A remote sensing assessment method and system for surface anomaly intensity based on multi-scale features

Abstract

The present application provides a method and system for remote sensing assessment of surface anomaly intensity based on multi-scale features, which belongs to the field of remote sensing data processing technology. A conceptual model of surface anomaly intensity is constructed from four scales: pixel scale, structure scale, object scale, and scene scale; a mathematical model of surface anomaly intensity is constructed based on the conceptual model of surface anomaly intensity; a remote sensing assessment model of surface anomaly intensity is constructed based on the conceptual model and the mathematical model; remote sensing image data is used as basic data to measure each remote sensing feature in the remote sensing feature system of surface anomaly intensity, and the weight of each remote sensing feature is calculated using the entropy weight method, and then the value of the comprehensive surface anomaly intensity of the study area is calculated through the mathematical model of surface anomaly intensity according to the value of each remote sensing feature and its corresponding weight. This scheme can adaptively adjust the feature weights at each scale, improving its ability to be applicable to different types of surface anomaly intensity assessments.

Description

Technical Field

[0001] The present application relates to the technical field of remote sensing data processing, and in particular to a method and system for remote sensing assessment of surface anomaly intensity based on multi-scale features. Background Art

[0002] Surface anomalies are short-term changes in the surface caused by external forces, and various activities that can be detected by remote sensing technology are closely related to the safety of human life and property. The intensity of surface anomalies can be defined as the degree of damage to surface elements caused by surface anomalies. The intensity of disasters is determined by two basic factors: one is the intensity of changes in disaster-causing factors, and the other is the population and economic density of the affected area and the ability to defend and withstand disasters.

[0003] At present, remote sensing assessment of surface anomaly intensity is mostly carried out for specific disaster types. Once the disaster type changes, the assessment model needs to be reselected and reconstructed, which has the problem of insufficient universality. Moreover, from the perspective of remote sensing, there are problems such as unclear sensitivity characteristics, unclear applicability of methods, and insufficient quantitative expression of the degree of change.

[0004] Therefore, it is necessary to provide an improved technical solution to address the above-mentioned deficiencies in the prior art. Summary of the invention

[0005] The purpose of this application is to provide a method and system for remote sensing assessment of surface anomaly intensity based on multi-scale features, so as to solve or alleviate the lack of universality of various types of surface anomaly intensity assessment in the above-mentioned prior art.

[0006] In order to achieve the above objectives, this application provides the following technical solutions:

[0007] In the first aspect, the present application provides a method for remote sensing assessment of surface anomaly intensity based on multi-scale features, comprising:

[0008] Obtain remote sensing image data before and after the occurrence of surface anomalies in the study area;

[0009] Construct a conceptual model of surface anomaly intensity from four scales: pixel scale, structure scale, object scale, and scene scale;

[0010] Constructing a mathematical model of surface anomaly intensity based on the surface anomaly intensity conceptual model;

[0011] Based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model, a surface anomaly intensity remote sensing assessment model is constructed; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and in the surface anomaly intensity remote sensing feature system, each scale includes multiple remote sensing features;

[0012] Among them, the pixel-scale remote sensing features are used to reflect the differences in individual pixel information generated by abnormal surface events; the structural-scale remote sensing features are used to reflect the pixel-to-pixel relationship between the abnormal surface pixels and the surrounding pixels; the object-scale remote sensing features are used to reflect the geometric features of a relatively independent set of pixels affected by abnormal surface events; the scene-scale remote sensing features are used to reflect the degree of abnormal surface state within a preset spatial scale;

[0013] Taking the remote sensing image data as the basic data, each remote sensing feature in the surface anomaly intensity remote sensing feature system is measured, and the weight of each remote sensing feature is calculated using the entropy weight method. Then, based on the value of each remote sensing feature and its corresponding weight, the value of the comprehensive surface anomaly intensity of the study area is calculated through the surface anomaly intensity mathematical model.

[0014] Preferably, the pre-constructed surface anomaly intensity mathematical model is as follows:

[0015]

[0016] Wherein, LSAI is the surface anomaly intensity, PI is the surface anomaly intensity at the pixel scale, SI is the surface anomaly intensity at the structure scale, OI is the surface anomaly intensity at the object scale, and LI is the surface anomaly intensity at the scene scale. PI, SI, OI, and LI are obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method.

[0017] Preferably, the pre-constructed surface anomaly intensity mathematical model is as follows:

[0018] LSAI=PI×w pi +SI×w si +OI×w oi +LI×w li ,

[0019] Where LSAI is the surface anomaly intensity, PI, w pi are the surface anomaly intensity at the pixel scale and its corresponding weight, SI, w si are the surface anomaly intensity and its corresponding weight at the structural scale, OI, w oi are the surface anomaly intensity at the object scale and its corresponding weight, LI, w liThey are the surface anomaly intensity at the scene scale and its corresponding weight; PI, SI, OI, and LI are obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and then performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method.

[0020] Preferably, the pixel-scale remote sensing features include spectral features and scattering features;

[0021] The spectral feature is obtained by calculating the difference between the spectral values ​​of the same pixel before and after the surface anomaly occurs and finding the absolute value;

[0022] The scattering feature is obtained by calculating the difference between the radar backscattering coefficients of the same pixel before and after the surface anomaly occurs and finding the absolute value.

[0023] Preferably, the spectral characteristics include:

[0024] The absolute value of the red light band difference (dRED), the absolute value of the green light band difference (dGREEN), the absolute value of the blue light band difference (dBLUE), the absolute value of the near-infrared band difference (dNIR), the absolute value of the two short-wave infrared band differences (dSWIR1 and dSWIR2) and three remote sensing index differences, including the absolute value of the normalized difference in vegetation index (dNDVI), the absolute value of the burned area index difference (dBAR) and the absolute value of the normalized burn index difference (dNBR).

[0025] Preferably, the scattering characteristics include:

[0026] VV polarization radar backscatter coefficient and VH polarization radar backscatter coefficient.

[0027] Preferably, the remote sensing characteristics of the structural scale are calculated by the following steps:

[0028] Using the remote sensing image data, the gray level co-occurrence matrix before and after the surface anomaly in the study area is calculated;

[0029] Calculate the difference between the characteristic quantity of the gray level co-occurrence matrix before the surface anomaly occurs in the study area and the characteristic quantity of the gray level co-occurrence matrix after the surface anomaly occurs, and take the absolute value to obtain the remote sensing characteristics of the structural scale;

[0030] The characteristic quantities include entropy, energy, contrast and correlation.

[0031] Preferably, the remote sensing features of the object scale include: the side length ratio and area of ​​each object after the surface anomaly occurs.

[0032] Preferably, the remote sensing features of the scene scale include: area ratio and number;

[0033] The area ratio refers to the ratio of the surface anomaly range in each scene to the total area of ​​the scene;

[0034] The number refers to the number of locations or areas where abnormal surface events occur simultaneously in each scene.

[0035] In a second aspect, this embodiment provides a surface anomaly intensity remote sensing assessment system based on multi-scale features, including:

[0036] A data acquisition unit configured to acquire remote sensing image data before and after the occurrence of surface anomalies in the study area;

[0037] A concept construction unit configured to construct a surface anomaly intensity conceptual model from four scales: pixel scale, structure scale, object scale, and scene scale;

[0038] A mathematical construction unit configured to construct a surface anomaly intensity mathematical model based on the surface anomaly intensity conceptual model;

[0039] An indicator construction unit is configured to construct a surface anomaly intensity remote sensing assessment model based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and in the surface anomaly intensity remote sensing feature system, each scale includes multiple remote sensing features;

[0040] Among them, the pixel-scale remote sensing features are used to reflect the differences in individual pixel information generated by abnormal surface events; the structural-scale remote sensing features are used to reflect the pixel-to-pixel relationship between the abnormal surface pixels and the surrounding pixels; the object-scale remote sensing features are used to reflect the geometric features of a relatively independent set of pixels affected by abnormal surface events; the scene-scale remote sensing features are used to reflect the degree of abnormal surface state within a preset spatial scale;

[0041] The comprehensive calculation unit is configured to use the remote sensing image data as basic data, measure each remote sensing feature in the surface anomaly intensity remote sensing feature system, calculate the weight of each remote sensing feature using the entropy weight method, and then calculate the value of the comprehensive surface anomaly intensity of the study area through the surface anomaly intensity mathematical model based on the value of each remote sensing feature and its corresponding weight.

[0042] In a third aspect, this embodiment provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for remote sensing assessment of surface anomaly intensity based on multi-scale features as described in any of the above embodiments is implemented.

[0043] In a fourth aspect, this embodiment provides an electronic device, comprising: a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for remote sensing assessment of surface anomaly intensity based on multi-scale features as described in any of the above embodiments is implemented.

[0044] The technical solution of the embodiment of the present application has the following beneficial effects:

[0045] The embodiment of the present application proposes a more comprehensive and universal surface anomaly intensity assessment method and system, which constructs a surface anomaly intensity conceptual model from four scales: pixel scale, structure scale, object scale and scene scale, and then constructs a mathematical model. Based on the conceptual model and the mathematical model, a surface anomaly intensity remote sensing feature system including multi-scale features at the pixel level, structure level, object level and scene level is constructed. The weights of remote sensing features in different types of surface anomaly intensity assessments are adaptively adjusted by introducing the entropy weight method. At the same time, the weights of surface anomaly intensity at each scale are adaptively adjusted, so that different remote sensing features and surface anomaly intensities at different scales can automatically change their sensitivity in the assessment according to the different types of surface anomalies, thereby achieving the effect of quantitatively, comprehensively, completely and accurately expressing different types of surface anomaly intensities, and providing a new theoretical system and technical framework for quantitative assessment of surface anomaly intensity. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic flow chart of a method for remote sensing assessment of surface anomaly intensity based on multi-scale features provided according to some embodiments of the present application.

[0047] Figure 2 A schematic diagram of a conceptual model of surface anomaly intensity provided according to some embodiments of the present application.

[0048] Figure 3 A schematic diagram of a surface anomaly intensity remote sensing assessment model provided according to some embodiments of the present application.

[0049] Figure 4 A schematic diagram of a remote sensing feature system for assessing forest fire intensity provided according to some embodiments of the present application.

[0050] Figure 5 A schematic diagram of the calculation results of pixel-level features for forest fire intensity assessment provided according to some embodiments of the present application.

[0051] Figure 6 A schematic diagram of the calculation results of structural-level characteristics for forest fire intensity assessment provided according to some embodiments of the present application.

[0052] Figure 7A schematic diagram of the calculation results of object-level features for forest fire intensity assessment provided according to some embodiments of the present application.

[0053] Figure 8 A schematic diagram of the calculation results of scene-level features for forest fire intensity assessment provided according to some embodiments of the present application.

[0054] Fig. 9 A schematic diagram of forest fire intensity assessment results provided according to some embodiments of the present application.

[0055] Fig.10 A schematic diagram of earthquake intensity assessment results provided according to some embodiments of the present application.

[0056] Fig.11 A schematic diagram of flood intensity assessment results provided according to some embodiments of the present application.

[0057] Fig.12 A schematic diagram of the structure of a surface anomaly intensity remote sensing assessment system based on multi-scale features according to some embodiments of the present application. DETAILED DESCRIPTION

[0058] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0059] In the following description, the terms "first / second / third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first / second / third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present disclosure belongs. The terms used herein are only for the purpose of describing the embodiments of the present disclosure and are not intended to limit the present disclosure.

[0061] Method Embodiment

[0062] This embodiment provides a method for remote sensing assessment of surface anomaly intensity based on multi-scale features. The execution subject of this method can be a personal computer, server, tablet computer or other electronic device, such as Figure 1 As shown, the method includes:

[0063] Step S101: Acquire remote sensing image data before and after the surface anomaly in the study area.

[0064] In this embodiment, the types of surface anomalies may be, for example, fires, earthquakes, floods, etc. These disasters will send strong interference to the surface in a short period of time, causing damage to surface elements, and the degree of damage is sufficient to be detected by satellite remote sensing technology. Then, the advantages of satellite remote sensing, such as wide detection range, spatial continuity, rich information, and good dynamics, can be utilized to analyze the changes in remote sensing before and after the surface anomaly from the perspective of satellite remote sensing, thereby expressing the intensity of the surface anomaly in a timely and quantitative manner, and providing a basis for humans to deal with the impact of various surface anomalies (natural disasters, ecological damage, illegal development, etc.) on life and property.

[0065] The study area can be any area on the earth's surface, and the scope of the area can be the area where the disaster actually occurred. It can be obtained by comparing the differences in remote sensing image data before and after the surface anomaly occurs and performing data extraction.

[0066] In this embodiment, the remote sensing image data may be image data of ground information obtained by observing and measuring the ground through unmanned vehicles (such as satellites, airplanes, drones, etc.). For ease of description, the subsequent steps will use satellite remote sensing image data as an example to illustrate the technical solution.

[0067] Step S102: constructing a conceptual model of surface anomaly intensity from four scales: pixel scale, structure scale, object scale and scene scale.

[0068] Among them, the conceptual model is an abstract representation of surface anomalies, which is used to describe the key concepts of systems, things or phenomena of surface anomalies and their relationships. By capturing the core concepts and the interactions between them, it helps people better understand, analyze and design evaluation models for complex surface anomalies.

[0069] In this embodiment, the conceptual model of surface anomaly intensity is constructed from four scales: pixel scale, structure scale, object scale, and scene scale. Among them, pixel is the basic unit of remote sensing image. Pixel scale can reflect the degree of change of physical characteristics such as the color tone of the ground object caused by the abnormal surface event, reflect the difference of individual information, and generate the surface anomaly intensity from the degree of change of radiation brightness based on the pixel. The structure scale expresses the degree of grayscale change between the pixel of the surface anomaly and the surrounding pixels, which can reflect the degree of change of physical characteristics such as the roughness and flatness of the ground object caused by the abnormal surface event, reflect the change of the relationship between pixels, and generate the surface anomaly intensity. The object scale can reflect the information of relatively independent pixel sets, express the state and size of the surface abnormal pixel spots in four or eight neighborhoods, reflect the geometric characteristics such as the fragmentation, shape and size of the impact range of the surface abnormal event, quantitatively evaluate the impact range of the surface abnormal event, and further generate the surface anomaly intensity. The scene scale can express the degree of surface anomaly within a certain spatial scale, express the spatial position, spatial relationship and other information of the research object, and reflect the geometric characteristics of the size and shape of the objects. These geometric characteristics can be quantified into the intensity of surface anomaly, thus realizing the generation of remote sensing early warning knowledge of surface anomaly intensity.

[0070] In short, the remote sensing features at the pixel scale are used to reflect the differences in individual pixel information caused by abnormal surface events; the remote sensing features at the structure scale are used to reflect the pixel-to-pixel relationship between the pixels of the surface anomaly and the surrounding pixels; the remote sensing features at the object scale are used to reflect the geometric characteristics of relatively independent pixel sets affected by abnormal surface events; the remote sensing features at the scene scale are used to reflect the degree of the size of the surface abnormality within the preset spatial scale.

[0071] Among them, the geographic entities of the structural scale of the surface anomaly are pre-acquired, and the geographic entities of the object scale are defined as a set of pixels adjacent to four or eight neighborhoods in the input target space.

[0072] From the time dimension, the definition of surface anomaly intensity is the degree to which the surface anomaly deviates from the normal state, where "normal" means that the surface is in a stable state for a long time. The remote sensing early warning of surface anomaly intensity refers to the analysis of satellite remote sensing spectrum, radiation, scattering, geometry, texture and other remote sensing response characteristic signs / critical points of surface anomalies based on the surface anomaly detection results, the extraction of surface anomaly remote sensing diagnostic characteristic thresholds and comprehensive indexes, and the transformation of surface anomaly satellite images to surface anomaly intensity knowledge. For example, Figure 2As shown in the figure, in the conceptual model of surface anomaly intensity, the remote sensing warning intensity of surface anomaly is evaluated from four scales: pixel scale, structure scale, object scale and scene scale. Each of the four scales contains the evaluation content of its own scale, that is, the remote sensing warning intensity of surface anomaly is evaluated from pixel level, structure level, object level and scene level, with time change and surface change (land cover) as clues and the same position as the spatial framework, to finally obtain the evaluation result of surface anomaly intensity.

[0073] This embodiment constructs a conceptual model to clarify the various components and their interactions in the complex assessment process of different types of surface anomaly intensities from four scales: pixel, structure, object, and scene, laying the foundation for the construction and implementation of the subsequent multi-scale feature assessment model.

[0074] Step S103: constructing a surface anomaly intensity mathematical model based on the surface anomaly intensity conceptual model.

[0075] In this embodiment, the surface anomaly intensity mathematical model is constructed based on the surface anomaly intensity conceptual model, which can be specifically implemented by the following steps: identifying the key elements of the evaluation from the key concepts of the surface anomaly intensity conceptual model, and expressing them as variables of the mathematical model, and establishing corresponding mathematical relationships and equations according to the relationship between the elements in the conceptual model. By expressing the surface anomaly intensity conceptual model as a mathematical model by adopting methods such as algebraic equations, differential equations or logical expressions, the abstract concepts and relationships are converted into quantifiable and computable mathematical expressions, which is conducive to more accurate calibration and optimization of the model.

[0076] Specifically, since the conceptual model of surface anomaly intensity is constructed from four scales: pixel scale, structure scale, object scale and scene scale, the surface anomaly intensity of the four scales is the key concept in this conceptual model and should be expressed as variables of the mathematical model. In addition, there is a spatial position relationship between pixels, structures, objects and scenes, that is, they express the surface anomaly intensity at the same spatial position from four spatial scales. Therefore, their relationship is a comprehensive relationship between various scales, and this synthesis can be linear or nonlinear.

[0077] In some embodiments, the pre-constructed mathematical model of surface anomaly intensity is as follows:

[0078]

[0079] Wherein, LSAI is the surface anomaly intensity, PI is the surface anomaly intensity at the pixel scale, SI is the surface anomaly intensity at the structure scale, OI is the surface anomaly intensity at the object scale, and LI is the surface anomaly intensity at the scene scale. PI, SI, OI, and LI are obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method. It represents the integration between features of different scales. Integration is the combination of features, which can be linear or nonlinear. That is, different scale features are combined into a total surface anomaly intensity in a linear or nonlinear way.

[0080] Among them, the linear or nonlinear mathematical model is constructed by using different scale features. The construction steps of the existing linear model or nonlinear model can be referred to, and this embodiment will not be described in detail here.

[0081] In some other embodiments, the pre-constructed mathematical model of surface anomaly intensity is as follows:

[0082] LSAI=PI×w pi +SI×w si +OI×w oi +LI×w li ,

[0083] Where LSAI is the surface anomaly intensity, PI, w pi are the surface anomaly intensity at the pixel scale and its corresponding weight, SI, w si are the surface anomaly intensity and its corresponding weight at the structural scale, OI, w oi are the surface anomaly intensity at the object scale and its corresponding weight, LI, w li They are the surface anomaly intensity at the scene scale and its corresponding weight; PI, SI, OI, and LI are obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and then performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method.

[0084] The surface anomaly intensity at the pixel scale, structure scale, object scale, and scene scale is integrated by weighted summation, and different weights are assigned to each feature using the weighted summation method. This allows the model to adjust the importance of each feature according to actual needs and comprehensively consider the information of multiple features to provide a more comprehensive and accurate result.

[0085] Step S104: Based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model, a surface anomaly intensity remote sensing assessment model is constructed; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and the surface anomaly intensity remote sensing feature system includes multiple remote sensing features at each scale.

[0086] The core features of different types of surface anomalies are extracted through the conceptual model of surface anomaly intensity, and the remote sensing feature system of surface anomaly intensity is formed according to the scale of the features. The relationship between the core features is then accurately expressed through a mathematical model, and then a remote sensing assessment model of surface anomaly intensity is obtained. The assessment model consists of three parts: an abstract conceptual model and a mathematical model, and a feature system. The feature system includes four scales: pixel, structure, object, and scene. Each scale includes multiple remote sensing features. Each remote sensing feature and the total features of each scale are expressed as variables of the mathematical model. The relationship between the variables is constructed based on the relationship between the core concepts extracted by the conceptual model. It can be a linear relationship or a nonlinear relationship. In this way, an assessment model suitable for different types of disasters is established, and relevant personnel can carry out specific assessments of surface anomaly intensity based on the assessment model.

[0087] Step S105: Using remote sensing image data as basic data, each remote sensing feature in the surface anomaly intensity remote sensing feature system is measured, and the weight of each remote sensing feature is calculated using the entropy weight method. Then, based on the value of each remote sensing feature and its corresponding weight, the value of the comprehensive surface anomaly intensity of the study area is calculated through the surface anomaly intensity mathematical model.

[0088] In step S105, the weight of the indicator can be automatically determined based on the entropy weight method, and the weight of the feature with more information will be higher. Therefore, for the feature with less information, the weight obtained based on the entropy weight method will be lower, so the role played in the assessment of surface anomaly intensity will be relatively reduced, while the feature with more information has a higher weight and plays a leading role in the assessment of surface anomaly intensity, realizing the adaptive selection of features driven by feature distribution.

[0089] In summary, the method provided in this embodiment constructs a conceptual model of surface anomaly intensity from four scales: pixel scale, structure scale, object scale and scene scale, and then constructs a mathematical model. Based on the conceptual model and the mathematical model, a remote sensing feature system of surface anomaly intensity including multi-scale features at the pixel level, structure level, object level and scene level is constructed. The weights of remote sensing features in the assessment of different types of surface anomaly intensity are adaptively adjusted by introducing the entropy weight method. At the same time, the weights of surface anomaly intensity at each scale are adaptively adjusted, so that surface anomaly intensity of different remote sensing features and different scales can automatically change its sensitivity in the assessment according to the different types of surface anomalies, thereby achieving the effect of quantitatively, comprehensively, comprehensively and accurately expressing different types of surface anomaly intensity, and providing a new theoretical system and technical framework for quantitative assessment of surface anomaly intensity.

[0090] In some optional schemes, the remote sensing characteristics at the pixel scale include spectral characteristics and scattering characteristics; the spectral characteristics are obtained by calculating the difference between the spectral values ​​of the same pixel before and after the surface anomaly occurs and finding the absolute value; the scattering characteristics are obtained by calculating the difference between the radar backscattering coefficients of the same pixel before and after the surface anomaly occurs and finding the absolute value.

[0091] In this embodiment, the features used to evaluate the intensity at the pixel scale mainly include spectral features and scattering features. Spectral features are important features of remote sensing images, which are mainly used to describe the statistical characteristics of pixels in the image in each band. Commonly used spectral features include brightness value, standard deviation, etc. The difference in spectral features can reflect the difference in physical properties such as the color tone of objects caused by abnormal surface events, and the greater the difference, the greater the corresponding surface anomaly intensity. Commonly used scattering features include radar backscattering coefficients, etc., which can better reflect the changes in other physical properties such as the roughness of objects caused by abnormal surface events, and the amplitude of feature changes is proportional to the intensity of surface anomalies. The calculation formula for defining the pixel-level intensity PI is as follows:

[0092]

[0093] Where PI represents the pixel-level intensity, L represents the spectral characteristics, and S represents the scattering characteristics.

[0094] In some optional schemes, the spectral characteristics include: the absolute value of the red light band difference (dRED), the absolute value of the green light band difference (dGREEN), the absolute value of the blue light band difference (dBLUE), the absolute value of the near-infrared band difference (dNIR), the absolute value of the two short-wave infrared band differences (dSWIR1, dSWIR2) and three remote sensing index differences, including the absolute value of the normalized vegetation index difference (dNDVI), the absolute value of the burned area index difference (dBAR) and the absolute value of the normalized burn index difference (dNBR).

[0095] In some optional schemes, the scattering characteristics include: VV polarization radar backscattering coefficient and VH polarization radar backscattering coefficient.

[0096] In some optional schemes, the remote sensing characteristics of the structural scale are calculated through the following steps: using remote sensing image data, the gray level co-occurrence matrix before and after the surface anomaly occurs in the study area is calculated; the difference between the characteristic quantity of the gray level co-occurrence matrix before the surface anomaly occurs in the study area and the characteristic quantity of the gray level co-occurrence matrix after the surface anomaly occurs is calculated, and the absolute value is taken to obtain the remote sensing characteristics of the structural scale.

[0097] Among them, the feature quantities include entropy, energy, contrast, and correlation.

[0098] In this embodiment, the features used in the structural scale assessment strength mainly include texture and variance. Texture, i.e., the frequency of the change in the color tone of the ground object, is an important feature reflecting the structural information of the ground object. Variance, i.e., the sum of the squares of the difference between the pixel value and the average value, can represent the degree of discreteness of the ground object pixel value. Texture difference and variance difference can better reflect the changes in other physical properties of the ground object such as roughness and smoothness caused by abnormal surface events, and the amplitude of feature changes can further reflect the strength of the surface anomaly.

[0099] The calculation formula for defining the structural strength SI is as follows:

[0100]

[0101] Where SI is the structural strength, C is the texture, and V is the variance.

[0102] In some optional schemes, the remote sensing characteristics of the object scale include: the side length ratio and area of ​​each object after the surface anomaly occurs.

[0103] In this embodiment, the geometric features of the surface anomaly range are mainly used to evaluate the strength at the object scale. Among them, the fragmentation feature, as an important geometric feature, can better realize the assessment of the strength of the surface anomaly. The boundary fragmentation of the surface anomaly range can be expressed by the side length ratio, and its calculation formula is the ratio of the patch perimeter to the patch area. The larger the fragmentation feature, the more fragmented the corresponding surface anomaly event impact range, and the smaller the surface anomaly intensity. In addition, the object area can also characterize the surface anomaly intensity at the object scale. The larger the object area, the more continuous the impact range of the surface anomaly event and the greater the surface anomaly intensity.

[0104] The calculation formula for object-level intensity OI is defined as follows:

[0105]

[0106] Where OI represents the object-level intensity, E represents the side length ratio, and A is the object area.

[0107] In some options, scene-scale remote sensing features include: area ratio and number; area ratio refers to the proportion of the surface anomaly range in each scene to the total area of ​​the scene; number refers to the number of locations or regions where surface anomaly events occur simultaneously in each scene.

[0108] In this embodiment, scene-level indicators are defined on a pre-designed grid, and the features used for scene-level intensity assessment mainly include area ratio and number. Among them, the area ratio is the proportion of the surface anomaly range within the unit grid area. The larger the proportion of the surface anomaly range area, the longer the surface anomaly event lasts and the wider the impact range, and thus the corresponding surface anomaly intensity is often greater. The number is the number of locations or areas where surface anomaly events occur simultaneously within a unit grid. For example, in a fire, the number can be the number of fire points that occur simultaneously within a unit grid. For another example, in a flood, the number can be the number of independent flooded areas within a unit grid. It should be understood that the more locations / areas where surface anomaly events occur in the same scene, the greater the corresponding surface anomaly intensity.

[0109] The calculation formula for scene-level intensity LI is defined as follows:

[0110]

[0111] Where LI is the scene level intensity, P is the area ratio, and N is the number.

[0112] Figure 3 This is a schematic diagram of the remote sensing assessment model for surface anomaly intensity. Figure 3 As shown, the surface anomaly intensity includes pixel-level intensity, structure-level intensity, object-level intensity and scene-level intensity, among which the pixel-level intensity includes spectral characteristics and scattering characteristics, the structure-level intensity includes texture characteristics and variance, the object-level intensity includes side length ratio and object area, and the scene-level intensity includes area ratio and number.

[0113] The surface anomaly intensity assessment method based on multi-scale features provided in this embodiment constructs a unified surface anomaly intensity assessment framework, analyzes the response relationship between surface changes and remote sensing features, establishes an index system for quantitative assessment of surface anomaly intensity of multi-scale features at the pixel level, structure level, object level and scene level, allocates weights for quantitative assessment of surface anomaly intensity of multi-scale features based on the entropy weight method, establishes a quantitative assessment model for surface anomaly intensity, quantitatively analyzes the remote sensing response characteristics of surface anomaly intensity, and clarifies the intrinsic relationship between surface anomaly intensity and remote sensing response characteristics. This method uses remote sensing to timely assess the intensity of surface anomalies, which is conducive to effective macro-control by decision-making departments, scientifically mobilizes human, material and financial resources, minimizes the damage to the natural environment, casualties and property losses caused by surface anomalies, and provides support for timely rescue of surface anomalies.

[0114] Example 1

[0115] The implementation process of the method provided in this embodiment is explained below by taking forest fire as an example.

[0116] Forest fire is a major natural disaster, characterized by strong suddenness, great destructiveness, high frequency of occurrence, difficulty in prediction and difficulty in extinguishing. In recent years, forest fires have occurred frequently in forested countries around the world, and have become a global problem that endangers social development and the national economy. Effective analysis and quantification of the scale and intensity of forest fires are crucial for forest fire management, emergency response and minimizing the losses caused by forest fires. In recent years, the continuous optimization of space observation methods, sensor technology and computing power, as well as the continuous improvement of spatial resolution, spectral resolution and temporal resolution of remote sensing data, have provided rich remote sensing data support for assessing the intensity of forest fires based on remote sensing images. Remote sensing technology has the characteristics of rich information extraction, wide observation range, fast observation speed and short cycle, and has become an effective means of assessing the intensity of forest fires.

[0117] Since forest fire is a typical type of abnormal surface event, the industry has also conducted some research on it. For example, in the early days, the manual sample plot survey method was mostly used. Although this method has high accuracy, it has a strong subjective selection effect, requires a lot of time and manpower, and needs to be implemented by experienced personnel because it needs to be combined with local actual conditions. The development of remote sensing technology has provided a large amount of reliable data for the quantitative evaluation of forest fire intensity, which can realize synchronous forest fire monitoring within a large scale range and has become an important method for evaluating forest fire intensity. The commonly used method in current research is to establish the correlation between the remote sensing spectral index and the actual forest fire intensity of the ground field survey, and to evaluate the forest fire intensity based on the threshold segmentation method. However, the current forest fire intensity assessment based on remote sensing index mainly focuses on the single remote sensing index threshold method, and lacks research on remote sensing quantitative evaluation of forest fire intensity based on multi-scale features. In addition, the forest fire intensity of the field survey is the basis for model construction, and the difficulty in realizing real-time field survey is also one of the limitations of this type of method.

[0118] In this embodiment, a forest fire that occurred in a certain area of ​​Chongqing from August 21 to 26, 2022 is selected as an example. In this fire accident, due to the dense vegetation in the mountainous area where the fire was caused, the fire spread quickly and lasted for a long time, resulting in serious damage to forest resources in the mountainous area, and the personal and property safety and normal life and production of residents in the surrounding villages were greatly affected.

[0119] In this exemplary application scenario, remote sensing images include image data from Sentinel-1 and Sentinel-2. Remote sensing images collected at two time points, before and during a forest fire, that can cover the study area are selected. The remote sensing image data of the study area are shown in Table 1. Table 1 is as follows:

[0120] Table 1 List of remote sensing image data of the study area

[0121]

[0122] In order to make the surface anomaly range correspond better to the images before and after the surface anomaly, the Sentinel-2 true color image taken on August 24, 2022 is consistent with the remote sensing image used to obtain the surface anomaly range. For this image, the forest fire burnt surface is extracted as the surface anomaly range by fusing remote sensing and spatial features, and the surface anomaly intensity assessment study is carried out within this range.

[0123] In addition, 100 random points were selected in the area to verify the accuracy of the model.

[0124] Before conducting forest fire assessment, the step of preprocessing the above remote sensing image data is also included. Specifically, for Sentinel-2 data, since the downloaded experimental data level is L2A, there is no need for radiation calibration and atmospheric correction, so it is only necessary to unify the image resolution of each band to 10 meters and perform band synthesis, and perform range clipping according to the study area. For Sentinel-1 data, the two images selected in this embodiment are both GRD products of Sentinel-1, which need to be preprocessed by orbit correction, thermal noise removal, radiation calibration, coherent speckle filtering, terrain correction and decibelization. The processed images include two bands, VV and VH. The above preprocessing is used to improve the data quality and the accuracy of forest fire intensity assessment.

[0125] Based on the conceptual model and mathematical model of surface anomaly intensity, we selected indicators at four scales: pixel, structure, object and scene, and constructed a remote sensing feature system for forest fire intensity assessment. We then calculated forest fire intensity based on the remote sensing feature system. The specific steps include:

[0126] 1) Pixel scale characteristics

[0127] Forest fires can cause damage to healthy vegetation, collapse of buildings, etc., causing changes in the physical characteristics of the ground objects. The higher the intensity of the forest fire, the greater the change in the physical characteristics of the ground objects, and the greater the difference reflected in the remote sensing indicators. Therefore, the intensity of the forest fire can be defined as the area where the forest fire occurs and the degree to which each indicator deviates from the normal. Based on the above idea, taking the absolute value of the feature difference before and after the forest fire within the forest fire range as the feature input can better reflect the intensity of the forest fire.

[0128] In this embodiment, the selected pixel-level features mainly include two types of features: spectral features and scattering features. Based on the images before and during the forest fire, the absolute value of the feature difference is calculated.

[0129] a. Spectral characteristics

[0130] The spectral characteristics include the absolute value of the red light band difference (dRED), the absolute value of the green light band difference (dGREEN), the absolute value of the blue light band difference (dBLUE), the absolute value of the near-infrared band difference (dNIR), the absolute value of the two short-wave infrared band differences (dSWIR1 and dSWIR2) and three remote sensing index differences, including the absolute value of the normalized difference in vegetation index (dNDVI), the absolute value of the burned area index difference (dBAR) and the absolute value of the normalized burn index difference (dNBR).

[0131] Among the above features, the red, green and blue bands can better reflect the physical characteristics of the ground object color tone, while the difference between the burned surface and the healthy vegetation in the near-infrared band is large. The selection of the absolute value of the near-infrared band difference can better reflect the difference between the burned surface before and after the disaster. The short-wave infrared band has a certain ability to penetrate smoke. In view of the smoke obstruction problem that may occur before the fire is extinguished in the forest fire assessment, the introduction of the short-wave infrared band as a feature can eliminate this interference to a certain extent. At the same time, the pixel value of the open fire in the short-wave infrared band is significantly higher than that of other ground objects. Therefore, the absolute value of the short-wave infrared band difference in the open fire area before and after the forest fire is large. Using the absolute value of the short-wave infrared band difference as a feature input for forest fire intensity assessment can realize the timely detection and intensity assessment of the open fire. In addition, the vegetation is destroyed after the forest fire, and the NDVI value of the pixel will be significantly reduced. Therefore, applying dNDVI to the forest fire intensity assessment can better reflect the degree of forest damage. BAR uses red light and near-infrared bands to enhance surface information after a fire. The BAR value of the burned surface will increase significantly, which is significantly different from the "normal" state of the surface. The NBR value will decrease significantly after a fire, which is also significantly different from the "normal" state of the surface. Therefore, it is also widely used in forest fire intensity assessment. In addition, there may be smoke obstruction in the early stages of a forest fire. The introduction of the normalized combustion index of the short-wave infrared band can eliminate the impact of smoke to a certain extent.

[0132] b. Scattering characteristics

[0133] The selected scattering features include VV and VH polarimetric radar backscatter coefficients. After a forest fire, the vegetation is destroyed and the image texture changes significantly. The use of VV and VH polarimetric radar backscatter coefficients can better reflect the changes in other physical properties such as surface roughness and flatness caused by surface anomalies, thereby further expressing the intensity of surface anomalies.

[0134] 2) Structural scale characteristics

[0135] At the structural scale, similar to the pixel scale, changes in the physical properties of objects caused by forest fires will cause changes in the texture characteristics of objects, and the greater the intensity of the forest fire, the greater the difference in the texture characteristics of objects. Therefore, the absolute value of the feature difference before and after the forest fire within the forest fire range is used as the feature input for evaluating the intensity of the forest fire. Based on the images before and during the forest fire, the absolute value of the feature difference is calculated.

[0136] The gray-level co-occurrence matrix can reflect the texture characteristics of the image. In this embodiment, the absolute values ​​of four characteristic quantity differences of the gray-level co-occurrence matrix are selected to describe the image texture differences before and after the disaster, including the absolute value of entropy difference (dEntropy), the absolute value of energy difference (dEnergy), the absolute value of contrast difference (dContrast), and the absolute value of correlation difference (dCorrelation).

[0137] Among them, the entropy value reflects the complexity of the grayscale distribution of the image, the energy reflects the coarseness of the image texture, the contrast reflects the depth of the texture grooves, and the correlation is used to measure the similarity of the grayscale levels of the image in the row or column direction.

[0138] In addition to the gray level co-occurrence matrix, the variance can also reflect the surface roughness and flatness. The absolute value of the variance difference can also be used to reflect the intensity of forest fires. The above five features can be used to more comprehensively express the texture characteristics of the image in the study area and reflect the difference in image texture before and after the fire.

[0139] 3) Object scale characteristics

[0140] In this embodiment, two fragmentation indices, side length ratio and patch area, are used to evaluate the intensity of forest fires at the object scale. The higher the fragmentation of the burned surface patches, the more discontinuous the burned surface in the area is, and the corresponding forest fire intensity is lower. The side length ratio is the ratio of the perimeter to the area of ​​the burned surface patches. The larger the side length ratio, the more fragmented the burned surface is, and the smaller the forest fire intensity is. The object area is the area of ​​the burned surface patches. The larger the object area, the larger the impact range of the forest fire is and the lower the fragmentation of the burned surface is, and the corresponding forest fire intensity is higher.

[0141] 4) Scene scale characteristics

[0142] In the forest fire application scenario, due to the single abnormal surface event, the number of scene-level indicators is consistent within the research scope of this application scenario, so only the proportion of burned area is selected as the feature at the scene scale for forest fire intensity assessment.

[0143] Specifically, a 0.5km×0.5km grid was constructed for the study area, and the ratio of the burned area in each grid to the grid area was calculated to obtain the burned area ratio. The larger the burned area ratio in the grid, the larger the area affected by the forest fire and the greater the intensity of the forest fire.

[0144] Figure 4 The remote sensing feature system for forest fire intensity assessment is shown, such as Figure 4 As shown in the figure, the remote sensing index system for forest fire intensity assessment includes pixel-level features, structure-level features, object-level features and scene-level features, and each level (scale) includes multiple specific remote sensing features.

[0145] Based on the above remote sensing feature system, the entropy weight method is used to calculate the weight of each remote sensing feature. The specific steps are as follows:

[0146] Step S1001: To unify the dimensions, all remote sensing features are standardized. If the remote sensing feature value is positive, the following formula is used for standardization:

[0147]

[0148] If the remote sensing feature value is negative, the following formula is used for standardization:

[0149]

[0150] In the formula, r ij represents the value of the jth remote sensing feature of the ith sample after normalization. The sample can be determined by randomly selecting points within the study area. i ' j Represents r ij The corresponding original remote sensing feature value, max means taking the maximum value, and min means taking the minimum value.

[0151] Step S1002: Calculate the characteristic weight after normalization, the formula is as follows:

[0152]

[0153] In the formula, f ij represents the feature weight, and m represents the total number of samples.

[0154] Step S1003: Calculate the characteristic entropy value, the formula is as follows:

[0155]

[0156] In the formula, H j Represents the entropy value of the jth remote sensing feature.

[0157] Step S1004: Calculate feature weights, the formula is as follows:

[0158]

[0159] In the formula, w j represents the weight of the jth feature.

[0160] The weights of the remote sensing features at each scale obtained through the above steps S1001 to S1004 are specifically shown in Table 2. Table 2 is as follows:

[0161] Table 2 Weights of forest fire intensity assessment characteristic system

[0162]

[0163]

[0164] It can be seen from Table 2 that in the assessment of the intensity of surface anomalies caused by forest fires, the weights of structure-level indicators, pixel-level indicators, scene-level indicators, and object-level indicators decrease in sequence.

[0165] Combined with the analysis of actual experience and knowledge, it can be seen that after a forest fire occurs, for pixel-level indicators, the surface color changes from green to dark purple and black before and after the fire, and the color tone of the ground objects changes greatly. The burning of vegetation will also cause changes in remote sensing indices such as NDVI, and the higher the intensity, the more drastic the change. For structural-level indicators, the burning of the forest before and after the fire causes significant changes in texture, and the higher the intensity, the greater the change. The object-level and scene-level indicators are the manifestation of comprehensive information at the grid scale. The extracted burned surface is relatively continuous in space as a whole, with fewer fragmented patches, and the differences in side length ratio, object area and proportion of burned surface are small. Therefore, the weights of object-level and scene-level indicators are lower than those of pixel-level and structural-level indicators.

[0166] On this basis, the remote sensing features at each scale are weighted and integrated to obtain the forest fire intensity values ​​at each scale. The formula is as follows:

[0167] p=∑c j × j ,

[0168] Where p represents the forest fire intensity value at any scale among the four scales of pixel, structure, object, and scene, and c j represents the eigenvalue of the jth remote sensing feature at this scale, w j is the weight of the jth remote sensing feature at this scale.

[0169] Figure 5 , Figure 6 , Figure 7 , Figure 8The calculation results of pixel-level features, structure-level features, object-level features and scene-level features and their corresponding forest fire intensity are shown respectively. The forest fire intensity is divided into levels 1, 2, 3 and 4, and the higher the level, the more serious the fire. Figure 5 Represents the red (DRED) spectral feature, which is a pixel-level feature. Figure 6 Indicates the absolute value of the entropy (dEntropy) value difference, which belongs to the structural level feature. Figure 7 Represents the area of ​​the object, which is an object-level feature. Figure 8 Indicates the proportion of burned area, which is a scene-level feature.

[0170] The obtained forest fire intensity values ​​at the pixel, structure, object, and scene scales are weighted and synthesized using the entropy weight method to generate the final forest fire intensity, that is, to obtain the forest fire intensity assessment result, such as Fig. 9 shown.

[0171] In order to verify the accuracy of the method, this embodiment also includes a step of evaluating the accuracy of forest fire intensity with the help of a confusion matrix. Specifically, based on the Sentinel-2 images of the study area before and during the disaster, 100 verification points are randomly set in the burned area, the intensity level of the verification points is manually determined visually, and the confusion matrix is ​​constructed to statistically classify the accuracy. The calculation formula is as follows:

[0172]

[0173] In the formula, Accuracy is the classification accuracy, T is the number of samples classified correctly, and N is the total number of samples.

[0174] Determine the characteristic information and remote sensing features corresponding to the intensity of forest fires on different burned surfaces, as shown in Table 3:

[0175] Table 3 Information characteristics of different forest fire intensities in the burned areas

[0176]

[0177] Based on Table 3, the actual forest fire intensity levels at 100 verification points were determined, and compared with the quantitative evaluation results of forest fire intensity calculated in the previous steps. The results are as follows:

[0178] Table 4. Results of the accuracy assessment of forest fire intensity classification mapping

[0179]

[0180] The results shown in Table 4 indicate that the method provided in this embodiment constructs a forest fire intensity evaluation index system based on pixel level, structure level, object level and scene level, which can quantitatively express the forest fire intensity, and the evaluation accuracy can reach 75.4%. The overall accuracy of the intensity evaluation is relatively high, proving that the proposed four-level index (four scales) can achieve a comprehensive quantitative expression of forest fire intensity with high accuracy, and is more effective in expressing more serious areas.

[0181] Example 2

[0182] In this embodiment, taking the 2018 Palu earthquake and the 2019 Midwestern United States flood cases as examples, the application effect of the surface anomaly intensity assessment method based on multi-scale features in other types of surface anomaly intensity assessments is illustrated, and the universality of the method in the intensity assessment of various types of surface abnormal events is verified.

[0183] First, the surface anomaly intensity of the 2018 Palu earthquake is evaluated, and its characteristic weights are shown in Table 5. Table 5 is as follows:

[0184] Table 5 Strength evaluation feature weights

[0185]

[0186]

[0187] It can be seen from Table 5 that in earthquake intensity assessment, the weight of structural indicators is relatively higher, while the weight of pixel-level indicators, object-level indicators and scene-level indicators is relatively lower. It can be seen that according to the heterogeneity of different types of data, the entropy weight method can automatically determine the indicator weights and reduce the role of low-information features in intensity expression. This method is universal for all types of surface abnormal events.

[0188] Combined with practical experience, it can be known that earthquakes mainly cause damage or collapse of buildings, damage to surface facilities, etc., and the surface texture characteristics change significantly, so the weight of the structure-level index is relatively higher. However, since the spectral characteristics do not change much before and after the building collapses, and the impact object of the earthquake is a single building or a group of buildings, the difference in the object and scene scale is not significant, so the pixel-level index, object-level index and scene-level index weight are relatively low. In other words, the method provided in the present embodiment, by adjusting the weight, enables the evaluation model to automatically adapt to the changes in the surface anomaly caused by the earthquake, thereby improving the universality of the method.

[0189] At the same time, since the scattering characteristics (radar characteristics) are sensitive to changes in surface roughness, the introduction of scattering characteristics can ensure that the model is applicable to scenarios where the spectral characteristics do not change much but the surface roughness changes greatly, further ensuring its applicability. For example, in the application of earthquake intensity assessment, the weight of spectral characteristics is relatively low at the pixel scale, while the damage to surface facilities such as buildings caused by earthquakes will cause changes in surface roughness, so the difference in radar characteristics is more significant, and it contributes more to the pixel-level surface anomaly intensity assessment.

[0190] The earthquake intensity assessment results obtained based on the characteristic weights are as follows: Fig.10 shown.

[0191] Secondly, the surface anomaly intensity of the 2019 Midwestern flood in the United States was evaluated, and its characteristic weights are shown in Table 6. Table 6 is as follows:

[0192] Table 6 Strength evaluation feature weights

[0193]

[0194] It can be seen from Table 6 that in the assessment of the intensity of surface anomalies caused by floods, the weight of pixel-scale features is relatively high, while the weights of features at other scales are low.

[0195] Combined with the analysis of actual experience and knowledge, it can be seen that flood disasters cause floods to submerge land, increase the turbidity of water bodies, cause changes in the spectral characteristics of water bodies, and the areas that originally showed land spectral characteristics now show water spectral characteristics. The spectral characteristics of ground objects have changed significantly, so the pixel scale characteristics are more sensitive to the assessment of surface anomaly intensity. In addition, because the water body basically maintains homogeneity and the texture characteristics change little, the structural scale characteristics are less different, and the sensitivity to the assessment of surface anomaly intensity is low. For the pixel scale, the spectral feature weight is relatively high, which is related to the fact that floods mainly affect the spectral characteristics of water bodies and their surrounding areas. For the structural scale, the contrast difference weight is relatively high, and the reason is that after the flood submerges part of the land, the depth of the grooves in the image changes. The weight of the object area at the object scale is much higher than the side length ratio, which may be because the flood-affected areas are mostly continuous surface areas spreading outward along the edge of the water body, and the patch shapes are similar, so the side length ratio of the object entity is small, and the area will become the dominant factor affecting the intensity assessment.

[0196] The above analysis also shows that the method provided in this embodiment enables the evaluation model to automatically adapt to changes in surface anomalies caused by floods by adjusting the weights, which verifies the universality of the method.

[0197] The flood intensity assessment results obtained based on feature weights are as follows: Fig.11 shown.

[0198] Based on the above embodiments, the surface anomaly intensity remote sensing assessment method based on multi-scale features proposed in this application can realize quantitative assessment of surface anomaly intensity for different types of surface anomaly events (forest fires, earthquakes, floods). The determination of feature weights based on the entropy weight method is completely dependent on the remote sensing data itself, without the introduction of other empirical knowledge and manual participation. The feature weights determined by the entropy weight method based on mathematical mechanisms are also consistent with empirical knowledge and physical mechanisms. Features that are greatly affected by surface anomalies should play a leading role in their intensity assessment, and these features can often obtain higher weights using the entropy weight method. Unimportant features will be weakened due to their low weights, thereby realizing automated feature selection.

[0199] In summary, the surface anomaly intensity remote sensing assessment method based on multi-scale features proposed in this application is universal for the intensity assessment of various types of surface anomaly events. It can adaptively adjust the weights of remote sensing features in the assessment of different types of surface anomaly intensity and the weights of surface anomaly intensity at each scale, so that the surface anomaly intensity of different remote sensing features and different scales can automatically change its sensitivity in the assessment according to the different types of surface anomalies, thereby realizing the quantitative expression of the intensity of different types of surface anomaly events.

[0200] System Example

[0201] This embodiment provides a surface anomaly intensity remote sensing assessment system based on multi-scale features, which can run on electronic devices such as personal computers, mobile phones, tablets, and edge devices.

[0202] like Fig.12 As shown, the surface anomaly intensity remote sensing assessment system based on multi-scale features includes: a data acquisition unit 1201, a concept construction unit 1202, a mathematical construction unit 1203, an indicator construction unit 1204 and a comprehensive calculation unit 1205. Among them:

[0203] The data acquisition unit 1201 is configured to acquire remote sensing image data before and after the surface anomaly of the study area occurs;

[0204] The concept construction unit 1202 is configured to construct a surface anomaly intensity concept model from four scales: pixel scale, structure scale, object scale and scene scale;

[0205] A mathematical construction unit 1203 is configured to construct a surface anomaly intensity mathematical model based on the surface anomaly intensity conceptual model;

[0206] The indicator construction unit 1204 is configured to construct a surface anomaly intensity remote sensing assessment model based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and in the surface anomaly intensity remote sensing feature system, each scale includes multiple remote sensing features;

[0207] Among them, the remote sensing features at the pixel scale are used to reflect the differences in individual pixel information caused by abnormal surface events; the remote sensing features at the structure scale are used to reflect the pixel-to-pixel relationship between the abnormal surface pixels and the surrounding pixels; the remote sensing features at the object scale are used to reflect the geometric features of relatively independent pixel sets affected by abnormal surface events; the remote sensing features at the scene scale are used to reflect the degree of abnormal surface state within the preset spatial scale;

[0208] The comprehensive calculation unit 1205 is configured to use the remote sensing image data as the basic data, measure each remote sensing feature in the remote sensing feature system of surface anomaly intensity, calculate the weight of each remote sensing feature using the entropy weight method, and then calculate the value of the comprehensive surface anomaly intensity of the study area through the surface anomaly intensity mathematical model based on the value of each remote sensing feature and its corresponding weight.

[0209] The surface anomaly intensity remote sensing assessment system based on multi-scale features provided in this embodiment can implement the process and steps of the surface anomaly intensity remote sensing assessment method based on multi-scale features provided in any of the above embodiments, and achieve the same technical effects, which will not be repeated here one by one.

[0210] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A remote sensing assessment method for surface anomaly intensity based on multi-scale features, characterized in that: include: Obtain remote sensing image data before and after the occurrence of surface anomalies in the study area; Construct a conceptual model of surface anomaly intensity from four scales: pixel scale, structure scale, object scale, and scene scale; Constructing a mathematical model of surface anomaly intensity based on the surface anomaly intensity conceptual model; Based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model, a surface anomaly intensity remote sensing assessment model is constructed; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and in the surface anomaly intensity remote sensing feature system, each scale includes multiple remote sensing features; The pixel-scale remote sensing features are used to reflect the differences in individual pixel information caused by abnormal surface events; the structure-scale remote sensing features are used to reflect the relationship between pixels of abnormal surface pixels and surrounding pixels; the object-scale remote sensing features are used to reflect the geometric features of relatively independent pixel sets affected by abnormal surface events; The remote sensing characteristics of the scene scale are used to reflect the degree of abnormal state of the surface within the preset spatial scale; Taking the remote sensing image data as the basic data, each remote sensing feature in the surface anomaly intensity remote sensing feature system is measured, and the weight of each remote sensing feature is calculated using the entropy weight method. Then, based on the value of each remote sensing feature and its corresponding weight, the value of the comprehensive surface anomaly intensity of the study area is calculated through the surface anomaly intensity mathematical model.

2. The method according to claim 1, characterized in that The mathematical model of surface anomaly intensity is as follows: ， In the formula, LSAI is the surface anomaly intensity, PI is the surface anomaly intensity at the pixel scale, SI is the surface anomaly intensity at the structural scale, OI is the surface anomaly intensity at the object scale, LI is the intensity of surface anomalies at the scene scale; PI, SI, OI, LI It is obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method.

3. The method according to claim 1, characterized in that The mathematical model of surface anomaly intensity is as follows: ， In the formula, LSAI is the surface anomaly intensity, PI, w pi are the surface anomaly intensity at the pixel scale and its corresponding weight, SI, w si are the surface anomaly intensity at the structural scale and its corresponding weight, OI, w oi are the surface anomaly intensity at the object scale and its corresponding weight, LI, w li are the surface anomaly intensity and its corresponding weight at the scene scale, respectively; PI, SI, OI, LI It is obtained by measuring each remote sensing feature in the surface anomaly intensity remote sensing feature system and performing weighted summation at different scales. The weight of each remote sensing feature and the weight of the surface anomaly intensity at each scale are determined by the entropy weight method.

4. The method according to claim 1, characterized in that: The pixel-scale remote sensing characteristics include spectral characteristics and scattering characteristics; The spectral feature is obtained by calculating the difference between the spectral values ​​of the same pixel before and after the surface anomaly occurs and finding the absolute value; The scattering feature is obtained by calculating the difference between the radar backscattering coefficients of the same pixel before and after the surface anomaly occurs and finding the absolute value.

5. The method according to claim 4, characterized in that The spectral characteristics include: The absolute value of the red light band difference, the absolute value of the green light band difference, the absolute value of the blue light band difference, the absolute value of the near-infrared band difference, the absolute value of the two short-wave infrared bands and the three remote sensing index differences, including the absolute value of the normalized vegetation index difference, the absolute value of the burned area index difference and the absolute value of the normalized burn index difference.

6. The method according to claim 4, characterized in that The scattering characteristics include: VV polarization radar backscatter coefficient and VH polarization radar backscatter coefficient.

7. The method according to claim 1, characterized in that The remote sensing characteristics of the structural scale are calculated by the following steps: Using the remote sensing image data, the gray level co-occurrence matrix before and after the surface anomaly in the study area is calculated; Calculate the difference between the characteristic quantity of the gray level co-occurrence matrix before the surface anomaly occurs in the study area and the characteristic quantity of the gray level co-occurrence matrix after the surface anomaly occurs, and take the absolute value to obtain the remote sensing characteristics of the structural scale; The characteristic quantities include entropy, energy, contrast and correlation.

8. The method according to claim 1, characterized in that: The remote sensing features of the object scale include: the side length ratio and area of ​​each object after the surface anomaly occurs.

9. The method according to claim 1, characterized in that: The remote sensing characteristics of the scene scale include: area ratio and number; The area ratio refers to the ratio of the surface anomaly range in each scene to the total area of ​​the scene; The number refers to the number of locations or areas where abnormal surface events occur simultaneously in each scene.

10. A surface anomaly intensity remote sensing assessment system based on multi-scale features, characterized in that: include: A data acquisition unit configured to acquire remote sensing image data before and after the occurrence of surface anomalies in the study area; A concept construction unit configured to construct a conceptual model of surface anomaly intensity from four scales: pixel scale, structure scale, object scale, and scene scale; A mathematical construction unit configured to construct a surface anomaly intensity mathematical model based on the surface anomaly intensity conceptual model; An indicator construction unit is configured to construct a surface anomaly intensity remote sensing assessment model based on the surface anomaly intensity conceptual model and the surface anomaly intensity mathematical model; the surface anomaly intensity remote sensing assessment model includes a surface anomaly intensity remote sensing feature system, and in the surface anomaly intensity remote sensing feature system, each scale includes multiple remote sensing features; Among them, the pixel-scale remote sensing features are used to reflect the differences in individual pixel information generated by abnormal surface events; the structural-scale remote sensing features are used to reflect the pixel-to-pixel relationship between the abnormal surface pixels and the surrounding pixels; the object-scale remote sensing features are used to reflect the geometric features of a relatively independent set of pixels affected by abnormal surface events; the scene-scale remote sensing features are used to reflect the degree of abnormal surface state within a preset spatial scale; The comprehensive calculation unit is configured to use the remote sensing image data as basic data, measure each remote sensing feature in the surface anomaly intensity remote sensing feature system, calculate the weight of each remote sensing feature using the entropy weight method, and then calculate the value of the comprehensive surface anomaly intensity of the study area through the surface anomaly intensity mathematical model based on the value of each remote sensing feature and its corresponding weight.

Images