A method and system for extracting earthquake-damaged buildings based on multi-temporal sar images

Abstract

The application discloses a method and system for extracting earthquake-damaged buildings based on multi-temporal SAR images, and relates to the technical field of post-disaster loss assessment, and comprises the following steps: collecting SAR image interannual observation data covering buildings in a core earthquake area; performing process-based processing on the multi-temporal SAR images to generate time sequence correlation data of an observation area; performing differential interference processing based on InSAR technology to generate time sequence coherence data of the observation area; performing change detection on the multi-temporal images based on a quartile outlier test principle, setting a threshold, and combining building vector data to obtain an earthquake-damaged building distribution situation diagram. The application combines SAR image intensity correlation, SAR image phase coherence and the quartile outlier test principle, and improves the recognition ability of earthquake-damaged buildings in a complex environment.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of post-disaster loss assessment, and particularly relates to a method and system for extracting earthquake-damaged buildings based on multi-temporal SAR images. BACKGROUND

[0002] Earthquake is one of the natural disasters that endanger human beings the most. Strong earthquakes often cause serious casualties and economic losses. With the current scientific and technological level, it is still impossible to predict the occurrence of earthquakes, so obtaining the disaster situation in a short time is the focus of earthquake emergency work. The energy released after an earthquake can cause buildings to collapse and damage, and the trapped personnel will cause casualties if they are not rescued in time. Traditional ground investigation work is time-consuming and dangerous, while remote sensing technology has the characteristics of macro-quick and not being limited by the ground, which can provide technical support for earthquake relief work. Among them, Synthetic Aperture Radar (SAR) has the characteristics of all-weather and all-day, and can obtain disaster area images in bad weather conditions such as cloudy, rainy and foggy, which is an important data source for post-earthquake disaster assessment. SAR images have intensity and phase information, which can be designed from these two aspects. The direct echo intensity of SAR image is represented by backscatter coefficient, and the difference in linear relationship between two SAR images can be measured by using the correlation coefficient in statistics. This method considers the adjacent area in the sliding window instead of a single pixel, which can reduce the interference of noise. When the ground surface changes in a short time, theoretically, the correlation will weaken, so the intensity correlation can be used to detect earthquake-damaged buildings. Interferometric Synthetic Aperture Radar (InSAR) technology is widely used in ground deformation monitoring, and the phase coherence can be obtained after differential interference processing. The phase coherence is an important indicator for evaluating the similarity of two SAR images, and like intensity correlation, it can also be used to identify earthquake-damaged buildings.

[0003] At present, the method for extracting earthquake-damaged building information from SAR images can be based on the multi-temporal correlation or coherence indicators mentioned in the previous paragraph to detect changes and extract abnormal value regions, which can be considered as potential earthquake-damaged building regions. However, the processing of multi-temporal remote sensing images involves high-dimensional matrix operations, and common data outlier detection methods such as hypothesis testing require data to meet the normal distribution probability model. Methods such as density clustering are relatively complex and difficult to quickly analyze outliers in multi-temporal remote sensing images, which are limited in actual emergency work.

[0004] Therefore, there is an urgent need to provide a solution to improve the above problems and quickly extract abnormal value regions from multi-temporal images. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a method and system for extracting earthquake-damaged buildings based on multi-temporal SAR images, which can quickly perform outlier detection and data analysis on multi-temporal SAR images, and obtain an earthquake-damaged building distribution map in combination with building vector data. It is worth noting that the present application takes a sequence of earthquakes in a certain city on October 7, 2023 as a real application case for illustration, but the method proposed by the present application is not limited to this case.

[0006] In order to solve the above technical problems, the present application is implemented as follows:

[0007] In a first aspect, the embodiments of the present application provide a method for extracting earthquake-damaged buildings based on multi-temporal SAR images, which comprises:

[0008] Collecting relevant data covering the observation area, the observation area including an area where there are earthquake-damaged buildings, and the relevant data including image, time, terrain and location information;

[0009] Performing a flow processing on the multi-temporal SAR images, generating backscattering coefficients through radiation calibration, multi-view processing and registration steps, further generating time series correlation data of the observation area, and taking pre-earthquake correlation as a reference image to perform histogram matching on other correlation images to obtain a set of matched correlation images;

[0010] Performing differential interference processing on the set of matched correlation images using InSAR technology, and performing baseline optimization, again performing differential interference processing to generate time series coherence data of the observation area, and taking pre-earthquake coherence as a reference image to perform histogram matching on other coherence images to obtain a set of matched coherence images;

[0011] Performing normalized difference processing on the set of matched correlation images and the set of matched coherence images adjacent in time to construct a time series correlation difference or time series coherence difference data set, including a pre-earthquake correlation difference, a pre-earthquake coherence difference data set and a co-seismic correlation difference, a co-seismic coherence difference data set;

[0012] Based on the quartile outlier detection principle, performing outlier detection analysis on the time series correlation difference or time series coherence difference data set, and obtaining an earthquake-damaged building distribution map in combination with building vector data.

[0013] Optionally, the step of collecting relevant data covering the observation area comprises:

[0014] Determining the vector range of the disaster area according to the epicenter position;

[0015] According to the vector range, a rising track or a descending track SAR image set, digital elevation model data and precise orbit data are acquired to collect relevant data covering the observation area.

[0016] Optionally, the specific steps of generating the time correlation data of the observation area include:

[0017] Radiometric calibration is performed on the SAR image set to obtain a backscattering coefficient with actual physical meaning; and the calibrated SAR image set is processed according to a 4:1 multi-view ratio for multi-view processing;

[0018] All images in the calibrated SAR image set are registered to obtain an aligned SAR image set, and a pre-earthquake SAR image in the calibrated SAR image set is taken as a master image, all slave images are registered with the master image to obtain registered slave images;

[0019] Resampling is performed on each registered slave image to obtain a resampled slave image, and an aligned SAR image set is obtained; all images in the aligned SAR image set are arranged in time sequence, and a master-slave relationship is established between adjacent images;

[0020] The absolute values of correlation coefficients are calculated based on a sliding window according to the master-slave relationship between adjacent SAR images in time to generate a correlation data set;

[0021] Taking a pre-earthquake correlation image as a reference image, histogram matching is performed on other correlation images to obtain a matched correlation image set of the observation area.

[0022] Optionally, the specific steps of generating the time correlation data of the observation area include:

[0023] All images in the SAR image set are registered to obtain an aligned SAR image set, and a pre-earthquake SAR image in the SAR image set is taken as a master image, all slave images are registered with the master image to obtain registered slave images;

[0024] Resampling is performed on each registered slave image to obtain a resampled slave image, and an aligned SAR image set is obtained; all images in the aligned SAR image set are arranged in time sequence, and a master-slave relationship is established between adjacent images;

[0025] Difference interference processing is performed on adjacent SAR images in time according to the master-slave relationship by using InSAR technology to remove terrain and reference surface phases in an initial interferogram, quality control interferograms and preliminary coherence are generated;

[0026] Filtering is performed on the quality control interferograms to obtain filtered interferograms; phase unwrapping processing is performed on the filtered interferograms to obtain unwrapped interferograms;

[0027] The baseline optimization is performed on the unwrapped interferogram, the baseline optimization is performed by combining the ground control points, and the time series coherence data set is obtained by performing differential interference processing again based on the precise baseline;

[0028] The post-earthquake coherence is taken as a reference image, and histogram matching is performed on other coherence images to obtain a matched coherence image set of the observation area.

[0029] Optionally, the specific process of constructing the time series correlation difference or time series coherence difference data set comprises the following steps:

[0030] All images in the matched correlation or matched coherence data set are arranged in time sequence, and normalized difference operation is performed between adjacent images to obtain the correlation difference or coherence difference data set, and the last image is referred to as the co-seismic correlation difference or co-seismic coherence difference.

[0031] Optionally, the specific steps of obtaining the building damage distribution diagram comprise the following steps:

[0032] Based on the pre-earthquake time series correlation difference or coherence difference data set, image synthesis is performed to obtain the 25% quantile image Q1 and the 75% quantile image Q3, and the difference between Q3 and Q1 is obtained to obtain the interquartile range image IQR, wherein the pre-earthquake time series correlation difference or coherence difference data set does not contain the co-seismic correlation difference or co-seismic coherence difference.

[0033] Based on the principle of quartile anomaly value test, the sum of the image Q3 and the image IQR is regarded as a background threshold image, that is, the upper boundary of the data, and the value exceeding the limit is regarded as an anomaly value.

[0034] The co-seismic correlation difference or co-seismic coherence difference exceeding the background threshold image and being positive is regarded as a potential building damage area.

[0035] Based on the building vector data, the building position is located, a threshold is set for the potential building damage area, and the greater the pixel value, the greater the building damage degree.

[0036] In the second aspect, the embodiments of the present application provide a system for extracting building damage based on multi-temporal SAR images, and the system comprises:

[0037] The regional data module is used to collect relevant data covering the observation area, the observation area comprises a region where there is building damage, and the relevant data comprises image, time, terrain and position information;

[0038] The calculation time correlation module performs process-based processing on the multi-temporal SAR images, generates the backscattering coefficient through radiation calibration, multi-view processing and registration steps, and further generates the time series correlation data of the observation area; the pre-earthquake correlation is taken as a reference image, and histogram matching is performed on other correlation images to obtain a matched correlation image set.

[0039] The time-series correlation difference or time-series coherence difference module performs normalized difference processing on the time-series correlation and coherence images adjacent in time to construct a time-series correlation difference or time-series coherence difference data set, including a pre-seismic correlation difference, a pre-seismic coherence difference data set and a co-seismic correlation difference, a co-seismic coherence difference data;

[0040] The time-series correlation difference or time-series coherence difference module performs normalized difference processing on the time-series correlation and coherence images adjacent in time to construct a time-series correlation difference or time-series coherence difference data set, including a pre-seismic correlation difference, a pre-seismic coherence difference data set and a co-seismic correlation difference, a co-seismic coherence difference data;

[0041] The multi-temporal image change detection module performs outlier detection analysis on the time-series correlation difference or time-series coherence difference data set based on the quartile outlier detection principle, and obtains a distribution map of damaged buildings in combination with the building vector data.

[0042] In a third aspect, an electronic device is provided, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the method of the first aspect.

[0043] In a fourth aspect, a readable storage medium is provided, which stores a program or instructions, and the program or instructions, when executed by a processor, implement the steps of the method of the first aspect.

[0044] Compared with the prior art, the method for extracting damaged buildings based on multi-temporal SAR images has the following beneficial effects: first, the correlation data of the observation area is obtained, wherein the observation area includes the area of the core earthquake zone, and the correlation data includes image, time, terrain and position information; second, the multi-temporal SAR images are processed in a flow, and the time-series correlation data of the observation area is generated; third, the InSAR technology is used for differential interference processing, and the time-series coherence data of the observation area is generated; and fourth, the quartile outlier detection principle is used for change detection of the multi-temporal images, a threshold is set, and the distribution map of damaged buildings is obtained in combination with the building vector data. The quartile outlier detection method does not need to assume that the data conforms to any probability model and is simple to operate. The method combines SAR image intensity correlation, SAR image phase coherence and the quartile outlier detection principle, can more objectively identify the abnormal high-value area in the co-seismic correlation or coherence difference, and improves the recognition ability of damaged buildings in complex environments by combining the results of the two. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1Fig. 1 shows a flow chart of a method for extracting earthquake-damaged buildings based on intensity correlation of multi-temporal SAR images according to the first embodiment of the present application;

[0046] Figure 2 Fig. 2 shows a flow chart of a method for extracting earthquake-damaged buildings based on phase coherence of multi-temporal SAR images according to the second embodiment of the present application;

[0047] Figure 3 Fig. 3 shows a result of processing of pre-earthquake time-series intensity difference image set according to the first embodiment of the present application;

[0048] Figure 4 Fig. 4 shows a result of quartile outlier test on multi-temporal intensity difference according to the first embodiment of the present application;

[0049] Figure 5 Fig. 5 shows a result of processing of pre-earthquake time-series phase coherence difference image set according to the second embodiment of the present application;

[0050] Figure 6 Fig. 6 shows a result of quartile outlier test on multi-temporal phase coherence difference according to the second embodiment of the present application;

[0051] Figure 7 Fig. 7 shows a result of extracting earthquake-damaged buildings in a sequence of villages in a city according to the first embodiment of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0053] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein. In addition, "and / or or" in the specification and claims means at least one of the connected objects, and the character "or" generally means an "or" relationship between the front and rear associated objects.

[0054] The method and system for extracting earthquake-damaged buildings based on multi-temporal SAR images according to the embodiments of the present application will be described in detail below with reference to the drawings and specific embodiments and their application scenarios.

[0055] Embodiment 1

[0056] The linear relationship difference between two SAR images can be measured by using the correlation coefficient in statistics, which considers the adjacent areas in the sliding window instead of a single pixel, and can reduce the interference of noise. When the ground surface changes in a short time, theoretically, the correlation will weaken, so the intensity correlation can be used to detect the damaged buildings. The embodiment is based on the characteristics of the intensity correlation of multi-temporal SAR images to design a method for extracting damaged buildings based on the intensity correlation of multi-temporal SAR images.

[0057] Specifically, the application provides a method for extracting damaged buildings based on the intensity correlation of multi-temporal SAR images, which comprises steps S1 to S5.

[0058] S1, collect relevant data covering the observation area, the observation area including the area where the damaged buildings exist, and the relevant data including image, time, terrain and location information.

[0059] Specifically, the observation area includes the area where the damaged buildings exist, and the relevant data includes image, time, terrain and location information, specifically including the vector range of the core disaster area; the ascending or descending SAR image set, the accurate time of the earthquake occurrence and the epicenter position, the digital elevation model and the precise orbit data are obtained according to the vector range.

[0060] Further, the relevant data covering the area where the damaged buildings exist are obtained, including SAR image data and auxiliary data, the SAR image data including ascending SAR images or descending SAR images, and the auxiliary data including the accurate time of the earthquake occurrence and the epicenter position, the digital elevation model data and the precise orbit data; the image selection includes all SAR images within one year before the earthquake and one SAR observation image immediately after the earthquake.

[0061] S2, process the multi-temporal SAR images in a flow, and generate the backscattering coefficient through radiation calibration, registration and multi-view processing, and further generate the time correlation data of the observation area.

[0062] Specifically, the SAR image set is subjected to radiation calibration processing to generate the backscattering cross section, and then the backscattering coefficient with actual physical significance is obtained through linear logarithmic conversion; the calibrated SAR image set is subjected to multi-view processing according to the multi-view ratio of 4:1 to enhance the signal-to-noise ratio of the SAR data.

[0063] Specifically, all images in the SAR image set are registered to obtain the aligned SAR image set; the image closest to the earthquake in the SAR image set is selected as the main image, and all slave images are registered with the main image to obtain the registered slave images; each registered slave image is resampled to obtain the resampled slave image; the coordinate position of the resampled slave image is consistent with that of the main image, and the resampled slave image and the main image form the registered SAR image set.

[0064] Specifically, for temporally adjacent registered SAR images, the absolute value of the correlation coefficient is calculated based on a sliding window according to the master-slave relationship, generating a correlation dataset; the correlation coefficient... The expression is:

[0065] ;

[0066] Statistical correlation coefficient The value range of is [-1, 1]. The correlation proposed in this invention refers to the correlation coefficient. Take the absolute value, with a range of [0,1]. The closer the absolute value of the two images is to 1, the stronger the linear relationship between them, that is, the higher their similarity. The closer the absolute value of is to 0, the worse the linear relationship between the two images, that is, the greater the difference. (In the formula...) and This refers to the pixel values ​​at the same position in two images. In practical applications, a 3×3 sliding window is selected, and the correlation is calculated through convolution operations. and This represents the mean. and This method represents the number of pixels, taking into account adjacent areas within the sliding window rather than individual pixels, which reduces noise interference and represents the degree of variation of that pixel.

[0067] S3. Using the pre-earthquake correlation as a reference image, perform histogram matching on other correlation images to obtain the matched adjacent correlation images.

[0068] Specifically, based on the temporal correlation dataset from the previous step, this invention selects pre-earthquake correlation as a reference image and performs histogram matching on other correlated images to ensure that the correlated images have a consistent data distribution, thus obtaining a matched temporal correlation dataset. Change detection not only requires high spatial registration accuracy for multi-temporal data, but also requires that the data distribution be as similar or identical as possible. The basic steps of histogram matching include: first, calculating the probability density function (PDF) and cumulative distribution function (CDF) of the source image and the reference image respectively; then, for each pixel value in the source image, finding its corresponding cumulative probability in the source image's CDF; next, finding the gray value corresponding to the value closest to this probability in the reference image's CDF, and using it as the new pixel value; finally, applying this mapping relationship to all pixels of the source image, thereby obtaining an output image that is as consistent as possible with the reference image in terms of grayscale distribution. This process essentially achieves probability distribution alignment through the inverse function of the CDF.

[0069] S4, performing normalized difference value processing on adjacent correlation images to construct a time-series correlation difference dataset, the correlation difference dataset including a pre-earthquake correlation difference dataset and a co-seismic correlation difference dataset.

[0070] Specifically, the present application arranges all images in the matched time-series correlation dataset in time sequence, performs normalized difference value operation between adjacent images to obtain a time-series coherence difference dataset, and the last image is called a co-seismic coherence difference. In a possible implementation, the correlation difference proposed by the present application refers to the normalized difference image of the pre-correlation and the post-correlation, similar to the normalized vegetation index NDVI, with a value range of [-1, 1]. This processing can highlight the difference between the pre-image and the post-image, and the data distribution histogram presents a normal distribution form.

[0071] S5, performing quartile outlier detection on the correlation difference dataset to obtain a building damage distribution map.

[0072] Specifically, based on the quartile outlier detection principle, the present application sets a threshold and combines building vector data to obtain a building damage distribution map, including: based on the time-series correlation difference dataset (not including the co-seismic correlation difference here), image synthesis is performed to obtain a background threshold image, that is, the upper boundary of the data, and values exceeding this boundary are regarded as outliers; the pixel value of the co-seismic correlation difference that exceeds the background threshold image and is positive is regarded as a potential building damage area, and the greater the value, the more serious the building damage.

[0073] Specifically, the present application uses the principle of quartile outlier detection to identify outliers in the co-seismic correlation difference as potential building damage areas. Quartiles are values that divide a set of data into four parts, including lower quartile (Q1), median (Q2) and upper quartile (Q3). IQR is the difference between Q3 and Q1, that is, IQR = Q3-Q1. The outliers in the present application are defined as values greater than Q3+IQR, and by comparing each data point with the above boundary, data points greater than the upper limit are regarded as outliers. This method is simple and effective, suitable for most cases, and does not require data to conform to any probability model. The background threshold map is defined as the sum of the upper quartile synthesis image and the quartile range synthesis image, which can extract only the correlation changes caused by earthquakes from natural changes, thereby achieving robust building damage information extraction. Based on the building vector data, the building position is located, the correlation influence caused by other ground object changes is excluded, and then a threshold is set for the potential building damage area, and values greater than the threshold are regarded as building damage, and the greater the pixel value, the more serious the building damage.

[0074] Example 2

[0075] Phase coherence is an important indicator to evaluate the similarity of two SAR images. When the ground surface changes in a short time, the coherence will theoretically weaken, which can be used to identify the damaged buildings after the earthquake. The embodiment designs a method for extracting earthquake-damaged buildings based on the phase coherence of multi-temporal SAR images and quartile outlier test.

[0076] Specifically, refer to Figure 2 , which represents the overall flowchart of the method for extracting earthquake-damaged buildings based on the phase coherence of multi-temporal SAR images provided by the present application, including steps S1 to S5.

[0077] S1, collect relevant data covering the observation area, the observation area including the area where there are earthquake-damaged buildings, and the relevant data including image, time, terrain and location information.

[0078] Specifically, the processing flow here is the same as S1 in Embodiment 1, which will not be repeated.

[0079] S2, difference interferometric processing is performed on the SAR images adjacent in time, and baseline optimization is performed to generate time series coherence data of the observation area.

[0080] Specifically, the present application uses InSAR technology to perform difference interferometric processing on SAR images adjacent in time, adopts a method combining ground control points to perform baseline optimization, and generates time series coherence data of the observation area based on precise baseline difference interferometric processing. Specifically, it includes: using InSAR technology to perform difference interferometric processing on SAR images adjacent in time to obtain an interferogram and preliminary coherence; filtering the interferogram to remove noise, and obtaining an unwrapped interferogram according to the filtered interferogram; using a method combining ground control points to perform baseline optimization, and generating time series coherence data of the observation area based on precise baseline difference interferometric processing. Coherence is an indicator to measure the similarity of two SAR images in InSAR. In the following formula, represents the correlation coefficient of two SAR images, and represents the complex value of a certain pixel of the SAR image, is the conjugate complex value, N represents the size of the sliding window. Essentially, the calculation formula of coherence is exactly the same as correlation, except that the operation object changes from real number to complex number. In practical application, a 3x3 sliding window is also selected.

[0081] ;

[0082] Specifically, the present application uses InSAR technology to obtain a time-series coherence dataset, specifically including: registering all images in a SAR image set to obtain an aligned SAR image set; specifically, selecting an image closest to the earthquake occurrence in the SAR image set as a master image, and registering all slave images with the master image to obtain registered slave images; performing resampling on each registered slave image to obtain a resampled slave image; wherein the coordinate position of the resampled slave image is consistent with the coordinate position of the master image, and the resampled slave image and the master image form the registered SAR image set.

[0083] Further, the present application arranges the registered SAR image set in chronological order, and establishes a master-slave relationship between adjacent images, which ensures that the time baseline between the interference pairs is minimized; difference interference processing is performed on the adjacent images after registration to remove the terrain and reference surface phases, and an interferogram and preliminary coherence are generated; the interferogram is processed using the Goldstein adaptive filtering algorithm to obtain a filtered interferogram; the filtered interferogram is unwrapped using the minimum cost flow method to obtain an unwrapped interferogram; and the unwrapped interferogram is baseline optimized, the baseline optimization is performed using a method combining ground control points, and a time-series coherence dataset is obtained based on the precise baseline through differential interference processing again.

[0084] Further, the present application uses SRTM 30m DEM data to remove the terrain phase in the interferogram, and then uses the Goldstein adaptive filtering method to filter all interferograms, with the filtering index set to 0.5; the filtered interferogram is unwrapped using the minimum cost flow method, with the coherence threshold set to 0.75 and the intensity threshold set to 0.05, and the unwrapping reference point is selected as a high-coherence point in the city, which ensures the quality of unwrapping and prepares for subsequent baseline optimization. The baseline optimization is performed using a method combining ground control points, a time-series coherence dataset is obtained based on the precise baseline through differential interference processing again, residual linear trend phases are removed, and a higher-quality coherence dataset is obtained.

[0085] S3, taking the pre-earthquake coherence as a reference image, histogram matching is performed on other coherence reference images to obtain matched adjacent coherence images.

[0086] Specifically, based on the time-series coherence dataset of the previous step, the pre-earthquake coherence is selected as a reference image, and histogram matching is performed on other coherence images to ensure that the coherence images have consistent data distribution, and a matched time-series coherence dataset is obtained. Coherence is easily affected by various factors, mainly including geometric decorrelation, time decorrelation, volume scattering decorrelation, thermal noise decorrelation, Doppler centroid decorrelation, and data processing decorrelation. The present application uses histogram matching technology to calibrate the data distribution of the coherence, and reduces the influence of decorrelation on subsequent processing. The basic steps of histogram matching are the same as S3 in Embodiment 1, and will not be repeated here.

[0087] S4, performing normalized difference value processing on adjacent coherence images to construct a time-series coherence difference dataset, the coherence difference dataset including a pre-earthquake coherence difference dataset and a co-seismic coherence difference.

[0088] Further, all images in the matched time-series coherence dataset are arranged in time sequence, normalized difference value operation is performed between adjacent images to obtain a time-series coherence difference dataset, and the last image is referred to as a co-seismic coherence difference. The calculation method of the coherence difference is the same as S4 in Embodiment 1, and will not be described here.

[0089] S5, performing quartile outlier detection on the coherence difference dataset to obtain a building damage distribution map.

[0090] Specifically, based on the quartile outlier detection principle, a threshold is set and combined with building vector data to obtain a building damage distribution map, including: based on the time-series coherence difference dataset (not including the co-seismic coherence difference here), a background threshold image is obtained by image synthesis, that is, the upper boundary of the data, and values exceeding this boundary are regarded as outliers; the pixel value of the co-seismic coherence difference image that exceeds the background threshold image and is positive is regarded as a potential building damage area, and the greater the value, the more serious the building damage. The principle of quartile outlier detection is the same as S5 in Embodiment 1, and will not be described here. The final building damage distribution map is the result of fusing Embodiment 1 and Embodiment 2, that is, the processing results of multi-time correlation and multi-time coherence, Figure 7 The results of extracting building damage in a village in a certain city in 2023 using the method proposed in the application are shown.

[0091] Embodiment 3

[0092] The third embodiment of the application further provides a system for extracting building damage based on multi-time SAR images, including the following parts:

[0093] The regional data module is used to collect relevant data covering the observation area, the observation area including an area where there are building damage, and the relevant data including image, time, terrain and location information;

[0094] The time-series correlation calculation module performs flow processing on the multi-time SAR images, generates backscattering coefficients after the radiation calibration, multi-view processing and registration steps, and further generates time-series correlation data of the observation area; taking the pre-earthquake correlation as a reference image, histogram matching is performed on other correlation images to obtain a matched correlation image set;

[0095] The time series correlation difference or time series coherence difference module carries out normalized difference processing on the time series correlation and coherence images adjacent in time, so as to construct a time series correlation difference or time series coherence difference data set, including a pre-earthquake correlation difference data set, a pre-earthquake coherence difference data set, and a co-seismic correlation difference data set and a co-seismic coherence difference data set.

[0096] The time series correlation difference or time series coherence difference module carries out normalized difference processing on the time series correlation and coherence images adjacent in time, so as to construct a time series correlation difference or time series coherence difference data set, including a pre-earthquake correlation difference data set, a pre-earthquake coherence difference data set, and a co-seismic correlation difference data set and a co-seismic coherence difference data set.

[0097] The multi-temporal image change detection module carries out outlier detection analysis on the time series correlation difference or time series coherence difference data set based on the quartile outlier detection principle, and obtains a building damage distribution map in combination with the building vector data.

[0098] The system for extracting building damage based on multi-temporal SAR images in the embodiment of the application can be a device, a component, an integrated circuit or a chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), and the like, and the non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, and the like, and the embodiment of the application is not limited in this regard.

[0099] The system for extracting building damage based on multi-temporal SAR images in the embodiment of the application can represent a device with an operating system. The operating system can represent an Android operating system, can represent an ios operating system, and can also represent other possible operating systems, and the embodiment of the application is not limited in this regard.

[0100] The system for extracting building damage based on multi-temporal SAR images provided in the embodiment of the application can realize the method for extracting building damage based on multi-temporal SAR images Figures 1 to 7 The method for extracting building damage based on multi-temporal SAR images in the method embodiment can realize each process of the method, and the description will not be repeated here to avoid repetition.

[0101] Optionally, the embodiment of the present application further provides an electronic device, comprising a processor, a memory, a program or instructions stored in the memory and executable in the processor, which realizes each process of the above-mentioned method embodiment for extracting earthquake-damaged buildings based on multi-temporal SAR images and achieves the same technical effects, and the description is omitted here.

[0102] The embodiment of the present application further provides a readable storage medium, which stores a program or instructions, which realizes each process of the above-mentioned method embodiment for extracting earthquake-damaged buildings based on multi-temporal SAR images and achieves the same technical effects, and the description is omitted here.

[0103] The processor represents the processor in the electronic device in the above-mentioned embodiment. The readable storage medium comprises a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disc or an optical disc, etc.

[0104] It should be noted that, in the present application, the terms “comprise”, “contain” or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to the process, method, article or device. Without more limitations, the element defined by the statement “comprising a …” does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present application is not limited to the order of functions performed as shown or discussed, but can also include the functions performed in a substantially simultaneous manner or in the opposite order according to the functions involved, for example, the described method can be performed in an order different from that described, and various steps can also be added, omitted or combined. In addition, the features described with reference to certain examples can be combined in other examples.

[0105] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of contribution to the prior art can be embodied in the form of software product, the computer software product is stored in a storage medium (such as ROM or RAM, magnetic disc, optical disc), including a plurality of instructions to make a terminal (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) executes the method described in various embodiments of the present application.

[0106] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-mentioned specific embodiments, the above-mentioned specific embodiments are only illustrative, but not limited, those skilled in the art can make many forms under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims, all of which belong to the protection of the present application.

Claims

1. A method for extracting earthquake-damaged buildings based on multi-temporal SAR images, characterized in that, The application relates to a method for extracting earthquake-damaged building distribution information from SAR images. The method comprises the following steps: collecting relevant data covering an observation area, wherein the observation area comprises an area where earthquake-damaged buildings exist, and the relevant data comprises images, time, terrain and position information; processing multi-temporal SAR images in a flow, generating backscattering coefficients through radiation calibration, multi-view processing and registration steps, further generating time-series correlation data of the observation area, taking pre-earthquake correlation as a reference image, performing histogram matching on other correlation images to obtain a matched correlation image set; performing differential interference processing on the matched correlation image set by using InSAR technology, optimizing the baseline, performing differential interference processing again to generate time-series coherence data of the observation area, taking pre-earthquake coherence as a reference image, performing histogram matching on other coherence images to obtain a matched coherence image set; performing normalized difference processing on the matched correlation image set and the matched coherence image set which are adjacent in time to construct time-series correlation difference or time-series coherence difference data sets, including pre-earthquake correlation difference, pre-earthquake coherence difference data sets and co-seismic correlation difference, co-seismic coherence difference data sets; 2.The method for extracting earthquake-damaged buildings based on multi-temporal SAR images according to claim 1, characterized in that, performing abnormal value test analysis on the time-series correlation difference and the time-series coherence difference data sets based on the quartile abnormal value test principle, combining with building vector data to obtain an earthquake-damaged building distribution graph. The step of collecting relevant data covering the observation area comprises the following steps: determining the vector range of the disaster area according to the epicenter position; 3.The method of claim 1, wherein, acquiring ascending or descending SAR image sets, digital elevation model data and precise orbit data according to the vector range to collect relevant data covering the observation area. The specific steps of generating time-series correlation data of the observation area comprise the following steps: performing radiation calibration processing on the SAR image set to obtain backscattering coefficients with actual physical significance; performing multi-view processing on the calibrated SAR image set according to a 4:1 multi-view ratio; registering all images in the calibrated SAR image set to obtain an aligned SAR image set, taking a pre-earthquake SAR image in the calibrated SAR image set as a master image, registering all slave images with the master image to obtain registered slave images; performing resampling on each registered slave image to obtain resampled slave images, obtaining an aligned SAR image set, and arranging all images in the aligned SAR image set in time sequence, and establishing a master-slave relationship between adjacent images; calculating the absolute value of the correlation coefficient based on a sliding window according to the master-slave relationship to generate a correlation data set; 4.The method of claim 1, wherein, taking the pre-earthquake correlation as a reference image, performing histogram matching on other correlation images to obtain a matched correlation image set of the observation area. The specific steps of generating time-series coherence data of the observation area comprise the following steps: registering all images in the SAR image set to obtain an aligned SAR image set, taking a pre-earthquake SAR image in the SAR image set as a master image, registering all slave images with the master image to obtain registered slave images; Resampling is performed on each of the registered slave images to obtain resampled slave images, and an aligned SAR image set is obtained, and all images in the aligned SAR image set are arranged in time sequence, and a master-slave relationship is established between adjacent images; SAR images adjacent in time are subjected to differential interferometric processing in a master-slave relationship by using InSAR technology, and terrain and reference surface phases in an initial interferogram are removed to generate a quality control interferogram and preliminary coherence; The quality control interferogram is filtered to obtain a filtered interferogram, and the filtered interferogram is subjected to phase unwrapping processing to obtain an unwrapped interferogram; Baseline optimization is performed on the unwrapped interferogram, a method combining ground control points is used for baseline optimization, and a time series coherence data set is obtained by again performing differential interferometric processing based on a precise baseline; The pre-earthquake coherence is taken as a reference image, and histogram matching is performed on other coherence images to obtain a matched coherence image set of the observation area. 5.The method of claim 1, wherein, The specific process of constructing a time series correlation difference or time series coherence difference data set includes: All images in the matched correlation or matched coherence data set are arranged in time sequence, and a correlation difference or coherence difference data set is obtained by performing normalized difference value operation between adjacent images, and the last image is referred to as a co-seismic correlation difference or co-seismic coherence difference. 6.The method of claim 1, wherein, The specific steps of obtaining the distribution map of earthquake-damaged buildings include: Based on the pre-earthquake time series correlation difference or coherence difference data set, image synthesis is performed to obtain a 25% quantile image Q1 and a 75% quantile image Q3, and Q3 is subtracted from Q1 to obtain a quartile range image IQR, wherein the pre-earthquake time series correlation difference or coherence difference data set does not contain the co-seismic correlation difference or co-seismic coherence difference; Based on the quartile anomaly value test principle, the sum of the image Q3 and the image IQR is regarded as a background threshold image, that is, the upper boundary of the data, and a value exceeding this boundary is regarded as an abnormal value; Pixels of the co-seismic correlation difference or co-seismic coherence difference that exceed the background threshold image and are positive are regarded as potential earthquake-damaged building regions; Based on building vector data, the positions of buildings are located, a threshold is set for the potential earthquake-damaged building regions, and a larger pixel value represents a greater degree of building damage. 7.A system for extracting a quake-damaged building based on multi-temporal SAR images, characterized in that, It includes: A regional data module is used to collect relevant data covering an observation area, the observation area includes a region where earthquake-damaged buildings exist, and the relevant data includes image, time, terrain and location information; A time series correlation calculation module performs process-based processing on multi-temporal SAR images, generates backscatter coefficients after radiation calibration, multi-view processing and registration steps, and further generates time series correlation data of the observation area; the pre-earthquake correlation is taken as a reference image, and histogram matching is performed on other correlation images to obtain a matched correlation image set; A time series coherence calculation module uses InSAR technology to perform differential interferometric processing on registered adjacent SAR data, and performs baseline optimization, and again performs differential interferometric processing to generate time series coherence data of the observation area; the pre-earthquake coherence is taken as a reference image, and histogram matching is performed on other coherence images to obtain a matched coherence image set. The time-series correlation difference or time-series coherence difference module performs normalized difference processing on the correlation and coherence images adjacent in time to construct time-series correlation difference or time-series coherence difference data sets, including pre-seismic correlation difference, pre-seismic coherence difference data sets and co-seismic correlation difference, co-seismic coherence difference data sets; The multi-temporal image change detection module performs outlier test analysis on the time-series correlation difference or time-series coherence difference data sets based on the quartile outlier test principle, and obtains a building damage distribution map in combination with building vector data.

8. An electronic device, comprising: The computer readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the method for extracting building damage based on multi-temporal SAR images according to any one of claims 1-6.

9. A readable storage medium, characterized by, The computer readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the method for extracting building damage based on multi-temporal SAR images according to any one of claims 1-6.

Images