Quantitative assessment method of earthquake damage to buildings using radar remote sensing based on polarization characteristic parameters

Abstract

The invention relates to the technical field of earthquake damage assessment, in particular to a radar remote sensing building earthquake damage quantitative assessment method based on novel polarization characteristic parameters. The method comprises the following steps: polarization decomposition is performed on pre-processed post-earthquake full polarization synthetic aperture radar data by using a Yamaguchi four-component polarization decomposition method to extract three scattering components; the invention uses a Touzi polarization decomposition method to polarization decompose the pre-processed post-earthquake full polarization synthetic aperture radar data to extract polarization decomposition parameters, calculates polarization characteristic parameters MRMPT, divides the full polarization synthetic aperture radar image elements corresponding to the dominant volume scattering components obtained by the Yamaguchi four-component polarization decomposition into collapsed buildings and intact buildings by using MRMPT, further supplements and identifies the collapsed buildings by using MOSP parameters, selects a suitable threshold for MOSP, calculates a regional comprehensive average earthquake damage index according to a formula, converts and merges the index into a specified earthquake damage level in combination with earthquake damage level assessment requirements, and draws an earthquake damage assessment map for the building.

Description

Technical Field

[0001] The invention relates to the technical field of earthquake damage assessment, and in particular to a method for quantitatively assessing earthquake damage to buildings using radar remote sensing based on polarization characteristic parameters. Background Art

[0002] Building collapse caused by earthquakes often causes a large number of casualties. Post-earthquake building damage assessment is one of the most important contents of earthquake damage assessment and an important basis for earthquake economic loss assessment. Full polarimetric synthetic aperture radar (SAR) data is rich in information and ground imaging is not restricted by weather conditions. Using post-earthquake single-phase full polarimetric SAR data for building damage identification and assessment can achieve both accuracy and speed. Traditional polarimetric decomposition methods are prone to produce damage assessment results that over-exaggerate the extent of building damage. Although, in general, the scattering intensity of collapsed buildings is weaker than that of non-collapsed buildings, the layout of buildings in the earthquake zone is diverse, with intact parallel buildings with strong scattering intensity and intact azimuth buildings with weak scattering intensity. In full polarimetric SAR images, collapsed buildings and azimuth buildings have similar scattering characteristics and are easily confused. To this end, two new polarimetric features, MRMPT and MOSP, are proposed based on two polarimetric decomposition methods, Touzi polarimetric decomposition and Yamaguchi four-component decomposition, to improve the problem of overestimation of building collapse rate. At the same time, the present invention combines the Yamaguchi four-component decomposition method with MRMPT and MOSP to propose a building damage identification scheme based on a single-view full-polarization synthetic aperture radar image after an earthquake. In order to quantitatively evaluate the level of building damage, the building SAR remote sensing damage index DI is proposed based on Touzi polarization decomposition. SAR_PDM and DI SAR , and established the DI SAR The proposed method was verified in an experiment using full polarization synthetic aperture radar data after the Yushu earthquake in China, and obtained 81.02% accuracy in correctly identifying damaged buildings and 78.40% accuracy in correctly identifying undamaged buildings, as well as 73.61% accuracy in directly using the earthquake damage index to assess the earthquake damage level in the building area. Summary of the invention

[0003] The purpose of the present invention is to provide a radar remote sensing building earthquake damage quantitative assessment method based on polarization characteristic parameters, which can not only obtain safe and reliable radar remote sensing building earthquake damage assessment results, but also ensure high and balanced building earthquake damage assessment accuracy, while realizing flexible quantitative characterization of building earthquake damage levels.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters, the method comprising the following steps:

[0005] Step S1, polarization decomposition of the pre-processed post-earthquake full-polarization synthetic aperture radar data is performed using the Yamaguchi four-component polarization decomposition method to extract three scattering components, which are secondary scattering component, volume scattering component and odd scattering component;

[0006] Step S2: Perform Touzi polarization decomposition on the pre-processed post-earthquake full polarization synthetic aperture radar data to extract the polarization decomposition parameter α si and Φ αsi (i=1, 2, 3), and calculate α according to formula (1) s and Φ αs ;

[0007] Step S3: α calculated according to formula (2) and step S2 s and Φ αs , calculate the polarization characteristic parameter MRMPT, and calculate the polarization characteristic parameter MOSP according to formula (2) and the odd-order scattered components produced by polarization decomposition;

[0008] Step S4, selecting a suitable threshold for MRMPT, and dividing the fully polarized synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by the Yamaguchi four-component polarization decomposition into collapsed buildings and azimuth buildings according to the classification rule shown in formula (6);

[0009] Step S5, using MOSP parameters to further supplement the identification of collapsed buildings, selecting a suitable threshold for MOSP, extracting collapsed buildings from the fully polarized synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by Yamaguchi four-component polarization decomposition according to the identification rule of collapsed buildings shown in formula (7), and adding these collapsed building identification results to the collapsed building identification results of MRMPT as the total collapsed building identification results. For the orientation building elements identified by MRMPT, if they are not classified as collapsed buildings by MOSP, they are still classified as orientation building categories;

[0010] Step S6, classifying the to-be-identified ground objects in the full polarization synthetic aperture radar data corresponding to the secondary scattered components as parallel buildings, merging the parallel buildings and azimuth buildings into undamaged buildings, and defining the collapsed buildings as damaged buildings;

[0011] Step S7: α calculated using formula (3) and step S2 s and Calculation of SAR Remote Sensing Damage Index DI for Buildings Based on Touzi Decomposition SAR_PDM , assuming that there are m areas to be assessed in the earthquake zone, the comprehensive average earthquake damage index (DI) of the m areas is calculated according to formula (4) and formula (5): SAR )j , (j = 1, ..., m), according to the obtained m [(DI SAR )1,…,(DI SAR ) j ,…,(DI SAR ) m ] values, draw the building damage level map of m areas in the earthquake zone, and according to the earthquake damage level assessment requirements, transform and merge them into the specified earthquake damage level framework to obtain the building earthquake damage assessment map.

[0012] Preferably, in step S4, formula (6), i.e., the classification rule of MRMPT, can be expressed as:

[0013] x∈volume_dominated buildings

[0014] if MRMPT≤ε1

[0015] x∈collapsed buildings

[0016] if MRMPT>ε1

[0017] x∈oriented standing buildings

[0018] Among them, ε1 represents the threshold for dividing collapsed buildings and oriented buildings using MRMPT parameters.

[0019] Preferably, in step S5, formula (7), i.e., the calculation method of the threshold ε1, is as follows:

[0020]

[0021] Where, mean(·) represents the calculated mean value, MRMPT CB and MRMPT OB They represent the MRMPT value set of collapsed building samples and the MRMPT value set of oriented building samples respectively.

[0022] Preferably, in step S5, the rule of extracting collapsed buildings by MOSP can be expressed as follows:

[0023] x∈volume_dominated buildings

[0024] if MOSP ≥ ε2

[0025] x∈collapsed buildings

[0026] Among them, ε2 represents the threshold value of extracting collapsed buildings using MOSP parameters. The calculation method of the threshold ε2 is as follows:

[0027]

[0028] Where, mean(·) represents the calculated average value, MOSP CB and MOSP OB They represent the MOSP value set of collapsed building samples and the MOSP value set of oriented building samples respectively.

[0029] Preferably, in step S7, the earthquake damage levels of buildings are divided into three levels: slight, moderate and severe damage.

[0030] Preferably, in step S7, the ratio of damaged buildings in each block to all buildings in the block is calculated as the collapse rate of the block, which can be expressed as follows:

[0031]

[0032] if I(i)∈the damaged buildings

[0033] x DB,i =1;

[0034] else x DB,i =0.

[0035] if I(i)∈the undamaged buildings

[0036] x UDB,i =1;

[0037] else x UDB,i =0.

[0038] Among them, CR_classify j represents the collapse rate of the j-th block, and I(i) represents the i-th pixel in the j-th block.

[0039] Preferably, in step S7, DI SAR The value is converted into collapse rate, and the DI of different blocks SAR The linear fitting was performed on the CR value to obtain the following fitting relationship:

[0040] CR_fit = 0.49 × DI SAR +0.27

[0041] Among them, CR_fit represents the collapse rate fitted by the earthquake damage index.

[0042] Preferably, in step S7, the earthquake damage level classification standard is expressed by the formula:

[0043] if collapserate≥0.5

[0044] damage level∈serious damage;

[0045] if 0.3 <collapse rate<0.5

[0046] damage level∈moderate damage;

[0047] if collapserate≤0.3

[0048] damage level∈slight damage.

[0049] Among them, collapse rate represents the collapse rate, which is CR_fit and CR_classify. CR_classify is the collapse rate of the block, and CR_fit represents the collapse rate fitted by the earthquake damage index.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present invention uses a Yamaguchi four-component polarization decomposition method to polarize preprocessed post-earthquake full-polarization synthetic aperture radar data, and performs Touzi polarization decomposition on the preprocessed post-earthquake full-polarization synthetic aperture radar data, extracts polarization decomposition parameters, calculates polarization characteristic parameters, and then divides the full-polarization synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by the Yamaguchi four-component polarization decomposition into collapsed buildings and intact buildings, uses MOSP parameters to further supplement and identify collapsed buildings, selects a suitable threshold for MOSP, calculates the comprehensive average earthquake damage index of each area according to a formula, combines the earthquake damage level assessment requirements, converts and merges them into a specified earthquake damage level framework, and draws a building earthquake damage assessment map. The method has higher precision, recognition rate and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the method of the present invention;

[0053] Figure 2 This is a false color composite image of the full polarization synthetic aperture radar data of Jiegu Town in the urban area of ​​Yushu County, with verification samples of three types of earthquake-affected buildings marked;

[0054] Figure 3 A reference information map for earthquake damage assessment of buildings in Jiegu Town, Yushu County, based on block scale;

[0055] Figure 4 It is the sample value distribution diagram of MRMPT parameters and MOSP parameters of azimuth buildings and collapsed buildings;

[0056] Figure 5 This is a schematic diagram of the extraction results of three types of buildings in the urban area after the earthquake;

[0057] Figure 6 is the comprehensive average earthquake damage index DI of the blocks in Yushu County SAR Continuous distribution graph;

[0058] Figure 7 is the comprehensive average earthquake damage index DI of different blocks SAR Schematic diagram of the fitting relationship with the block collapse rate;

[0059] Figure 8 This is the result of the earthquake damage level assessment at the block level in Yushu urban area based on the earthquake damage level classification of CR_classify;

[0060] Fig. 9 This is the result of the earthquake damage level assessment at the block scale in Yushu urban area based on the earthquake damage level classification of CR_fit. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] A quantitative assessment method for earthquake damage of buildings using radar remote sensing based on polarization characteristic parameters, the method comprising the following steps:

[0063] Step S1, polarization decomposition of the pre-processed post-earthquake full-polarization synthetic aperture radar data is performed using the Yamaguchi four-component polarization decomposition method to extract three scattering components, which are secondary scattering component, volume scattering component and odd scattering component;

[0064] Step S2: Perform Touzi polarization decomposition on the pre-processed post-earthquake full polarization synthetic aperture radar data to extract the polarization decomposition parameter α si and And calculate α according to formula (1) s and Formula (1) is as follows:

[0065]

[0066] Among them, λ i is the eigenvalue of the coherence matrix;

[0067] Step S3: α calculated according to formula (2) and step S2 s and Calculate the polarization characteristic parameter MRMPT, and calculate the polarization characteristic parameter MOSP according to formula (2) and the odd-order scattered components produced by polarization decomposition. Formula (2) is as follows:

[0068]

[0069] MOSP=mean(10×lgI odd )

[0070] Among them, e is the Euler number, lg represents the logarithm with base 10, mean represents the mean calculation, and I odd represents the odd-order scattered power from the Yamaguchi four-component polarization decomposition;

[0071] Step S4, select a suitable threshold for MRMPT, and divide the fully polarized synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by the Yamaguchi four-component polarization decomposition into collapsed buildings and azimuth buildings according to the classification rule shown in formula (6). Formula (6), i.e., the classification rule of MRMPT, can be expressed as:

[0072] x∈volume_dominated buildings

[0073] if MRMPT≤ε1

[0074] x∈collapsed buildings

[0075] if MRMPT>ε1

[0076] x∈oriented standing buildings

[0077] Among them, ε1 represents the threshold for dividing collapsed buildings and oriented buildings using MRMPT parameters, and formula (7), i.e., the calculation method of the threshold ε1, is as follows:

[0078]

[0079] Where, mean(·) represents the calculated mean value, MRMPT CB and MRMPT OB They represent the MRMPT value set of collapsed building samples and the MRMPT value set of oriented building samples respectively;

[0080] Step S5: When using MRMPT to classify collapsed buildings and oriented buildings, collapsed buildings may be missed. Insufficient recognition of collapsed buildings will delay timely rescue of buried personnel. In order to reduce this risk, the correct recognition rate of undamaged buildings should be guaranteed while collapsed buildings should be recognized as much as possible. In this step, MOSP parameters are used to further supplement the recognition of collapsed buildings. A suitable threshold is selected for MOSP. According to the recognition rule of collapsed buildings shown in formula (7), the corresponding volume scattering component is obtained from the Yamaguchi four-component polarization decomposition. The collapsed buildings are extracted from the fully polarized synthetic aperture radar image elements, and the collapsed building identification results are added to the collapsed building identification results of MRMPT as the total collapsed building identification results. That is to say, in the azimuth building results classified by MRMPT, the elements identified as collapsed buildings by MOSP are also classified into the collapsed building category at this time, and are no longer classified as azimuth building categories. For the azimuth building elements identified by MRMPT, if they are not classified as collapsed buildings by MOSP, they are still classified as azimuth building categories. The rules for MOSP to extract collapsed buildings can be expressed by the formula:

[0081] x∈volume_dominated buildings

[0082] if MOSP ≥ ε2

[0083] x∈collapsed buildings

[0084] Among them, ε2 represents the threshold value of extracting collapsed buildings using MOSP parameters. The calculation method of the threshold ε2 is as follows:

[0085]

[0086] Where, mean(·) represents the calculated average value, MOSP CB and MOSP OB They represent the MOSP value set of collapsed building samples and the MOSP value set of oriented building samples respectively;

[0087] Step S6, classifying the to-be-identified ground objects in the full polarization synthetic aperture radar data corresponding to the secondary scattered components as parallel buildings, merging the parallel buildings and azimuth buildings into undamaged buildings, and defining the collapsed buildings as damaged buildings;

[0088] Step S7: α calculated using formula (3) and step S2 s and Calculation of SAR Remote Sensing Damage Index DI for Buildings Based on Touzi Decomposition SAR_PDM, assuming that there are m areas to be assessed in the earthquake zone, the comprehensive average earthquake damage index (DI) of the m areas is calculated according to formula (4) and formula (5): SAR ) j , (j = 1, ..., m), according to the obtained m [(DI SAR )1,…,(DI SAR ) j ,…,(DI SAR ) m ] values, draw the building damage level map of the m areas in the earthquake zone, and transform and merge them into the specified earthquake damage level framework according to the earthquake damage level assessment requirements to obtain the quantitative assessment results of building damage (i.e., building damage assessment map), and the building SAR remote sensing earthquake damage index DI based on Touzi decomposition SAR_PDM , formula (3) is as follows:

[0089]

[0090] The subscript nor[0,1] indicates that DI SAR_PDM The value of is normalized to the interval [0, 1], and formula (4) is as follows:

[0091] The average SAR remote sensing earthquake damage index of buildings with earthquake damage level i is defined as for:

[0092]

[0093] Among them, N PDM represents the total number of building units (the total number of pixels or other scale objects) with earthquake damage level i, n SAR Indicates that the building belongs to the earthquake damage level i. When the kth SAR image unit (pixel or other scale object) belongs to the building of earthquake damage level i, n SAR The value is 1, otherwise it is 0. Formula (5) is the comprehensive average earthquake damage index (DI SAR ) j , and define the index as:

[0094]

[0095] Among them, N L Indicates the total number of earthquake damage levels, A PDM It represents the SAR image area or number of pixels of the building with earthquake damage level i.

[0096] Example:

[0097] The present invention selects the "Ms7.1 Yushu Earthquake" that occurred in Yushu County, Qinghai Province, China on April 14, 2010 as a research example. The epicenter of the earthquake is located at Yushu County is a high-altitude area with a dry and cold climate. The vegetation in the area is sparse and mostly low-lying, which has almost no effect on the extraction of earthquake damage information of buildings in the earthquake zone and can be ignored. Therefore, the classification of vegetation land types was not considered in this experiment.

[0098] The experimental data used in the present invention is a high-resolution airborne fully polarized synthetic aperture radar data of Jiegu Town in Yushu County after the earthquake. Figure 2 In this experiment, the boundary data of Yushu County (i.e., .shp data) was used to mask out the mountains around the urban area, and the earthquake damage identification experiment was only carried out in the urban area. This experiment used the optical remote sensing image of Google Maps as a reference, and selected 25,000 verification samples for collapsed buildings, oriented buildings, and parallel buildings. Figure 2 They are marked in red, blue and green respectively.

[0099] The entire city is divided into 72 blocks. The earthquake damage level assessment results are shown in the figure below. Figure 3 As shown. According to the building collapse rate of each block, the building damage level of the 72 blocks is divided into three levels: slight, moderate and severe damage. The classification standard is: if half or more of the buildings collapsed, the earthquake damage level of the block is severe damage; if less than one-third of the buildings collapsed, the block is slightly damaged; the earthquake damage level between slight damage and severe damage is considered moderate damage.

[0100] According to the building earthquake damage information extraction process shown in Figure 1, the damaged buildings and undamaged buildings in the full polarization synthetic aperture radar data of the experiment of the present invention are extracted. Figure 1 The method flow shown in the figure first performs Yamaguchi four-component polarization decomposition and Touzi polarization decomposition on the experimental data to obtain the secondary scattering component, volume scattering component and odd scattering component of the polarization scattering component and the Touzi polarization decomposition parameter α si and Where i=1, 2, 3. The secondary scattered components are classified as parallel buildings. According to formula (1) and formula (2), the two polarization classification parameters MRMPT and MOSP of the present invention are calculated. According to formula (6), MRMPT is used to classify the pixels dominated by the body scattering component to obtain collapsed buildings and azimuth buildings. Then, according to the rule formula for extracting collapsed buildings by MOSP, MOSP is used to extract collapsed buildings from the pixels dominated by the body scattering component, and the extraction result is incorporated into the collapsed building result obtained by MRMPT classification to form the final collapsed building extraction result. The collapsed building elements identified by MOSP are removed from the azimuth buildings generated by the MRMPT classification as the final azimuth building identification result. The parallel buildings identified by the secondary scattered components are merged with the azimuth building identification results as undamaged buildings, and the identification results of the collapsed buildings are taken as damaged buildings. According to formula (7) and the calculation formula of the threshold ε2, the collapsed building extraction result is adopted. Figure 2 The sample set of collapsed buildings and oriented buildings marked in the figure is used to calculate the classification thresholds ε1 and ε2 of the two polarization classification parameters for collapsed buildings and oriented buildings. The calculated results of ε1 and ε2 are 48 and 44.4 respectively, which is consistent with Figure 4 The difference boundaries of the MRMPT and MOSP scatter plots of collapsed buildings and oriented buildings are basically consistent. Figure 4 In the figure, the MRMPT value of the collapsed building is generally smaller than that of the oriented building, and the watershed between the two types of buildings appears at the calculated MRMPT threshold of 48. Figure 4 In the experiment, the MOSP value of the collapsed building is significantly greater than that of the oriented building, and there is an obvious dividing line between the MOSP values ​​of the two buildings at the calculated threshold of 44.4. Therefore, in this experiment, when the collapsed buildings and oriented buildings are classified using the MRMPT parameters according to formula (6), the classification threshold ε1 is set to 48. Correspondingly, when the collapsed buildings are extracted using the MOSP parameters according to the MOSP rule formula, the identification threshold ε2 is set to 44.4. Figure 4 The sample value distribution diagram of MRMPT parameters and MOSP parameters of azimuth buildings and collapsed buildings is shown in Figure 1. The blue points and red points in the figure are 1000 azimuth building samples and 1000 collapsed building samples, respectively. The extraction results of building damage information in this experimental earthquake case are shown in Figure 2. Figure 5 shown. Figure 5 This is an experimental result map showing only the earthquake damage distribution of buildings in the urban area after masking the mountains around the urban area using the boundary data of the urban area of ​​Yushu County. Figure 2 The verification samples of the three types of buildings marked in the figure are the collapsed building, the azimuth building and the parallel building. Figure 5 The accuracy of the experimental results of building earthquake damage information extraction is evaluated, and the accuracy evaluation confusion matrix of the experimental results is listed in Table 1. Table 1 not only lists Figure 5The extraction results of the three types of buildings in the earthquake zone identified in the experiment are shown in the figure. The collapsed buildings are regarded as damaged buildings, and the azimuthal buildings and parallel buildings are merged into undamaged buildings. The recognition accuracy of damaged buildings and undamaged buildings is evaluated. As can be seen from Table 1, the total number of diagonal samples whose experimental results are consistent with the reference samples accounts for 79.27% ​​of the total samples, that is, the overall accuracy (Overall accuracy) given in the table. In other words, the overall correct recognition accuracy of the three types of buildings in the study area identified by the method proposed in the present invention reaches 79.27%. At the same time, Table 1 also shows the user accuracy (producer's accuracies) of damaged buildings and undamaged buildings. 'User accuracy refers to Figure 2 The probability that the ground truth reference data of this category marked in is correctly classified. From Table 1, we can see that according to the user accuracy of the damaged building is 81.02%, it can be considered that the correct recognition rate of the damaged building is 81.02%; similarly, the user accuracy of the undamaged building is 78.40%, that is, the correct recognition accuracy of the undamaged building is 78.40%.

[0101] Table 1. Confusion matrix of accuracy evaluation of three building recognition results

[0102]

[0103] The building damage is quantitatively assessed at the block scale, and the overall damage level of the block is directly quantified using the damage index value. The comprehensive average damage index (DI) of the 72 blocks in the earthquake zone is calculated according to formulas (3)-(5). SAR ) j , (j=1,…,72). DI of each block SAR The value represents the earthquake damage degree of the block. SAR The level to which the value belongs represents the earthquake damage level of the block. SAR The value range is between [0, 1], and the DI of 72 blocks SAR The continuous distribution of values ​​is as follows Figure 6 As shown. SAR The higher the value, the more serious the damage to the buildings in the block after the earthquake, and the higher the earthquake damage level. In order to evaluate the accuracy of the earthquake damage index proposed in this invention in reflecting the degree of earthquake damage to buildings, the earthquake damage index should be converted into collapse rate and compared with Figure 3 The earthquake damage level assessment results based on collapse rate are compared with the reference figure shown in the figure to evaluate the accuracy of the proposed earthquake damage index. Based on the earthquake zone building identification results of the above experiment, the ratio of damaged buildings in each block to all buildings in the block is calculated as the collapse rate of the block, which can be expressed by the following formula (Formula (10)):

[0104]

[0105] if I(i)∈the damaged buildings

[0106] x DB,i =1;

[0107] else x DB,i =0.

[0108] if I(i)∈the undamaged buildings

[0109] x UDB,i =1;

[0110] else x UDB,i =0.

[0111] Among them, CR_classify j represents the collapse rate of the j-th block, and I(i) represents the i-th pixel in the j-th block.

[0112] In order to DI SAR The value is converted into collapse rate, and the DI of different blocks SAR The linear fitting of the CR value and the CR value was performed to obtain the following fitting relationship (Formula (11)):

[0113] CR_fit = 0.49 × DI SAR +0.27

[0114] Among them, CR_fit represents the collapse rate fitted by the earthquake damage index. For the convenience of presentation, the block collapse rate calculated by formula (10) is represented by CR_classify. The fitting relationship diagram is shown in Figure 7 As shown. Figure 7 It can be seen that the distribution trend of CR_classify represented by the red circle is consistent with that of the fitting line, indicating that the proposed earthquake damage index can accurately reflect the degree of earthquake damage to buildings in the assessment area. Figure 3 The earthquake damage level classification standard shown in the figure is used to classify CR_fit and CR_classify, and to draw the building damage level map of the earthquake zone respectively, forming Figure 8 The earthquake damage level assessment results are shown in Figure 1. The earthquake damage level classification standard can be expressed by formula (12):

[0115] if collapse rate ≥ 0.5

[0116] damage level∈serious damage;

[0117] if 0.3 <collapse rate<0.5

[0118] damage level∈moderate damage;

[0119] if collapse rate≤0.3

[0120] damage level∈slight damage.

[0121] Where collapse rate represents the collapse rate, which can be CR_fit and CR_classify. In order to verify the accuracy of the fitting results and the proposed earthquake damage index, Figure 8 The accuracy of the earthquake damage level assessment results shown in the figure is evaluated, and the accuracy evaluation data is listed in Table 2. From Table 2, it can be seen that the overall accuracy of the earthquake damage level assessment based on CR_classify is 77.78%, and the overall accuracy of the earthquake damage level assessment based on CR_fit is 73.61%. The accuracy of the two earthquake damage assessments is not much different. It can be seen that the proposed earthquake damage index can accurately map the average earthquake damage level of buildings in the assessment area.

[0122] Table 2. Accuracy evaluation of block earthquake damage level assessment results

[0123]

[0124] It can be seen from Table 1 that the correct recognition rate of damaged buildings and the correct recognition rate of undamaged buildings obtained by using the method proposed by the present invention to identify earthquake-affected buildings is 81.02%, and the correct recognition rate of undamaged buildings is 78.40%. The overall correct recognition accuracy of damaged and undamaged buildings is 79.27%. The three recognition accuracies are slightly different, and the correct recognition rate of damaged buildings is guaranteed to be above 80%. Therefore, it is believed that the earthquake damage recognition result of this building is both balanced and safe. Because the missed recognition of collapsed buildings will produce the risk of delayed rescue or missed rescue, and the over-recognition of collapsed buildings will lead to the waste of effective resources and resource allocation errors. Therefore, it is not possible to simply pursue a very high overall recognition accuracy or a high recognition accuracy of a certain type of building, but to pursue a balanced recognition accuracy of damaged and undamaged buildings. In order to avoid the missed recognition of collapsed buildings due to excessive correction of the orientation building by MRMPT, the collapsed buildings considered by MRMPT and MOSP are all identified as collapsed buildings. This makes the recognition accuracy of collapsed buildings slightly higher than the recognition accuracy and overall recognition accuracy of non-collapsed buildings, and such recognition results are safe. The recognition results of this experiment neither overestimated the number of damaged buildings and led to underestimated the number of undamaged buildings, nor underestimated the number of damaged buildings and led to overestimated the number of undamaged buildings. Therefore, it can be said that the building earthquake damage information extraction method proposed by the present invention is safe and reliable.

[0125] From the statistical data in Table 1, it can be seen that 16,034 samples out of the 25,000 orientation building verification samples were correctly identified, with a correct identification rate of 64.14%. In other words, using the method proposed in the present invention, 64.14% of the orientation buildings were successfully separated from the volume scattering components. If these orientation buildings are all mistakenly classified as collapsed buildings, it will mislead a large number of rescue manpower and material resources to be incorrectly allocated. The traditional polarization decomposition method is to regard all the objects corresponding to the volume scattering components as collapsed buildings, and not to further accurately classify the image elements dominated by the volume scattering components. This approach will lead to a very high recognition rate for collapsed buildings and a very low recognition rate for undamaged buildings, resulting in a serious imbalance. At the same time, the recognition results produced in this way to guide emergency rescue will cause a considerable waste of rescue resources.

[0126] The radar remote sensing earthquake damage index proposed in the present invention can quantify the degree of damage to buildings in the earthquake damage assessment area. This quantitative assessment result is the basis for producing earthquake intensity maps using remote sensing means after the earthquake, and earthquake intensity maps are one of the most important requirements for earthquake emergency work. Generally, when using remote sensing technology for quantitative assessment, the building collapse rate is often used to quantitatively assess the degree of damage to buildings in a specific area. However, the collapse rate needs to be calculated based on the identification results of buildings in the earthquake zone. The building SAR remote sensing earthquake damage index DI proposed in the present invention SAR_PDM It can directly reflect the degree of damage to the building. SAR It can directly reflect the average earthquake damage degree of the building area. According to different earthquake damage level classification standards, the value of the earthquake damage index is directly segmented and matched with the earthquake damage level. There is no need to classify the buildings in the earthquake zone and then calculate the collapse rate. It is more intuitive and convenient. Although in this experiment, the earthquake damage level is divided into three levels according to the correspondence between the earthquake damage index and the collapse rate, the proposed earthquake damage index can be promoted and applied according to different needs. First, the earthquake damage index can be directly graded according to different earthquake damage degrees without converting the collapse rate and relying on the correspondence between the earthquake damage index and the collapse rate. Second, the earthquake damage index can be graded according to different rules to obtain the number of different earthquake damage levels and the corresponding earthquake damage degrees.

[0127] As shown in Table 2, 16 of the 72 blocks in Yushu city were misidentified by CR_classify and 19 blocks were misidentified by CR_fit. SARThere will be errors in the process of fitting CR_fit and CR_classify, so the CR_fit error rate calculated by the fitting relationship is slightly higher than that of CR_classify. There are actually 25 severely damaged blocks. CR_classify and CR_fit correctly extracted 21 severely damaged blocks, underestimated the earthquake damage level of 4 blocks, and the earthquake damage level of the underestimated 4 blocks was moderately damaged. There are actually 33 moderately damaged blocks. CR_classify and CR_fit correctly extracted 28 moderately damaged blocks. CR_classify underestimated the earthquake damage level of 1 block and overestimated the earthquake damage level of 4 blocks. CR_fit overestimated 5 blocks and there were no underestimated blocks. It can be seen that the main error of CR_classify and CR_fit in determining the moderate damage level is overestimating the earthquake damage level of the block. Of the actual 14 slightly damaged blocks, CR_classify correctly extracted 7 slightly damaged blocks, and overestimated the earthquake damage level of 7 slightly damaged blocks as moderate damage. However, only 4 slightly damaged blocks were correctly extracted using CR_fit, and the earthquake damage levels of 10 slightly damaged blocks were overestimated as moderate damage. When using CR_classify to identify slightly damaged blocks, the number of blocks with misidentified earthquake damage levels was equal to the number of blocks correctly identified; and when using CR_fit to divide the slightly damaged levels, the number of blocks with misclassified earthquake damage levels was even much larger than the number of blocks with correctly classified earthquake damage levels. It can be seen that when evaluating earthquake damage levels based on CR_classify and CR_fit, the main misestimation exists in the evaluation of the earthquake damage level of slightly damaged damage, and the misestimation of moderate damage and severe damage is far less serious than the misestimation of slight damage. At the same time, observing the elements outside the diagonal of the two precision evaluation matrices in Table 2, because the elements outside the diagonal are the number of blocks with misclassified earthquake damage levels, it can be seen that the total number of blocks in the upper right triangle is greater than the total number of blocks in the lower left triangle. These phenomena all reflect that the earthquake damage index and earthquake damage identification method proposed in the present invention mainly misclassify the earthquake damage level of the block in an overestimation, and underestimate it less. The total number of blocks in the upper right triangle outside the diagonal of the precision evaluation matrix of CR_fit is 4 blocks more than that of CR_classify, which means that the number of blocks overestimated by the earthquake damage assessment based on CR_fit is more than that of CR_classify. This is because in the earthquake damage information extraction method of calculating CR_classify, two polarization classification parameters are used to complementary identify collapsed buildings. This protection mechanism for collapsed buildings makes the number of collapsed buildings identified too many, resulting in a high CR_classify. In addition, when fitting the earthquake damage index DI SARWhen it comes to the relationship with CR_classify, the fitting coefficient is retained to 2 decimal places, and the reason for rounding is that the higher value is taken. Therefore, the earthquake damage level assessment based on CR_fit will estimate the block earthquake damage level higher than the CR_classify assessment, but not much higher, because there are only 4 more blocks. In addition, the earthquake damage level of all blocks whose earthquake damage levels were misestimated was only underestimated or overestimated by 1 level, and was not misestimated by 2 levels. Therefore, in general, the earthquake damage assessment results based on the proposed earthquake damage index are safe and reliable.

[0128] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A quantitative assessment method for earthquake damage to buildings using radar remote sensing based on polarization characteristic parameters, characterized in that: The method comprises the following steps: Step S1, using the Yamaguchi four-component polarization decomposition method to polarize the pre-processed post-earthquake full-polarization synthetic aperture radar (SAR) data to extract three scattering components, which are secondary scattering component, volume scattering component and odd scattering component; Step S2: Perform Touzi polarization decomposition on the pre-processed post-earthquake full polarization synthetic aperture radar data to extract the polarization decomposition parameter α si and And calculate α according to formula (1) s and Formula (1) is as follows: Among them, λ i is the eigenvalue of the coherence matrix; Step S3: α calculated according to formula (2) and step S2 s and Calculate the polarization characteristic parameter MRMPT, and calculate the polarization characteristic parameter MOSP according to formula (2) and the odd-order scattered components produced by polarization decomposition. Formula (2) is as follows: MOSP=mean(10×lgI odd ) Among them, e is the Euler number, lg represents the logarithm with base 10, mean represents the mean calculation, and I odd represents the odd-order scattered power from the Yamaguchi four-component polarization decomposition; Step S4, select a suitable threshold for MRMPT, and divide the fully polarized synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by the Yamaguchi four-component polarization decomposition into collapsed buildings and azimuth buildings according to the classification rule shown in formula (6). Formula (6), i.e., the classification rule of MRMPT, can be expressed as: x∈volume_dominated buildings if MRMPT≤ε1 x∈collapsed buildings if MRMPT>ε1 x∈oriented standing buildings Among them, ε1 represents the threshold for dividing collapsed buildings and oriented buildings using MRMPT parameters; Step S5, further supplement the identification of collapsed buildings using MOSP parameters, select a suitable threshold for MOSP, extract collapsed buildings from the fully polarized synthetic aperture radar image elements corresponding to the dominant volume scattering component obtained by Yamaguchi four-component polarization decomposition according to the identification rule of collapsed buildings shown in formula (7), and add these collapsed building identification results to the collapsed building identification results of MRMPT as the total collapsed building identification results. For the azimuth building elements identified by MRMPT, if they are not classified as collapsed buildings by MOSP, they are still classified as azimuth building categories. Formula (7), i.e., the calculation method of the threshold ε1, is as follows: Where, mean(·) represents the calculated mean value, MRMPT CB and MRMPT OB They represent the MRMPT value set of collapsed building samples and the MRMPT value set of oriented building samples respectively; Step S6, classifying the to-be-identified ground objects in the full polarization synthetic aperture radar data corresponding to the secondary scattered components as parallel buildings, merging the parallel buildings and azimuth buildings into undamaged buildings, and defining the collapsed buildings as damaged buildings; Step S7: α calculated using formula (3) and step S2 s and Φ αs , calculate the building SAR remote sensing damage index DI based on Touzi decomposition SAR_PDM , assuming that there are m areas to be assessed in the earthquake zone, the comprehensive average earthquake damage index (DI) of the m areas is calculated according to formula (4) and formula (5): SAR ) j , j = 1, ..., m, according to the obtained m [(DI SAR )1,…,(DI SAR ) j ,…,(DI SAR ) m ] values, draw the building damage level map of the m areas in the earthquake zone, and transform and merge them into the specified earthquake damage level framework according to the earthquake damage level assessment requirements to obtain the building earthquake damage assessment map. Formula (3) is as follows: The subscript nor[0,1] indicates that DI SAR_PDM The value of is normalized to the interval [0, 1]; Formula (4) is as follows: The average SAR remote sensing earthquake damage index of buildings with earthquake damage level i is defined as for: Among them, N PDM represents the total number of building units with earthquake damage level i, n SAR Indicates the building with earthquake damage level i; Formula (5) is the comprehensive average earthquake damage index (DI) of the jth region. SAR ) j , and define the index as: Among them, N L Indicates the total number of earthquake damage levels, A PDM It represents the SAR image area or number of pixels of the building with earthquake damage level i.

2. The method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters according to claim 1 is characterized in that: In step S5, the rule of extracting collapsed buildings by MOSP can be expressed as follows: x∈volume_dominated buildings if MOSP ≥ ε2 x∈collapsed buildings Among them, ε2 represents the threshold value of extracting collapsed buildings using MOSP parameters. The calculation method of the threshold ε2 is as follows: Where, mean(·) represents the calculated average value, MOSP CB and MOSP OB They represent the MOSP value set of collapsed building samples and the MOSP value set of oriented building samples respectively.

3. The method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters according to claim 1 is characterized in that: In step S7, the earthquake damage levels of buildings are divided into three levels: slight, moderate and severe damage.

4. The method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters according to claim 1 is characterized in that: In step S7, the ratio of damaged buildings in each block to all buildings in the block is calculated as the collapse rate of the block, which can be expressed as follows: if I(i)∈the damaged buildings x DB,i ＝1； else x DB,i =0. if I(i)∈the undamaged buildings x UDB,i ＝1； else x UDB,i =0. Among them, CR_classify j represents the collapse rate of the j-th block, and I(i) represents the i-th pixel in the j-th block.

5. The method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters according to claim 1 is characterized in that: In step S7, DI SAR The value is converted into collapse rate, and the DI of different blocks SAR The linear fitting of the value and the collapse rate (CR) value was performed to obtain the following fitting relationship: CR_fit=0.49×DI SAR +0.27 Among them, CR_fit represents the collapse rate fitted by the earthquake damage index.

6. The method for quantitatively assessing earthquake damage of buildings by radar remote sensing based on polarization characteristic parameters according to claim 5 is characterized in that: In step S7, the earthquake damage level classification standard is expressed by the formula: if collapserate≥0.5 damage level∈serious damage; if 0.3 <collapserate<0.5 damage level∈moderate damage; if collapserate≤0.3 damage level∈slight damage. Among them, collapse rate represents the collapse rate, which is CR_fit and CR_classify. CR_classify is the collapse rate of the block, and CR_fit represents the collapse rate fitted by the earthquake damage index.

Images