A global assessment method for war damage based on multi-source remote sensing data
By integrating multi-source remote sensing data, distinguishing between built-up and non-built-up areas, and establishing a differentiated assessment model, the problem of weak impact of existing war assessment methods in non-built-up areas is solved, achieving a comprehensive and accurate assessment of the impact of war, and supporting war impact monitoring and reconstruction decision-making.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2025-08-29
- Publication Date
- 2026-07-03
AI Technical Summary
Existing war assessment methods show significant effects in urban built-up areas but have a weak impact on unbuilt-up areas, making it difficult to fully cover all war-affected areas. Furthermore, relying solely on NTL data makes it difficult to capture changes in ecology and surface conditions, resulting in insufficient accuracy in assessment results.
By integrating multi-source remote sensing data such as nighttime light data, Sentinel-2 imagery, and land use data, and distinguishing between built-up areas and non-built-up areas, differentiated assessment models were established for each. The degree of urban destruction was measured using NTL data, and indicators such as NDVI, NBR, and LST were integrated to construct a war intensity index for non-built-up areas. Principal component analysis was used to improve the accuracy of the assessment.
It enables the classification and assessment of built-up and unbuilt-up areas, improves the comprehensiveness and accuracy of the spatial representation of war intensity, and can systematically quantify the conflict intensity in different regions, supporting wartime humanitarian relief and post-war reconstruction decisions.
Smart Images
Figure CN121388672B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of remote sensing image processing and analysis and surface change monitoring, especially the field of multi-source remote sensing image processing, and specifically relates to a method and system for comprehensive assessment of war damage based on multi-source remote sensing data. Background Technology
[0002] While traditional war survey and assessment methods can provide ground data, conducting ground surveys in war zones faces challenges such as security risks, transportation obstacles, and poor information flow, leading to difficulties in information collection and long data update cycles, hindering timely and wide-area monitoring. In contrast, satellite observation, due to its advantage of not being limited by ground conditions, is gradually becoming an ideal alternative. Remote sensing technology integrating multi-source satellites can efficiently acquire spatiotemporal data over large areas, tracking the multidimensional impacts of war on society, the economy, and the environment, providing strong support for comprehensive assessment.
[0003] Significant progress has been made in war impact assessment in recent years. Optical remote sensing imagery (such as the Landsat and Sentinel series) has been widely used for immediate damage assessment in conflict zones, revealing information such as infrastructure damage through post-war image changes. With advancements in remote sensing technology, satellite observation data from more sources has been introduced into the field of war assessment to analyze environmental damage, economic decline, and social structural changes caused by war. Among these, nighttime light data (NTL), as an effective indicator of human activity intensity, provides a quantifiable and traceable technical means for war damage assessment, exhibiting unique advantages, especially in macro-scale and long-term monitoring. Existing techniques include using statistical indicators such as the maximum, average, and total values of nighttime light data to measure the degree of urban functional decline caused by war, thus achieving a certain degree of characterization of the intensity of damage to built-up areas. However, existing research largely focuses on immediate damage to built-up areas during wartime, lacking a comprehensive assessment of the long-term impact of war by integrating multi-source remote sensing data. Built-up areas, as densely populated areas with concentrated infrastructure and active economic and social activities, are often the main targets of armed conflicts, exhibiting significant spatial damage characteristics and a decline in nighttime light intensity. Non-built-up areas encompass ecologically sensitive landscapes such as farmland, grassland, and forests. Although less frequently attacked, they often suffer from vegetation degradation, fire damage, and abnormal changes in the surface environment, affecting their ecological functions and regional resilience. Systemic conflict intensity assessments targeting non-built-up areas can help reveal the long-term impacts of war on ecological vulnerability and improve our understanding of the spatially differentiated impacts of war.
[0004] In summary, existing methods have the following limitations: nighttime light data shows a significant effect in urban built-up areas, but has a weak impact on non-built-up areas (such as farmland and forest land), making it difficult for assessment results to fully cover all areas affected by war; using NTL data alone is insufficient to capture changes in ecological and surface conditions, such as land burning, vegetation reduction, or increased surface temperature, resulting in insufficient dimensions for damage assessment; existing studies generally do not distinguish between built-up and non-built-up areas, easily mixing remote sensing changes under different regional characteristics, reducing the accuracy of the assessment.
[0005] This invention innovatively integrates multi-source remote sensing data, including nighttime light data, Sentinel-2 imagery, and land use data, to propose a framework for assessing the extent of war damage across the entire region. It quantifies the intensity and spatial extent of the impact of war on both built-up and unbuilt-up areas, revealing the long-term trends and spatiotemporal evolution characteristics of war impacts. This invention can be widely applied in fields such as war impact monitoring, post-war reconstruction decision support, and humanitarian aid planning, overcoming the shortcomings of existing methods in terms of assessment area coverage, indicator expression dimensions, and spatial classification accuracy. It possesses significant technological innovation and application value. Summary of the Invention
[0006] To address the aforementioned issues, this invention aims to propose a method for assessing the intensity of regional warfare by integrating multi-source remote sensing data. Based on news reports and related literature, there are significant differences in the intensity of warfare between built-up areas and non-built-up areas: built-up areas typically experience sustained high-intensity conflicts, while non-built-up areas mostly experience brief, low-intensity conflicts. Furthermore, built-up areas, dominated by man-made structures, exhibit phenomena such as nighttime light decay, abnormal surface temperature, and multispectral texture faults that directly reflect the degree of building damage; while war damage in non-built-up areas (such as farmland and woodland) is manifested by a sharp drop in vegetation index or abnormal heat radiation, with a relatively weak correlation to human activities. Therefore, this invention divides the study area into built-up and non-built-up areas, establishing differentiated assessment models for each. Through data fusion and algorithm optimization, misjudgments based on natural interference and man-made damage can be avoided, improving monitoring accuracy and decision-making relevance. Specifically, the differentiated assessment method employed is as follows: for built-up areas, a light intensity index is constructed based on NTL data to measure the degree of damage to urban areas; for non-built-up areas, multiple indicators such as NDVI, NBR, LST, and NTL are integrated, and a war intensity index for non-built-up areas is constructed through principal component analysis. This method enables the classification and assessment of built-up and unbuilt-up areas, improving the comprehensiveness, accuracy, and adaptability of the spatial representation of war intensity.
[0007] The technical solution of the present invention is as follows:
[0008] (1) Multi-source data collection and preprocessing: acquire multi-source remote sensing data of the study area, including land use data (LULC), nighttime light data (NTL), normalized vegetation index (NDVI), normalized burn index (NBR), land surface temperature data (LST), etc., and perform preprocessing operations such as spatial alignment and noise removal.
[0009] (2) Built-up area extraction: By integrating LULC and NTL data, the built-up area and non-built-up area ranges are extracted based on the local optimal threshold method;
[0010] (3) Characterization of war intensity in built-up areas: Within the built-up area extracted in step (2), war intensity indicators are constructed using total nighttime light (SNTL) and light ratio index (LIR) to describe the spatiotemporal variation characteristics of war damage within the built-up area;
[0011] (4) Assessment of war intensity in non-built areas: Within the non-built areas extracted in step (2), based on NDVI, NBR, LST and NTL data, principal component analysis (PCA) is used to comprehensively construct the war intensity index in non-built areas, so as to achieve a quantitative assessment of the war intensity in non-built areas;
[0012] (5) Analysis of the war intensity trend in non-built areas: The Sen slope estimation method is used to extract the time series trend of the war intensity index sequence in non-built areas obtained in step (4), and the Mann-Kendall test is used to detect the significance, and the changing trend of war damage and its spatial distribution characteristics are analyzed.
[0013] Specifically, the preprocessing of each auxiliary variable in step (1) is as follows:
[0014] ①NPP-VIIRS Nighttime Lighting Data Preprocessing:
[0015] VIIRS nighttime light data were extracted from the study area to ensure consistent data coverage. For monthly maps with few missing values, a neighborhood imputation method was used to fill in the gaps by interpolating data from adjacent areas to restore data integrity.
[0016]
[0017] The nighttime light data resolution is 500 meters. Indicates grid China and Israel pixels The average value is calculated for the neighborhood of the center (the neighborhood size is 5×5).
[0018] For monthly charts with many missing values, a time-trend imputation method is used. This method involves analyzing the changing trends of the time series data and using data from adjacent time points to fill in the missing values.
[0019]
[0020] in, Value Time Point The missing values that need to be filled by interpolation Is and and Known values at adjacent time points.
[0021] ② Normalized Difference Vegetation Index (NDVI):
[0022]
[0023] in, The data resolution is 10 meters. , These represent the red and near-infrared band information in the Sentinel-2 image, respectively.
[0024] ③ Normalized Burnability Index (NBR):
[0025]
[0026] in, The data resolution is 10 meters. , These represent the near-infrared and short-wave infrared band information in the Sentinel-2 image, respectively.
[0027] ④ Surface temperature (LST):
[0028] The surface temperature dataset is from the VIIRS VNP21A1D product, with a spatial resolution of 1000 meters.
[0029] To ensure the consistency of data spatial resolution, this invention uniformly resamples NDVI, NBR, and LST index data to a spatial resolution of 500 meters, matching the resolution of nighttime light remote sensing data (NTL).
[0030] Specifically, the extraction of built-up areas and non-built-up areas in step (2) is as follows:
[0031] ① Initial binary image acquisition:
[0032] First, all images are uniformly projected to create two LULC built-up area binary maps from different data sources, and the union of the two datasets is obtained. This means that the accumulated pixels in either dataset will be added to the output map.
[0033] ②Mode filtering:
[0034] The merged data contains a large number of isolated and scattered cells, most of which are rural settlements located in suburban and mountainous areas. To identify the war impacts in continuous built-up areas, rather than remote rural regions, a mode filter was applied that replaces the center pixel based on the majority of its neighboring pixels.
[0035] ③ Extraction of light-covered built-up areas:
[0036] The preprocessed light map is transformed into binary maps of the illuminated area with different thresholds, and then raster-fitted with the LULC binary map. The Pearson correlation coefficient (PCC) is used as the correlation index. The light threshold corresponding to the highest Pearson correlation coefficient between the illuminated area binary map and the LULC binary map is the built-up area extraction threshold. The specific formula is as follows:
[0037]
[0038] in, It is a binary image of the lit / unlit pixels in each region; It is the first The radiometric value of each pixel; It is a threshold.
[0039]
[0040] in, It is a binary map of the built-up area / unbuilt-up area of each region; It is the Pearson correlation coefficient between the two binary plots; It is to maximize the correlation coefficient .
[0041]
[0042] For each region, Representing the The original emissivity value of each pixel; Representing the The final nighttime luminous emissivity of each pixel.
[0043] Specifically, the total nighttime light value (SNTL) and light ratio index (LIR) war intensity indicators mentioned in step (3) are as follows:
[0044]
[0045] In the formula, This indicates the total illuminance index of a city at night; This represents the total number of pixels within the city area; Indicates the first The value of a pixel in the established application area and in the non-established area, within the established area. =1, outside the established area (i.e., not in the established area). =0; Indicates the first The brightness value of each pixel.
[0046]
[0047] In the formula, Indicates time Nighttime light ratio index Indicates time Nighttime light intensity values at any given time Indicates time The nighttime light intensity value at a given time, i.e., the nighttime light intensity at the previous time point.
[0048] Specifically, the non-built-up area war intensity index mentioned in step (4) is as follows:
[0049] ① Normalization process:
[0050] The NTL, NDVI, NBR, and LST values processed in step (1) are inconsistent in magnitude. To avoid significant deviations in subsequent calculation results, it is necessary to perform forward normalization to map their values uniformly to the [0,1] interval. This interval is calculated using the following formula:
[0051]
[0052] in, It is the index value after positive normalization. It is an indicator in image elements The value in It is the maximum value of the indicator. It is the minimum value of the indicator.
[0053] ②Data centerization:
[0054] The four normalized indices are combined by band to form a unified data matrix, and each band is centered.
[0055]
[0056] in, It is the first The pixel in the first Values on each band It is the first The average value of each band, This is the data after removing the mean.
[0057] ③ Construct the covariance matrix:
[0058] Next, we calculate the covariance matrix of the data. It represents the correlation between different bands.
[0059]
[0060] in, It is an n×m data matrix (n is the number of pixels, m is the number of bands), and the covariance matrix is an m×m matrix representing the covariance between each pair of bands.
[0061] ④ Eigenvalue decomposition:
[0062] For covariance matrix Perform eigenvalue decomposition. Eigenvectors represent the most important directions in the data, while eigenvalues represent the variance of each principal component. The larger the eigenvalue, the more variance the principal component contains.
[0063]
[0064] in, It is the eigenvector matrix, which contains the eigenvectors of the covariance matrix. It is a diagonal matrix that contains eigenvalues.
[0065] ⑤ Projected onto principal component space
[0066] The centralized data matrix Projecting onto the new principal component space yields the score for each pixel on each principal component.
[0067]
[0068] in, It is an n×k matrix containing the projection values of each pixel onto the first k principal components. It is a matrix containing the first k eigenvectors. In this method, the first principal component (PC1) is selected as the final result of the war intensity index for non-built-up areas because it carries the maximum variance information and can effectively reflect the intensity of war interference expressed by the comprehensive changes of various remote sensing indicators.
[0069] Principal component analysis (PCA) was performed on four remote sensing indices, and the first principal component (PC1) was extracted as the Non-built-up War Intensity Index (NWII). This index is used to quantify the intensity of war damage within a region; the higher the NWII value, the more significant the impact of war on the region. Specifically: a low NDVI (Normalized Difference Vegetation Index) indicates reduced vegetation cover, possibly due to surface damage caused by military activities; a low NBR (Normalized Burn Index) indicates significant burn marks, indicating the presence of high-intensity events such as explosions or fires; a high LST (Land Surface Temperature) reflects increased exposure of bare land and infrastructure, or enhanced anthropogenic heat sources; and an abnormal NTL (Nighttime Light Intensity) may indicate a concentration of unusual human activity or enhanced lighting at military facilities.
[0070] Specifically, the trend analysis described in step (5) is as follows:
[0071] This invention uses the Mann-Kendall test to examine the long-term trend significance of the NWII (War Index) series, assessing the changing trend of the long-term impact of war on the study area. Combined with the Sen slope estimator, the strength and direction of the war impact trend are further quantified, thus providing a scientific basis for the identification and assessment of regional change trends. The Sen slope estimator is a measure of all slopes... The median is calculated using the following formula:
[0072]
[0073] in, This represents all calculated slopes; and These represent the time points in the time series. and The NWII value; Indicates the time interval between two points in time; This represents the median of the Sen slope estimates.
[0074] The Mann-Kendall method was used to test the significance of the trend in the NWII series. The hypothesis test statistic was: The calculation formula is as follows:
[0075]
[0076] Among them, when | Values greater than 1.65, 1.96, and 2.58 indicate that the confidence levels for the significance trend test reach 90%, 95%, and 99%, respectively. This invention primarily uses the 95% confidence level for significance testing.
[0077] Compared with the prior art, the present invention has the following beneficial effects:
[0078] This paper proposes a differentiated war intensity assessment method that takes into account both built-up areas (urban areas) and unbuilt-up areas (suburbs and wilderness), improving the spatial accuracy and regional adaptability of war damage impacts. It utilizes multi-source remote sensing data (including nighttime light, vegetation index, combustion index, and surface temperature) to construct a comprehensive war intensity index for unbuilt-up areas, achieving a systematic and quantitative characterization of conflict intensity in different regions. By visualizing the war intensity index as a spatial distribution map, it intuitively reflects the degree of damage in conflict zones, facilitating scientific decision-making in wartime humanitarian relief and post-war reconstruction. This enriches and improves the technical means of war impact assessment, filling the gaps in existing methods regarding quantitative assessment of unbuilt-up areas and short-term time series analysis. Attached Figure Description
[0079] Figure 1 This is a proposed technical roadmap for assessing global war damage based on multi-source remote sensing data.
[0080] Figure 2 This is a damage intensity test map of the built-up area in a conflict zone.
[0081] Figure 3 This is a damage intensity measurement map of a non-built-up area in a conflict zone. Detailed Implementation
[0082] To better understand the technical solution of this invention, it is now further explained and described in conjunction with the accompanying drawings and specific embodiments. This invention proposes a method for comprehensive assessment of war damage based on multi-source remote sensing imagery. In this example, a war-torn region from January 2022 to March 2025 is used as a typical area. Multi-source remote sensing data of this region is collected, and built-up and non-built-up areas are distinguished according to the proposed technical framework. The overall war intensity is quantitatively assessed, revealing the spatiotemporal characteristics of war damage. The overall process is as follows: Figure 1 As shown. It mainly includes the following steps:
[0083] 1) Multi-source data collection and preprocessing:
[0084] NPP-VIIRS nighttime light data from January 2022 to March 2025 were obtained from NOAA's official website, and missing values were imputed. Sentinel-2 data from January 2022 to March 2025 with cloud cover less than 20% were downloaded from the Google Earth Engine platform. Based on this, the auxiliary variables NDVI, LST, and NBR required for subsequent processing were quantitatively retrieved, and the data were resampled to 500 meters using bilinear interpolation. Land use data from ESRI and ESA in 2021 were downloaded from the Google Earth Engine platform to prepare for subsequent built-up area extraction. The downloaded imagery was cropped using national administrative division vector data of a war-torn region. All geospatial data were processed using the WGS84 coordinate system.
[0085] 2) Extraction from built-up areas:
[0086] ① Initial Binary Map Acquisition. The 2021 ESRI and ESA land use data were transformed into binary maps of built-up and non-built-up areas, and then subjected to union processing.
[0087] ② Mode Filtering. Use the mode filtering tool in ArcGIS to filter the mode of the binary map, removing a large number of isolated and scattered pixels.
[0088] ③ Delineation of built-up areas. Continuously filter the light thresholds until the correlation coefficient between the obtained built-up area with light and the built-up area in the binary map is the largest; this is the final built-up area.
[0089] 3) Intensity of warfare in built-up areas:
[0090] Within the built-up area, the total nighttime light value and light ratio index are used to assess the intensity of war in the built-up area.
[0091] ①Synchronous Light Value (SNTL)
[0092]
[0093] In the formula, This indicates the total illuminance index of a city at night; This represents the total number of pixels within the city area; Indicates the first The value of a pixel in the established application area and in the non-established area, within the established area. =1, outside the established area (i.e., not in the established area). =0; Indicates the first The brightness value of each pixel.
[0094] ② Light ratio index (LIR)
[0095]
[0096] In the formula, Indicates time Nighttime light ratio index Indicates time Nighttime light intensity values at any given time Indicates time The nighttime light intensity value at a given time, i.e., the nighttime light intensity at the previous time point.
[0097] 4) Intensity of war in non-built-up areas:
[0098] Within the non-built-up area, the preprocessed NTL, NDVI, NBR, and LST data were normalized, data-centricated, and a covariance matrix was constructed using Google Earth Engine. Eigenvalue decomposition and projection onto the principal component space were then performed. Finally, the first principal component (PC1) was selected as the final result for the non-built-up area war intensity index. Details are as follows:
[0099] ① Normalization process:
[0100] The NTL, NDVI, NBR, and LST values processed in step (1) are inconsistent in magnitude. To avoid significant deviations in subsequent calculation results, it is necessary to perform forward normalization to map their values uniformly to the [0,1] interval. This interval is calculated using the following formula:
[0101]
[0102] in, It is the index value after positive normalization. It is an indicator in image elements The value in It is the maximum value of the indicator. It is the minimum value of the indicator.
[0103] ②Data centerization:
[0104] The four normalized indices are combined by band to form a unified data matrix, and each band is centered.
[0105]
[0106] in, It is the first The pixel in the first Values on each band It is the first The average value of each band, This is the data after removing the mean.
[0107] ③ Construct the covariance matrix:
[0108] Next, we calculate the covariance matrix of the data. It represents the correlation between different bands.
[0109]
[0110] in, It is an n×m data matrix (n is the number of pixels, m is the number of bands), and the covariance matrix is an m×m matrix representing the covariance between each pair of bands.
[0111] ④ Eigenvalue decomposition:
[0112] For covariance matrix Perform eigenvalue decomposition. Eigenvectors represent the most important directions in the data, while eigenvalues represent the variance of each principal component. The larger the eigenvalue, the more variance the principal component contains.
[0113]
[0114] in, It is the eigenvector matrix, which contains the eigenvectors of the covariance matrix. It is a diagonal matrix that contains eigenvalues.
[0115] ⑤ Projected onto principal component space
[0116] The centralized data matrix Projecting onto the new principal component space yields the score for each pixel on each principal component.
[0117]
[0118] in, It is an n×k matrix containing the projection values of each pixel onto the first k principal components. It is a matrix containing the first k eigenvectors. In this method, the first principal component (PC1) is selected as the final result of the war intensity index for non-built-up areas because it carries the maximum variance information and can effectively reflect the intensity of war interference expressed by the comprehensive changes of various remote sensing indicators.
[0119] Principal component analysis (PCA) was performed on four remote sensing indices, and the first principal component (PC1) was extracted as the Non-Built Area War Intensity Index (NWII). This index is used to quantify the intensity of war damage within a region; the higher the NWII value, the more significant the impact of war on the region. Specifically: a low NDVI (Normalized Difference Vegetation Index) indicates reduced vegetation cover, possibly due to surface damage caused by military activities; a low NBR (Normalized Burn Index) indicates significant burn marks, indicating the presence of high-intensity events such as explosions or fires; a high LST (Land Temperature) reflects increased exposure of bare land and infrastructure, or enhanced anthropogenic heat sources; and an abnormal NTL (Nighttime Light Intensity) may indicate a concentration of abnormal human activity or enhanced lighting at military facilities.
[0120] 5) Analysis of War Intensity Trends in Non-Built-in Areas
[0121]
[0122] in, This represents all calculated slopes; and These represent the time points in the time series. and The NWII value; Indicates the time interval between two points in time; This represents the median of the Sen slope estimates.
[0123] The Mann-Kendall method was used to test the significance of the trend in the NWII series. The hypothesis test statistic was: The calculation formula is as follows:
[0124]
[0125] Among them, when | Values greater than 1.65, 1.96, and 2.58 indicate that the confidence levels for the significance trend test reach 90%, 95%, and 99%, respectively. This invention primarily uses the 95% confidence level for significance testing.
[0126] The final assessment of war damage to built-up and unbuilt-up areas is as follows: Figure 2 and Figure 3 As shown, the war intensity assessed by the method of this invention has a more continuous spatial distribution and clearer boundaries. It can effectively distinguish areas of varying degrees of damage, clearly show the direction of war damage expansion, and demonstrate stronger sensitivity and spatial detail in non-built-up areas, revealing the damage zone on the outskirts of cities.
[0127] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. It will be apparent to those skilled in the art that various modifications can be made to the above embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.
Claims
1. A method for global assessment of war damage based on multi-source remote sensing data, characterized in that, Includes the following steps: (1) Multi-source data collection and preprocessing: acquire multi-source remote sensing data of the study area, including land use data (LULC), NPP-VIIRS nighttime light data (NTL), normalized vegetation index (NDVI), normalized burn index (NBR), and land surface temperature data (LST), and perform preprocessing operations such as coordinate system 1, spatial alignment and noise removal. (2) Built-up area extraction: By integrating LULC and NTL data, the spatial range of the built-up area is extracted based on the local optimum threshold method, and the remaining part is the spatial range of the non-built-up area; the specific method is as follows: 1) Initial binary image acquisition: First, project all images uniformly to create two LULC built-up area binary maps from different data sources, and obtain the union of the two datasets, that is, the accumulated pixels in either dataset will be added to the output map; 2) Mode filtering: The merged data contains a large number of isolated and scattered pixels, most of which are rural settlements in the suburbs and mountains. In order to identify the war impact in continuous built-up areas rather than remote rural areas, a mode filter is applied to replace the center pixel based on the majority of the neighboring pixels. 3) Extraction of light from built-up areas: The preprocessed light map is transformed into a binary map of the lighting area with different thresholds, and then raster-fitted with the LULC binary map. The Pearson correlation coefficient (PCC) is selected as the correlation index. When the Pearson correlation coefficient between the binary map of the lighting area and the LULC binary map is the largest, the corresponding light threshold is the extraction threshold of the built-up area. (3) Characterization of war damage in built-up areas: Within the built-up area extracted in step (2), the total nighttime light value (SNTL) and light ratio index (LIR) are used to develop evaluation indicators of war intensity in built-up areas, and to quantitatively describe the spatiotemporal variation characteristics of war damage in built-up areas; (4) Assessment of War Damage in Non-Built-in Areas: Within the non-built-in areas extracted in step (2), based on NDVI, NBR, LST, and NTL data, principal component analysis (PCA) is used to construct an evaluation index for the war intensity of non-built-in areas, thereby achieving a quantitative assessment of war damage in non-built-in areas; the specific non-built-in area war intensity index is as follows: 1) Normalization process: The NTL, NDVI, NBR, and LST values processed in step (1) are inconsistent in magnitude. Therefore, NTL, NDVI, NBR, and LST are forward normalized to map their values uniformly to the [0,1] interval, which is calculated using the following formula: ， wherein is the normalized index value, is the value of the index in the image element , is the maximum value of the index, is the minimum value of the index; 2) Data centerization: The four normalized indices are combined by wave band to form a unified data matrix, and each wave band is centered. ， in, It is the first The pixel in the first Values on each band It is the first The average value of each band, This is the data after removing the mean; 3) Construct the covariance matrix: Next, we calculate the covariance matrix of the data. It represents the correlation between different bands: ， in, It is an n×m data matrix, where n is the number of pixels and m is the number of bands. The covariance matrix is an m×m matrix that represents the covariance between each pair of bands. 4) Eigenvalue decomposition: For covariance matrix Eigenvalue decomposition is performed, where eigenvectors represent the most important directions in the data, and eigenvalues represent the variance of each principal component. The larger the eigenvalue, the more variance the principal component contains. ， in, It is an eigenvector matrix, containing the eigenvectors of the covariance matrix. It is a diagonal matrix containing eigenvalues; 5) Project onto principal component space The centralized data matrix Projecting onto the new principal component space, we obtain the score of each pixel on each principal component: ， in, It is an n×k matrix containing the projection values of each pixel onto the first k principal components. It is a matrix containing the first k eigenvectors; Principal component analysis (PCA) was performed on four remote sensing indices to extract the first principal component (PC1) as the Non-built-up War Intensity Index (NWII). This index is used to quantify the intensity of war damage within a region; the larger the NWII value, the more significant the impact of war on the region. (5) Spatiotemporal analysis of war damage in non-built areas: The Sen slope estimation method is used to extract the temporal trend of the war intensity index sequence in non-built areas obtained in step (4), and the Mann-Kendall test is used to detect the significance. The spatiotemporal differentiation characteristics of conflicts in non-built areas are analyzed.
2. The method for comprehensive assessment of war damage based on multi-source remote sensing data according to claim 1, characterized in that, The multi-source data preprocessing method in step (1) is as follows: 1) NPP-VIIRS Nighttime Lighting Data The nighttime light dataset comes from the Monthly DNB Composite product synthesized by the Suomi NPP satellite. First, nighttime light data was extracted from the study area at a spatial resolution of 500 meters to ensure consistent data coverage. For monthly imagery with few missing values, a neighborhood imputation method was used, interpolating data from neighboring areas to fill in the gaps and ensure data integrity. ， in, Indicates grid China and Israel pixels The average value is calculated for the neighborhood of the center. For monthly images with a large number of missing values, a time-series imputation method is used. This method analyzes the changing trends of the time series data and uses the valid data from adjacent time points to fill in the missing values. ， in, Value Time Point Missing values that need to be filled by interpolation Is and and Known values at adjacent time points; 2) Normalized Difference Vegetation Index (NDVI) ， in, The data resolution is 10 meters. , These represent the red and near-infrared band information in the Sentinel-2 monthly average composite image, respectively. 3) Normalized Burnability Index (NBR) ， in, The data resolution is 10 meters. , These represent the near-infrared and short-wave infrared band information in the Sentinel-2 monthly average composite image, respectively. 4) Land surface temperature (LST) The surface temperature dataset is from the VIIRS VNP21A1D product, with a resolution of 1000 meters; 5) Spatial matching of multi-source data To ensure consistency in spatial resolution, NDVI, NBR, and LST data were uniformly resampled to a spatial resolution of 500 meters to match the resolution of Night Light Data (NTL).
3. The method for comprehensive assessment of war damage based on multi-source remote sensing data according to claim 1, characterized in that, The total nighttime light intensity (SNTL) and light ratio index (LIR) war intensity indicators mentioned in step (3): ， In the formula, This indicates the total illuminance index of a city at night; This represents the total number of pixels within the city area; Indicates the first The value of a pixel in the established application area and in the non-established area, within the established area. =1, outside of the existing area (i.e., not in the existing area). =0; Indicates the first The brightness value of each pixel; ， In the formula, Indicates time Nighttime light ratio index, Indicates time Nighttime light intensity value at any given time Indicates time The nighttime light intensity value at a given time, i.e., the nighttime light intensity at the previous time point.
4. The method for comprehensive assessment of war damage based on multi-source remote sensing data according to claim 1, characterized in that, The trend analysis in step (5) is as follows: The Mann-Kendall test was used to examine the long-term trend significance of the NWII non-built-up area war index series, assessing the changing trend of the long-term impact of war on the study area. Combined with the Sen slope estimator, the strength and direction of the war impact trend were further quantified, thus providing a scientific basis for the identification and assessment of regional change trends. The Sen slope estimator is the median of all slopes Q, and the specific formula is as follows: ， in, This represents the calculated slope; and These represent the time points in the time series. and The NWII value; Indicates the time interval between two points in time; This represents the median of the Sen slope estimates; The Mann-Kendall method was used to test the significance of the trend in the NWII series. The hypothesis test statistic was: The calculation formula is as follows: ， Among them, when Values greater than 1.65, 1.96, and 2.58 indicate that the confidence levels for the significance trend test reach 90%, 95%, and 99%, respectively.