A disaster prediction method, device, equipment, and storage medium based on SAR satellites

Abstract

This application discloses a disaster prediction method, apparatus, device, and storage medium based on SAR satellites. The method includes: acquiring multi-temporal images collected by SAR satellites; calculating the temporal coherence coefficient at each pixel location for every two temporal images, and calculating the normalized scattering intensity at each pixel location for each temporal image; performing wavelet threshold denoising on each temporal coherence coefficient, and fusing the processed result with the normalized scattering intensity at the corresponding pixel location to obtain a feature vector; and obtaining the disaster prediction result corresponding to each pixel location based on the feature vector and a preset clustering method. Applying the solution provided in this application allows for joint feature detection based on both temporal coherence and scattering intensity for multi-temporal images. Compared to methods relying on single-temporal echo intensity, this approach can extract more sensitive data, enabling the detection of minute changes in surface deformation in the early stages of a disaster, thereby improving the accuracy of disaster detection.

Description

Technical Field

[0001] This application relates to the field of disaster prediction technology, and more specifically, to a disaster prediction method, apparatus, equipment, and storage medium based on SAR satellites. Background Technology

[0002] SAR (Synthetic Aperture Radar) is an active Earth observation system that can be installed on aircraft, satellites, and other flying platforms to conduct all-weather, 24 / 7 Earth observations and has a certain degree of ground penetration capability. Therefore, SAR systems have unique advantages in applications such as disaster monitoring, environmental monitoring, marine monitoring, resource exploration, crop yield estimation, and mapping, and can play a role that is difficult to achieve with other remote sensing methods.

[0003] When using SAR satellites for disaster prediction, the most common known method relies on single-phase echo intensity to monitor surface deformation. However, this method suffers from insufficient noise resistance in detecting subtle changes in the early stages of a disaster, resulting in poor accuracy. Therefore, improving the noise resistance of disaster detection, and consequently, its accuracy, has become a pressing technical challenge. Summary of the Invention

[0004] This application provides a disaster prediction method, apparatus, equipment, and storage medium based on SAR satellites to improve the noise resistance of disaster detection, thereby improving the accuracy of disaster detection. The specific technical solution is as follows.

[0005] In a first aspect, embodiments of this application provide a disaster prediction method based on synthetic aperture radar (SAR) satellites, the method comprising:

[0006] Acquire multi-temporal images from SAR satellites;

[0007] Calculate the temporal coherence coefficient of each pair of said temporal images at each pixel location, and calculate the normalized scattering intensity of each said temporal image at each pixel location;

[0008] Wavelet thresholding denoising is performed on each of the temporal coherence coefficients, and the processed result is fused with the normalized scattering intensity at the corresponding pixel position to obtain a feature vector;

[0009] Based on the feature vector, disaster prediction results corresponding to each pixel location are obtained using a preset clustering method.

[0010] In one embodiment of this application, the step of calculating the temporal coherence coefficient at each pixel position for every two temporal images includes:

[0011] Calculate the temporal coherence coefficient γ of temporal images i and j within the local region of pixel location (x, y) using the following formula. ij (x,y):

[0012]

[0013] Where N is the total number of pixel windows included in the local region; S i (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image i; |S i (x,y;n)| 2 For S i The signal strength of (x, y; n) S; j (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image j; |S j (x,y;n)| 2 For S j Signal strength of (x, y; n); For S j The conjugate signal of (x, y; n).

[0014] In one embodiment of this application, the step of calculating the normalized scattering intensity at each pixel location of each temporal image includes:

[0015] Calculate the normalized scattering intensity β(x,y) at pixel location (x,y) for any temporal image using the following formula:

[0016]

[0017] S(x,y;n) is the complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any temporal image; |S(x,y;n)| 2 S(x,y;n) represents the signal strength; N represents the total number of pixel windows included in the local region; and P0 represents the reference power factor.

[0018] In one embodiment of this application, the step of obtaining the disaster prediction result corresponding to each pixel position based on the feature vector and a preset clustering method includes:

[0019] For each pixel location's feature vector, if the temporal coherence coefficient included in the feature vector is less than or equal to a preset first threshold, and the normalized scattering intensity is greater than or equal to a second threshold, then the pixel location is determined to have a disaster risk.

[0020] In one embodiment of this application, after acquiring multi-temporal images collected by SAR satellites, the method further includes:

[0021] The time-phase images are subjected to time-series correction and registration, and adaptive filtering is performed using a preset confidence factor.

[0022] Secondly, embodiments of this application provide a disaster prediction device based on synthetic aperture radar (SAR) satellites, the device comprising:

[0023] The image acquisition module is used to acquire multi-temporal images collected by SAR satellites;

[0024] The parameter calculation module is used to calculate the temporal coherence coefficient of each pair of the time-phase images at each pixel position, and to calculate the normalized scattering intensity of each time-phase image at each pixel position.

[0025] The vector construction module is used to perform wavelet threshold denoising on each of the temporal coherence coefficients, and fuse the processed result with the normalized scattering intensity at the corresponding pixel position to obtain a feature vector;

[0026] The disaster prediction module is used to obtain the disaster prediction result corresponding to each pixel position based on the feature vector and a preset clustering device.

[0027] In one embodiment of this application, the parameter calculation module is specifically used for:

[0028] Calculate the temporal coherence coefficient γ of temporal images i and j within the local region of pixel location (x, y) using the following formula. ij (x,y):

[0029]

[0030] Where N is the total number of pixel windows included in the local region; S i (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image i; |S i (x,y;n)| 2 For S i The signal strength of (x, y; n) S; j (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image j; |S j (x,y;n)| 2 For S j Signal strength of (x, y; n); For S j The conjugate signal of (x, y; n).

[0031] In one embodiment of this application, the parameter calculation module is specifically used for:

[0032] Calculate the normalized scattering intensity β(x,y) at pixel location (x,y) for any temporal image using the following formula:

[0033]

[0034] S(x,y;n) is the complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any temporal image; |S(x,y;n)| 2 S(x,y;n) represents the signal strength; N represents the total number of pixel windows included in the local region; and P0 represents the reference power factor.

[0035] In one embodiment of this application, the disaster prediction module is specifically used for:

[0036] For each pixel location's feature vector, if the temporal coherence coefficient included in the feature vector is less than or equal to a preset first threshold, and the normalized scattering intensity is greater than or equal to a second threshold, then the pixel location is determined to have a disaster risk.

[0037] In one embodiment of this application, the apparatus further includes:

[0038] The image processing module is used to perform time-series correction and registration on each of the phase images, and to apply a preset confidence factor for adaptive filtering.

[0039] Thirdly, embodiments of this application provide a computer device, including: a memory and a processor, wherein the memory and the processor are coupled together;

[0040] The memory is used to store one or more computer instructions;

[0041] The processor is used to execute one or more computer instructions to implement the SAR satellite-based disaster prediction method as described in the first aspect.

[0042] Fourthly, embodiments of this application provide a computer-readable storage medium storing one or more computer instructions that are executed by a processor to implement the SAR satellite-based disaster prediction method as described in the first aspect above.

[0043] Fifthly, this application provides a computer program product, which includes a computer program that, when executed by a processor, implements the SAR satellite-based disaster prediction method described in the first aspect.

[0044] In this embodiment, after acquiring multi-temporal images from SAR satellites, the temporal coherence coefficients of every two temporal images at each pixel location are calculated, along with the normalized scattering intensity of each pixel location in each temporal image. Further, wavelet thresholding denoising is applied to each temporal coherence coefficient, and the processed result is fused with the normalized scattering intensity of the corresponding pixel location to obtain a feature vector. Then, based on the feature vector and a preset clustering method, the disaster prediction result corresponding to each pixel location can be obtained. In other words, joint feature detection based on temporal coherence and scattering intensity can be performed on multi-temporal images simultaneously. Compared to methods relying on single-temporal echo intensity, this approach can extract more sensitive data, enabling the detection of subtle changes in surface deformation in the early stages of a disaster, thereby improving the accuracy of disaster detection. Furthermore, in this embodiment, the clustering method outputs pixel-level anomaly probabilities, rather than fuzzy classification results, thus allowing for accurate location of disaster areas when anomalies are detected, further improving the accuracy of disaster prediction. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0046] Figure 1 A flowchart illustrating a disaster prediction method based on SAR satellites provided in an embodiment of this application is shown.

[0047] Figure 2 A schematic diagram of the structure of a disaster prediction device based on SAR satellite provided in an embodiment of this application is shown;

[0048] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0050] It should be noted that the terms "comprising" and "having," and any variations thereof, in the embodiments and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0051] This application discloses a disaster prediction method, apparatus, device, and storage medium based on SAR satellites, which can improve the noise resistance of disaster detection and thus improve the accuracy of disaster detection. The embodiments of this application are described in detail below.

[0052] Figure 1 This application illustrates a disaster prediction method based on SAR satellites, which can be applied to electronic devices. The process may include the following steps:

[0053] S110: Acquire multi-temporal images from SAR satellites.

[0054] In this embodiment of the application, the SAR installed on the satellite can acquire images of the monitored area at a certain frequency, which are called multi-temporal images. The resolution of the multi-temporal images can be set according to the actual monitored area, such as 5 meters, 6 meters, 8 meters, etc., which are all possible.

[0055] Furthermore, the multi-temporal images acquired by SAR satellites can be stored in a fixed storage space. Electronic devices can then retrieve these multi-temporal images from this storage space, enabling disaster prediction based on the acquired images. For example, Sentinel-1A dual-polarization images with a resolution of 5 meters can be acquired, covering a time range from January 2023 to December 2024, totaling 24 temporal phases.

[0056] It is understandable that although the acquired multi-temporal images are collected by SAR in the same area, in practical applications, due to factors such as weather conditions, the geographical locations corresponding to the acquired multi-temporal images may not be exactly the same, which will affect the accuracy of disaster prediction. In addition, speckle noise in the temporal images will also affect the accuracy of disaster prediction.

[0057] In this embodiment of the application, in order to improve the accuracy of disaster prediction, after acquiring multi-temporal images, each temporal image can be preprocessed first, and then disaster prediction can be performed based on the preprocessed temporal data. For example, temporal correction and registration can be performed on each temporal image, and adaptive filtering can be applied using a preset confidence factor.

[0058] Temporal correction and registration refers to the precise alignment of multiple SAR images from different times (or sensors) to a unified spatial reference system, so that their pixels have consistent positional and geometric correspondences. Its main function is to strictly align SAR images from multiple time periods, so that each pixel corresponds to the same geographical location in space.

[0059] During SAR image acquisition, if images acquired at different times exhibit positional drift (even by only a few pixels), it can lead to errors in coherence calculations and prevent accurate deformation extraction. Furthermore, SAR images are highly sensitive to phase information; even slight geometric misalignments can be interpreted as surface deformation, resulting in false positive disaster signals. Therefore, in this embodiment, temporal correction and registration of images from different time phases effectively avoids these problems and improves the accuracy of disaster prediction.

[0060] In one implementation, a cross-correlation matching algorithm based on ORB (ORiented Brief) feature points can be used to perform temporal correction and registration on images of each phase, ultimately achieving pixel-level registration accuracy better than 0.1 pixels.

[0061] Adaptive filtering, also known as "despeculiar filtering," is a unique image filtering technique in SAR image processing. It is specifically designed to handle speckle noise, a common feature in SAR images. For example, the Lee filter can be used to filter images at different time phases.

[0062] Lee filtering is a classic adaptive filtering method specifically designed for speckle removal in SAR images, balancing the goals of "smoothing noise" and "preserving edges." The processed image can smooth flat areas while maintaining clarity in textured areas. The algorithm principle is as follows:

[0063] Let the image pixels be I(x,y) and the local window be W (e.g., 3x3 or 5x5). Then, any temporal image can be filtered according to the following steps:

[0064] 1. Calculate the local mean μ and variance σ 2 :

[0065] μ = mean(W), σ 2 =var(W)

[0066] 2. Calculate the overall image noise ratio k:

[0067]

[0068] 3. Update the center pixel value:

[0069] I filtered(x,y)=μ+k·(I(x,y)-μ)

[0070] 4. Window sliding, repeat full image processing.

[0071] I(x,y) is the pixel intensity value (with speckle noise) at position (x,y) in the original SAR image, μ is the average value of pixels within a local window centered at (x,y) (e.g., 3x3 or 5x5); σ 2 is the pixel variance of the local window, reflecting regional texture fluctuations; k is the filter weight coefficient, determining the "mixing degree" between the original pixels and the mean; ENL (Equivalent Number of Looks) is the effective number of views, a parameter measuring the speckle noise intensity of the SAR system, usually a known constant or an estimated value; I filtered (x,y) represents the pixel values ​​of the output image after speckle removal (i.e., the smoothed pixels).

[0072] S120: Calculate the temporal coherence coefficients of each pair of time-phase images at each pixel location, and calculate the normalized scattering intensity of each time-phase image at each pixel location.

[0073] In this embodiment, the temporal coherence coefficients of every two temporal images at each pixel location can be calculated to evaluate the temporal coherence of every two temporal images at corresponding pixel locations. Temporal coherence refers to the correlation of time series data at different time points. In time series analysis, coherence analysis is an important tool used to reveal the intrinsic relationships and mutual influences between time series.

[0074] Furthermore, the normalized scattering intensity at each pixel location in each temporal image can be calculated to evaluate the scattered light intensity of the temporal image. This scattered light intensity can reflect terrain features; for example, the greater the terrain undulation, the greater the scattering intensity, and the smaller the terrain undulation, the smaller the scattering intensity.

[0075] In one implementation, the following formula can be used to calculate the temporal coherence coefficient γ of temporal images i and j within the local region of pixel location (x,y). ij (x,y):

[0076]

[0077] Where N is the total number of pixel windows included in the local region; S i (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image i; |S i (x,y;n)| 2 For S i The signal strength of (x, y; n) S; j(x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image j; |S j (x,y;n)| 2 For S j Signal strength of (x, y; n); For S j The conjugate signal of (x, y; n).

[0078] The normalized scattering intensity β(x,y) at pixel location (x,y) of any temporal image can be calculated using the following formula:

[0079]

[0080] S(x,y;n) is the complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any temporal image; |S(x,y;n)| 2 S(x,y;n) represents the signal strength; N represents the total number of pixel windows included in the local region; and P0 represents the reference power factor.

[0081] S130: Perform wavelet thresholding denoising on each temporal coherence coefficient, and fuse the processed result with the normalized scattering intensity at the corresponding pixel position to obtain the feature vector.

[0082] In this embodiment, the calculated temporal coherence coefficients can also be subjected to wavelet threshold denoising. This allows for multi-scale signal decomposition, utilizing the differences in wavelet coefficients between noise and signal at different frequency bands. Thresholding suppresses noise and reconstructs the signal, thereby improving the accuracy of disaster prediction. For example, wavelet threshold denoising can be performed on the temporal coherence coefficients according to the following steps:

[0083] Step 1: For γ ij Perform wavelet decomposition on (x,y);

[0084] For example, two-dimensional discrete wavelet transform (2D-DWT), such as the Daubechies-4 (db4) wavelet, can be used to perform level 1 or level 2 wavelet decomposition on the image. The one-dimensional original signal is then subjected to low-pass filtering, high-pass filtering, and binary downsampling to obtain the low-frequency component L and the high-frequency component H of the signal, ultimately resulting in multiple subbands: LL, LH, HL, and HH. Among them, the low-frequency component LL includes the contour structure of the image, while the high-frequency components LH, HL, and HH include the edges and noise of the image.

[0085] Step 2: Apply a threshold function to denoise in the wavelet domain;

[0086] Specifically, soft thresholding or hard thresholding methods can be used to process the high-frequency subbands (LH, HL, HH), as shown in the following formulas:

[0087] Soft threshold W':

[0088]

[0089] Where W is the wavelet coefficient and T is the threshold, which can be set to... Where σ is the standard deviation of the high-frequency coefficients, and N is the number of coefficients.

[0090] After wavelet thresholding denoising of each temporal coherence coefficient, the processed result can be fused with the normalized scattering intensity of the corresponding pixel position to obtain a feature vector. This feature vector can fully reflect the geomorphic features of the monitoring area. Based on this feature vector, disaster prediction can be performed in the monitoring area.

[0091] S140: Based on the feature vector, obtain the disaster prediction results corresponding to each pixel position using a preset clustering method.

[0092] In this embodiment, the disaster risk of the monitoring area can be predicted. For example, disasters may include landslides, land subsidence, etc. Furthermore, under normal circumstances, the temporal images of the same monitoring area without disaster risk will not differ significantly; that is, their temporal coherence should be high, and the corresponding temporal coherence coefficient should also be high. Conversely, the scattering intensity of temporal images of normal terrain should be low.

[0093] In one implementation of this application, disaster prediction results corresponding to each pixel location can be obtained based on the feature vector and a preset clustering method. For example, for the feature vector of each pixel location, if the temporal coherence coefficient included in the feature vector is less than or equal to a preset first threshold, and the normalized scattering intensity is greater than or equal to a second threshold, it can be determined that there is a disaster risk at that pixel location. The first and second thresholds can be set according to actual conditions, such as 0.6, 0.7, etc., which are all acceptable, and this application does not limit their specific values. Furthermore, the first and second thresholds can be the same or different, which is also acceptable.

[0094] For example, disaster prediction can be performed according to the following steps:

[0095] 1. Construct feature vectors;

[0096] For each pixel position (x, y) in the image, the following feature vector can be extracted:

[0097]

[0098] The explanation is as follows:

[0099]

[0100] In this way, an n-dimensional vector (e.g., 5-dimensional) can be formed for each pixel position, which can be used for subsequent clustering.

[0101] 2. GMM (Gaussian Mixture Module): Uses Gaussian mixture models for unsupervised clustering;

[0102] Gaussian Mixture Model (GMM) principle: Assume the data consists of K distinct Gaussian distributions, each representing a "pixel category"; use the Expectation-Maximization (EM) algorithm to fit the model, estimating the probability that each pixel belongs to each category. For each pixel i, the GMM outputs a vector:

[0103] P i =[p i1 ,p i2 ,...,p ik ]

[0104] Where p ik This represents the probability that pixel i belongs to the k-th Gaussian component.

[0105] In this embodiment of the application, the feature vector set of all pixels Perform Gaussian mixture modeling.

[0106] 3. Determine the anomaly category;

[0107] For example, classes containing characteristics such as average low coherence and strong wave scattering can be considered "abnormal distributions"; the corresponding pixel anomalous probability p a Take the probability that it belongs to this exception class:

[0108] p a (x,y)=p i Abnormal distribution

[0109] When p a When (x,y)>0.7, the pixel is determined to be a potential disaster area.

[0110] In this embodiment, after acquiring multi-temporal images from SAR satellites, the temporal coherence coefficients of every two temporal images at each pixel location are calculated, along with the normalized scattering intensity of each pixel location in each temporal image. Further, wavelet thresholding denoising is applied to each temporal coherence coefficient, and the processed result is fused with the normalized scattering intensity of the corresponding pixel location to obtain a feature vector. Then, based on the feature vector and a preset clustering method, the disaster prediction result corresponding to each pixel location can be obtained. In other words, joint feature detection based on temporal coherence and scattering intensity can be performed on multi-temporal images simultaneously. Compared to methods relying on single-temporal echo intensity, this approach can extract more sensitive data, enabling the detection of subtle changes in surface deformation in the early stages of a disaster, thereby improving the accuracy of disaster detection. Furthermore, in this embodiment, the clustering method outputs pixel-level anomaly probabilities, rather than fuzzy classification results, thus allowing for accurate location of disaster areas when anomalies are detected, further improving the accuracy of disaster prediction.

[0111] Figure 2 This illustration shows a structural diagram of a disaster prediction device based on a synthetic aperture radar (SAR) satellite, according to an embodiment of this application. The device includes:

[0112] Image acquisition module 210 is used to acquire multi-temporal images collected by SAR satellites;

[0113] The parameter calculation module 220 is used to calculate the temporal coherence coefficient of each pair of the time-phase images at each pixel position, and to calculate the normalized scattering intensity of each time-phase image at each pixel position.

[0114] The vector construction module 230 is used to perform wavelet threshold denoising on each of the temporal coherence coefficients, and fuse the processed result with the normalized scattering intensity at the corresponding pixel position to obtain a feature vector.

[0115] The disaster prediction module 240 is used to obtain the disaster prediction result corresponding to each pixel position based on the feature vector and a preset clustering device.

[0116] In one embodiment of this application, the parameter calculation module 220 is specifically used for:

[0117] Calculate the temporal coherence coefficient γ of temporal images i and j within the local region of pixel location (x, y) using the following formula. ij (x,y):

[0118]

[0119] Where N is the total number of pixel windows included in the local region; S i(x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image i; |S i (x,y;n)| 2 For S i The signal strength of (x, y; n) S; j (x, y; n) represents the complex scattering signal of the nth pixel in the local region of pixel position (x, y) of the temporal image j; |S j (x,y;n)| 2 For S j Signal strength of (x, y; n); For S j The conjugate signal of (x, y; n).

[0120] In one embodiment of this application, the parameter calculation module 220 is specifically used for:

[0121] Calculate the normalized scattering intensity β(x,y) at pixel location (x,y) for any temporal image using the following formula:

[0122]

[0123] S(x,y;n) is the complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any temporal image; |S(x,y;n)| 2 S(x,y;n) represents the signal strength; N represents the total number of pixel windows included in the local region; and P0 represents the reference power factor.

[0124] In one embodiment of this application, the disaster prediction module 240 is specifically used for:

[0125] For each pixel location's feature vector, if the temporal coherence coefficient included in the feature vector is less than or equal to a preset first threshold, and the normalized scattering intensity is greater than or equal to a second threshold, then the pixel location is determined to have a disaster risk.

[0126] In one embodiment of this application, the apparatus further includes:

[0127] The image processing module is used to perform time-series correction and registration on each of the phase images, and to apply a preset confidence factor for adaptive filtering.

[0128] In this embodiment, after acquiring multi-temporal images from SAR satellites, the temporal coherence coefficients of every two temporal images at each pixel location are calculated, along with the normalized scattering intensity of each pixel location in each temporal image. Further, wavelet thresholding denoising is applied to each temporal coherence coefficient, and the processed result is fused with the normalized scattering intensity of the corresponding pixel location to obtain a feature vector. Then, based on the feature vector and a preset clustering method, the disaster prediction result corresponding to each pixel location can be obtained. In other words, joint feature detection based on temporal coherence and scattering intensity can be performed on multi-temporal images simultaneously. Compared to methods relying on single-temporal echo intensity, this approach can extract more sensitive data, enabling the detection of subtle changes in surface deformation in the early stages of a disaster, thereby improving the accuracy of disaster detection. Furthermore, in this embodiment, the clustering method outputs pixel-level anomaly probabilities, rather than fuzzy classification results, thus allowing for accurate location of disaster areas when anomalies are detected, further improving the accuracy of disaster prediction.

[0129] The following describes a computer device provided in an embodiment of this application. Please refer to [link / reference needed]. Figure 3 , Figure 3 A schematic diagram of a computer device provided in an embodiment of this application, the computer device comprising:

[0130] One or more processors 40;

[0131] The processor 40 is coupled to a storage device 41, which is used to store one or more programs.

[0132] When the one or more programs are executed by the one or more processors 40, the electronic device performs the following functions: Figure 1 The technical solution of a disaster prediction method based on SAR satellites is described above.

[0133] This application also provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement... Figure 1 The technical solution of a disaster prediction method based on SAR satellites is described above.

[0134] This application provides a computer program product, which includes a computer program that, when executed by a processor, implements the following: Figure 1 The technical solution of a disaster prediction method based on SAR satellites is described above.

[0135] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of one embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application.

[0136] Those skilled in the art will understand that the modules in the apparatus of the embodiments can be distributed in the apparatus of the embodiments as described in the embodiments, or they can be located in one or more devices different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A disaster prediction method based on synthetic aperture radar (SAR) satellites, characterized in that, The method includes: Acquire multi-temporal images from SAR satellites; Calculate the pixel position of every two of the aforementioned temporal images. Temporal coherence coefficient And calculate the normalized scattering intensity of each phase image at each pixel position; Wavelet thresholding denoising is performed on each of the temporal coherence coefficients, and the processed result is fused with the normalized scattering intensity at the corresponding pixel position to obtain a feature vector; Based on the feature vector, disaster prediction results corresponding to each pixel location are obtained using a preset clustering method; The steps for obtaining the disaster prediction results corresponding to each pixel location include: Step 1, constructing a feature vector; for each pixel location in the image... The following feature vectors were extracted: ; in, At the pixel position The average coherence coefficient; For pixel position The average scattering intensity; The variance of the coherence variation; The degree of scattering intensity fluctuation; This represents a change in scattering trends; Step 2: Perform unsupervised clustering using a Gaussian Mixture Model (GMM). The data consists of K distinct Gaussian distributions, each representing a "pixel category." The EM algorithm is used for model fitting to estimate the probability of each pixel belonging to each category. For each pixel i, the GMM outputs a vector. ;in Represents the probability that pixel i belongs to the k-th Gaussian component; the set of feature vectors for all pixels. Perform Gaussian mixture modeling; Step 3: Determine the anomaly category: Determining the anomaly category includes classifying those containing average low coherence and strong wave scattering characteristics as "anomaly distributions"; the corresponding pixel anomaly probability. Take the probability that it belongs to that exception class. : ; when At that time, the pixel was determined to be a potential disaster area.

2. The method according to claim 1, characterized in that, The calculation of the temporal coherence coefficient at each pixel position for every two of the aforementioned temporal images. The steps include: Calculate the time-phase image using the following formula. i and phase image j Temporal coherence coefficients within the local region of pixel location (x,y) : ； in, The total number of pixel windows included in the local region; For phase images i At pixel position The complex scattering signal of the nth pixel within a local region; for The signal strength; For phase images j At pixel position The complex scattering signal of the nth pixel within a local region; for The signal strength; for The conjugate signal.

3. The method according to claim 1, characterized in that, The step of calculating the normalized scattering intensity at each pixel location of each of the temporal images includes: Calculate the normalized scattering intensity at pixel location (x, y) of any temporal image using the following formula. : The complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any given temporal image; for The signal strength; The total number of pixel windows included in the local region; This is the reference power factor.

4. The method according to claim 1, characterized in that, The step of obtaining the disaster prediction result corresponding to each pixel position based on the feature vector and a preset clustering method includes: For each pixel location's feature vector, if the temporal coherence coefficient included in the feature vector is less than or equal to a preset first threshold, and the normalized scattering intensity is greater than or equal to a second threshold, then the pixel location is determined to have a disaster risk.

5. The method according to any one of claims 1-4, characterized in that, After acquiring the multi-temporal images collected by SAR satellites, the method further includes: The time-phase images are subjected to time-series correction and registration, and adaptive filtering is performed using a preset confidence factor.

6. A disaster prediction device based on the disaster prediction method of synthetic aperture radar (SAR) satellite according to any one of claims 1-5, characterized in that, The device includes: The image acquisition module is used to acquire multi-temporal images collected by SAR satellites; The parameter calculation module is used to calculate the temporal coherence coefficient of each pair of the time-phase images at each pixel position, and to calculate the normalized scattering intensity of each time-phase image at each pixel position. The vector construction module is used to perform wavelet threshold denoising on each of the temporal coherence coefficients, and fuse the processed result with the normalized scattering intensity at the corresponding pixel position to obtain a feature vector; The disaster prediction module is used to obtain the disaster prediction result corresponding to each pixel position based on the feature vector and a preset clustering device.

7. The apparatus according to claim 6, characterized in that, The parameter calculation module is specifically used for: Calculate the time-phase image using the following formula. i and phase image j Temporal coherence coefficients within the local region of pixel location (x,y) : in, The total number of pixel windows included in the local region; For phase images i The complex scattering signal of the nth pixel within the local region of pixel position (x,y); for The signal strength; For phase images j The complex scattering signal of the nth pixel within the local region of pixel position (x,y); for The signal strength; for The conjugate signal.

8. The apparatus according to claim 6, characterized in that, The parameter calculation module is specifically used for: Calculate the normalized scattering intensity at pixel location (x, y) of any temporal image using the following formula. : The complex scattering signal of the nth pixel in the local region of pixel position (x,y) of any given temporal image; for The signal strength; The total number of pixel windows included in the local region; This is the reference power factor.

9. A computer device, characterized in that, include: The memory and the processor are coupled; The memory is used to store one or more computer instructions; The processor is used to execute one or more computer instructions to implement the disaster prediction method based on synthetic aperture radar (SAR) satellites as described in any one of claims 1 to 5.

10. A readable storage medium having stored thereon one or more computer instructions, characterized in that, The instruction is executed by the processor to implement the disaster prediction method based on synthetic aperture radar (SAR) satellites as described in any one of claims 1 to 5.

Images