculates the loss functions of the generator and discriminator of a WGAN network; trains the WGAN using the coherent image time series before a disaster; uses the trained WGAN network to predict coherent images after a disaster, and compares the predicted coherent images with the actual coherent images to obtain anomaly images. This method can provide a more accurate detection method for post-disaster building damage under all-day and all-weather conditions, improving the performance of building damage detection.

[0117] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for building damage detection based on sequential InSAR coherence images using a generative model, characterized in that: include: Acquire SAR images of the building area from time 1 to time T+1 to form a SAR image time series of the building area; wherein, the SAR images of the building area from time 1 to time T are SAR images of the building area before the disaster occurs; the SAR image of the building area at time T+1 is the SAR image of the building area after the disaster occurs; Perform registration and coherence calculation on the SAR image time series of the building area to obtain the InSAR coherent images of the building area from time 2 to time T+1, forming the InSAR coherent image time series of the building area; Inputting the InSAR coherent images of the building area from time 2 to time T-1 into the WGAN generative model to generate the InSAR coherent images of the building area from time 3 to time T; wherein the WGAN generative model includes a generator and a discriminator; using the InSAR coherent images of the building area from time 3 to time T generated by the WGAN generative model as generated image samples, and using the InSAR coherent images of the building area from time 3 to time T as real image samples; constructing a loss function for the generator with the goal of minimizing the Wasserstein distance and the absolute error loss between the real image samples and the generated image samples; constructing a loss function for the discriminator with the goal of minimizing the Wasserstein distance between the generated image samples and the real image samples, and training the WGAN generative model; The InSAR coherent images of the building area from time 2 to time T are input into the trained WGRN model to generate the InSAR coherent images of the building area from time 3 to time T+1 when no disaster occurs. The generated InSAR coherent image of the building area at time T+1 when no disaster occurs is subtracted from the InSAR coherent image of the building area at time T+1 after the disaster occurs to obtain an anomaly image. The building area map is converted into a binary mask image of the building area and multiplied with the anomaly image to extract multiple anomaly information values ​​of the building area. Each anomaly information value of the building area is compared with the set threshold. If the current anomaly information is greater than the threshold, the area corresponding to the current anomaly information is determined to be a building damaged area.

2. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: The SAR image time series of the building area is registered, and the SAR image time series of the building area after registration includes: The SAR image at the first moment in the SAR image time series of the building area is used as the main image, and the remaining images in the SAR image time series of the building area are used as the SAR image time series to be registered; The registration operation includes: a coarse registration stage and a fine registration stage; The coarse registration stage includes: Using the center point of the master image as the initial control point, an iterative algorithm is used to find the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point. Based on the coordinates of the initial control point and the coordinates of the same-name point in the SAR image to be registered at the current moment that corresponds to the initial control point, the coordinate offset between the SAR image to be registered at the current moment and the master image is calculated. The fine registration stage includes: Select multiple control points in the main image, and determine the same-name points corresponding to the current SAR image to be registered and each control point based on the coordinate offset between the current SAR image to be registered and the main image obtained in the coarse registration stage; The coordinates of the points with the same names corresponding to the control points in the SAR image to be registered at the current moment are screened, and the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment is fitted; based on the optimal coordinate transformation model corresponding to the SAR image to be registered at the current moment, the coordinate transformation and resampling of the SAR image to be registered at the current moment are performed to obtain the registered SAR image at the current moment; The registration operation is performed on the SAR images to be registered at each moment in turn to obtain the registered SAR images at each moment, and a time series of SAR images of the registered building area is constructed.

3. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 2, characterized in that: According to the coherence coefficient calculation formula, the coherence of the SAR image time series of the building area after registration is calculated, and the InSAR coherent image time series of the building area is obtained, including: The coherence coefficient calculation formula is expressed as: ； in, 、 They are the first Time, Coordinates in the SAR image at time The complex value at ; Represents the first SAR image at the moment and The pixels of the SAR image at the moment are conjugate multiplied to obtain the Interference images of moments; 、 Respectively represent the number of rows and columns of the data block during coherence calculation; ; The InSAR coherent images of the building area from time 2 to time T+1 are obtained in sequence to form a time series of InSAR coherent images of the building area.

4. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: The loss function of the generator is expressed as: ； in, represents the loss of the generator; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; Represents the discriminator for the The score of the generated image sample at the moment; The weight hyperparameter representing the loss; Indicates the The absolute error loss between the real image sample and the generated image sample at each moment; represents a real image sample.

5. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: The loss function of the discriminator is constructed as follows: ； in, represents the loss of the discriminator; represents a real image sample; Represents the distribution of real image samples; Indicates the Generate image samples at time instant; Represents the sample distribution generated by the generator; represents the mean; represents the discriminator network; represents the score of the discriminator for the real image sample; Represents the discriminator for the The score of the generated image sample at time t.

6. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: The generator consists of a GRU model with three hidden layers of size 128 and a fully connected layer.

7. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 6, characterized in that: The update expression of the GRU model is: ； ； ； ； in, represents the update gate; Represents the Sigmoid activation function, which maps the model input to the interval [0,1]; Indicates the The input sample corresponding to the moment, that is, A true image of the moment; 、 Represent the weight matrices of the update gate input state and hidden state respectively; represents the update gate bias term; Reset gate. 、 Represent the weight matrices of the reset gate input state and hidden state respectively; Represents the reset gate bias term; Indicates the Candidate hidden states at time t; Indicates the The hidden state of the moment; Indicates the The hidden state corresponding to the moment; 、 Represent the input state of the candidate hidden state and the weight matrix of the hidden state respectively; A bias term representing the candidate hidden state; Represents element-wise multiplication; No. The update expression of the hidden state at the moment is: ;in, Represents the GRU model.

8. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 6, characterized in that: The expression of the fully connected layer is: ； in, Indicates the The output sample at the moment is the output of the fully connected layer, that is, the Generate images at the moment; represents a fully connected layer; Indicates the The hidden state at the moment is the input of the fully connected layer.

9. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: The discriminator consists of a one-dimensional CNN model with a convolution kernel of 1×1 and 64 output channels and a fully connected layer.

10. The method for building damage detection based on sequential InSAR coherence images of a generative model according to claim 1, characterized in that: During the training process of the WGAN generative model, in each training cycle, the discriminator is updated five times in a cycle and then the generator is updated once.

Images