Geological disaster early warning method based on remote sensing monitoring

Abstract

The invention discloses a geological disaster early warning method based on remote sensing monitoring, and relates to the technical field of remote sensing monitoring, and the method comprises the following steps: obtaining a multi-temporal remote sensing image of a target monitoring area, constructing a standard deviation deformation field, and obtaining the standard deviation deformation field; performing regional sliding window analysis according to the standard deviation deformation field to obtain a disturbance image; performing boundary extraction according to the disturbance image to obtain a regional mask pattern; performing time sequence fitting according to the regional mask pattern to obtain a displacement rate image; performing stability analysis according to the displacement rate image to obtain a critical coefficient graph; and performing spatial clustering according to the critical coefficient graph to obtain a disaster early warning unit graph. According to the method, the response of the continuous boundary region is effectively enhanced, the influence of isolated noise points is reduced, the false alarm rate is reduced, the noise immunity is improved, the disturbance cluster and the background weak disturbance region can be accurately separated, the region recognition capability is improved, and the adaptability is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing monitoring, and in particular to a geological disaster early warning method based on remote sensing monitoring. Background Art

[0002] Geological disasters (such as landslides, collapses, and ground subsidence) are characterized by suddenness, concealment, and wide-ranging impacts. Especially in mountainous areas, reservoir slopes, and mined-out areas, their mechanisms are complex and difficult to predict. With the rapid development of remote sensing technology, geological disaster early warning methods based on remote sensing monitoring have become a research hotspot and a key focus of engineering deployment. Leveraging the high temporal and spatial resolution of remote sensing imagery, they enable large-scale, contactless, and periodic monitoring, providing an important supplement and extension to traditional on-site monitoring methods.

[0003] In the existing technology, there are deficiencies in deformation perturbation image extraction: most existing deformation perturbation image extraction methods use global statistical methods (such as overall variance, mean change, etc.) for analysis, have weak local feature recognition capabilities, ignore the spatial coherence and aggregation of deformation perturbations, and the perturbation enhancement capability relies on artificial parameters or low-order filters. As a result, they are easily affected by texture background and ground object changes, resulting in a high false alarm rate. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a geological disaster early warning method based on remote sensing monitoring to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a geological disaster early warning method based on remote sensing monitoring, comprising the following steps: S1. Acquire multi-temporal remote sensing images of the target monitoring area and construct the standard deviation deformation field to obtain the standard deviation deformation field; S2. Perform regional sliding window analysis based on the standard deviation deformation field to obtain a disturbance image; S3, extracting the boundary based on the disturbed image to obtain a regional mask image; S4. performing time series fitting according to the regional mask image to obtain a displacement rate image; S5. Perform stability analysis based on the displacement rate image to obtain a critical coefficient graph; S6. Perform spatial clustering based on the critical coefficient map to obtain a disaster warning unit map.

[0006] To further optimize this technical solution, the regional sliding window analysis in S2 includes: According to the obtained standard deviation deformation field, a sliding window area is defined, and the disturbance response model is used to convert the standard deviation deformation field into a disturbance image reflecting the degree of deformation anomaly aggregation.

[0007] To further optimize this technical solution, the sliding window area includes: ; ; in: : Sliding window side length; : sliding window area; : The radius of the sliding window area; : A pixel in the sliding window area centered on pixel (i, j).

[0008] To further optimize this technical solution, the disturbance response model includes: ; in: : The disturbance response value at pixel (i, j); : The structural difference between the sliding window centered on pixel point (i, j) and the eight neighboring windows; : The perturbation density of the sliding window centered at pixel (i, j).

[0009] To further optimize this technical solution, the structural differences include: ; in: : The perturbation density of the sliding window centered at pixel (m,n); : The center point coordinate set of 8 neighborhood windows centered on pixel point (i, j).

[0010] Further optimizing this technical solution, the disturbance density includes: ; ; in: : standard deviation of the shape variable at the pixel point (x, y); : The mean of the standard deviation of the deformation variable within the sliding window centered at pixel point (i, j).

[0011] To further optimize this technical solution, the time series fitting in S4 includes: According to the obtained regional mask map, the weighted perturbation trend fitting model is used to extract the surface displacement time series corresponding to each pixel in the target area in the multi-temporal remote sensing image, thereby calculating the average surface displacement rate per unit time and obtaining the displacement rate image.

[0012] To further optimize this technical solution, the weighted disturbance trend fitting model includes: ; in: : the long-term trend displacement rate of the target pixel; : represents the time index of remote sensing images; : time mean; : total time series length; : The remote sensing displacement value of the target pixel (i, j) at the tth moment; : the mean displacement of the target pixel; : Perturbation weight factor.

[0013] Further optimizing this technical solution, the disturbance weight factor includes: ; in: : The remote sensing displacement value of the target pixel (i, j) at the t-1th moment; : Perturbation amplification factor.

[0014] Further optimizing this technical solution, the displacement mean and time mean include: ; ; The displacement mean and time mean are used to center the series so that the fitting results do not depend on the initial position and time starting point.

[0015] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of a geological disaster early warning method based on remote sensing monitoring as described in the first aspect of the present invention are implemented.

[0016] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored, wherein: when the computer program instructions are executed by a processor, the steps of a geological disaster early warning method based on remote sensing monitoring as described in the first aspect of the present invention are implemented.

[0017] Compared with the existing technology, the present invention provides a geological disaster early warning method based on remote sensing monitoring, which has the following beneficial effects: This geological disaster early warning method based on remote sensing monitoring uses a disturbance response model and a sliding window analysis to effectively enhance the response of continuous boundary areas, reduce the impact of isolated noise points, reduce the false alarm rate, improve noise resistance, and accurately separate disturbance clusters from background weak disturbance areas. It is particularly suitable for identifying hidden landslides and potential active slopes, improving regional recognition capabilities and enhancing adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A schematic diagram of the process of a geological disaster early warning method based on remote sensing monitoring proposed by the present invention; Figure 2 This is a flow chart of a disturbance response model of a geological disaster early warning method based on remote sensing monitoring proposed by the present invention; Figure 3 A flow chart of a weighted disturbance trend fitting model for a geological disaster early warning method based on remote sensing monitoring proposed by the present invention; Figure 4 This is a flow chart of a stability critical analysis model for a geological disaster early warning method based on remote sensing monitoring proposed in the present invention. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1: Reference Figures 1 to 4 , which is the first embodiment of the present invention, provides a geological disaster early warning method based on remote sensing monitoring, comprising the following steps: S1. Acquire multi-temporal remote sensing images of the target monitoring area and construct the standard deviation deformation field to obtain the standard deviation deformation field.

[0024] In this embodiment, the acquisition of multi-temporal remote sensing images of the target monitoring area and the construction of the standard deviation deformation field include: Geological disasters such as landslides and ground subsidence are essentially temporal deformations of the earth's surface. These deformations may accumulate slowly over long periods of time before a disaster occurs. Using only the surface state at a single moment (such as a single remote sensing image) cannot effectively identify potential risk areas. Therefore, multi-temporal remote sensing observations are used to continuously monitor the changing trends of surface movement, thereby capturing subtle precursors to disasters. Time series data formed by superimposing multiple image periods can identify the deformation trajectories and spatial distribution characteristics of slow-changing and latent disasters.

[0025] The core purpose of this step is to establish a foundational raw deformation data field for subsequent deformation disturbance analysis, disaster identification, and early warning modeling. By calculating the standard deviation of the deformation time series, a standard deviation deformation field is constructed, which serves as a key input for identifying unstable areas of geological activity. The standard deviation reflects the degree of deformation volatility over time and is significant for distinguishing stable areas (low volatility) from areas of anomalous activity (high volatility). Acquiring multi-temporal remote sensing imagery and constructing a standard deviation deformation field provide a foundation for subsequent trend identification and change detection. This reveals the characteristics of temporal volatility and serves as a key indicator of geological stability. It helps identify areas of strong local disturbance and uneven deformation, facilitating the delineation of high-risk areas.

[0026] Remote sensing data selection and acquisition typically uses Sentinel-1A C-band SAR imagery or other spaceborne SAR data with known orbital parameters. Multi-temporal remote sensing image datasets covering the monitoring area are selected, with a time span of at least 60 days and at least ten image periods. After acquiring the multi-temporal remote sensing image datasets, data preprocessing is performed, including image registration (master image alignment of all images to ensure pixel alignment), interferometry (using SBAS-InSAR (Small Baseline Subset) technology to construct small baseline interferogram pairs and obtain differential interferograms), atmospheric phase correction, and terrain phase removal (using DEM auxiliary data to eliminate terrain effects and using meteorological models or temporal differencing methods to mitigate atmospheric delay). After preprocessing, time series reconstruction and deformation calculation are performed. For each pixel, the differential phase is inverted using the least squares method to reconstruct the deformation time series, and the standard deviation of the deformation is calculated over the time series. Finally, the standard deviation values of the deformation variables of all pixels are combined into a two-dimensional image, thereby obtaining a deformation basic data set with temporal continuity, spatial correspondence and statistical discrimination ability, namely the standard deviation deformation field.

[0027] S2. Perform regional sliding window analysis based on the standard deviation deformation field to obtain a disturbance image.

[0028] In this embodiment, the regional sliding window analysis includes: In step S1, the standard deviation deformation field is obtained from multi-temporal remote sensing image data. This image reflects the degree of deformation fluctuation of the surface during the monitoring period. However, the standard deviation image can only represent the statistical amplitude of the deformation variable in a local area, and its effectiveness in identifying potential disaster areas with sparse distribution or blurred boundaries is limited. For example, the boundary of a slowly developing landslide body often has a blurred transition with the background area and lacks clear mutation characteristics. Therefore, it is necessary to introduce spatial hierarchical analysis methods to further enhance the disturbance density response within the local area. The sliding window mechanism is used to capture the relative abnormal deformation aggregation at the regional scale, thereby forming a high-resolution, high-responsive deformation disturbance image to provide data support for subsequent boundary identification.

[0029] In this step, a sliding window area is defined based on the obtained standard deviation deformation field, and the disturbance response model is used to convert the standard deviation deformation field into a disturbance image that reflects the degree of deformation anomaly aggregation, thereby improving the contrast between the local geological disturbance area and the background area, enhancing the continuity and clarity of the regional boundary in the image, forming a concentrated response in areas with inconspicuous deformation but complex internal disturbances (such as landslide incubation areas), and segmenting and identifying areas with abnormal deformation to facilitate subsequent processing.

[0030] Furthermore, the sliding window area includes: ; ; in: : Sliding window side length, usually takes an odd value, recommended range is 5 to 15, set according to image resolution and terrain complexity; : The sliding window area is centered at pixel (i, j), with a size of W×W, and contains all pixels to be analyzed in the area; : The radius of the sliding window area; : A pixel in the sliding window area centered on pixel (i, j).

[0031] Furthermore, the disturbance response model includes: ; in: : The disturbance response value at pixel point (i, j). A larger value indicates that the disturbance density in the area is high and significantly different from the neighboring areas. It has a high risk potential and is suitable as a candidate area for potential landslides or other geological hazards. : The structural difference between the sliding window centered on pixel point (i, j) and the eight neighboring windows reflects whether the current window maintains perturbation consistency with the neighborhood. Each sliding window has the same size. : The perturbation density of the sliding window centered at pixel point (i, j), reflecting the degree of aggregation of local deformation fluctuations.

[0032] Furthermore, the structural difference includes: ; in: : The perturbation density of the sliding window centered at pixel point (m,n), reflecting the degree of aggregation of local deformation fluctuations; : The coordinate set of the center points of the 8 neighborhood windows centered on the pixel point (i, j). The eight directions include left, right, up, down, upper left, upper right, lower left, and lower right.

[0033] Furthermore, the disturbance density includes: ; ; in: : standard deviation of the shape variable at the pixel point (x, y); : The mean of the standard deviation of the deformation variable within the sliding window centered at pixel point (i, j).

[0034] This model describes how to perform regional sliding window analysis based on the standard deviation deformation field to segment and identify abnormal deformation areas.

[0035] Traditional sliding window analysis methods, such as the mean method and variance method, only process the interior of the window and are easily affected by noise. Isolated noise is easily enhanced, and noise resistance is weak. In addition, they have insufficient ability to identify areas where local abnormal deformation variables are concentrated, and cannot separate areas with strong disturbance responses from areas with weak background disturbances. However, this model not only considers the disturbance intensity within the window, but also introduces neighborhood structure comparison, effectively enhancing the response of continuous boundary areas, is insensitive to isolated noise points, avoids high response misjudgment, improves noise resistance, and can accurately separate disturbance clusters from background weak disturbance areas. It is particularly suitable for the identification of hidden landslides and potential active slopes, improves recognition ability, and enhances adaptability.

[0036] The steps for using this model include: Data acquisition: Get the standard deviation deformation field from step S1 and set the sliding window length , determine the sliding window boundary and build a sliding window around each valid pixel in the image , get the standard deviation of all shape variables in the sliding window ; Disturbance response calculation: Based on the acquired data, calculate the structural difference between the current window and the 8 neighboring windows and the perturbation density of the current window ; Perturbation image output: Based on the calculated data, calculate the disturbance response value at all pixels , all pixel coordinates and the corresponding disturbance response values compose a disturbance image. The high-value areas in the image correspond to disturbance aggregation and structural mutations, which are suitable as candidate areas for geological hazard risk identification.

[0037] S3. Extract the boundary based on the disturbed image to obtain a region mask.

[0038] In this embodiment, boundary extraction includes: In remote sensing geological disaster early warning, in order to carry out subsequent regional classification, disaster level analysis or risk grid generation, the continuous response area in the disturbance image must be structured, the boundaries must be clearly defined, and it must be discretized into specific regional units. The potential disaster feature space in the continuous pixel field must be explicitly extracted to form a potential disaster area mask map that can serve as the basic input for subsequent algorithms.

[0039] This step identifies the boundaries of high-response areas in the disturbed image obtained in step S2. The Canny edge detection algorithm, combined with connected component analysis, extracts the boundaries of numerically significant high-response areas in the disturbed image. Noise, isolated points, and unstructured regions in the deformation response are filtered out. A binary mask image is then constructed for subsequent regional analysis. Regions with a value of 1 in the mask represent potential geological hazard areas, while regions with a value of 0 represent background areas. This mask image is used to label the spatial distribution of potential hazard areas. This boundary extraction technique offers advantages such as openness, transparency, controllable parameters, and robust results.

[0040] To ensure the stability of edge detection response strength, the perturbed image is first linearly normalized and noise is suppressed using a Gaussian filter to reduce the influence of isolated pixels on boundary extraction. Edge detection is then performed using the Canny algorithm. This involves using the Sobel operator to obtain gradient magnitude and direction, non-maximum suppression to eliminate redundant edges, and a double threshold to connect edge structures to ensure boundary integrity. This generates a closed or nearly closed boundary map structure, resulting in an initial boundary image where 1 represents edge pixels and 0 represents non-edge pixels. Since the initial boundary image only represents the boundary outline of the region, the interior of the boundary is filled to generate a candidate mask region map. Next, a connected component analysis algorithm is used to extract all closed regions within the candidate regions and calculate their area (number of pixels). A lower threshold is set for the area, and noise filtering is performed to remove small interference regions while retaining the main continuous deformation structure. Finally, based on the noise filtering results, a mask map of potential disaster areas is generated. This map has the same spatial dimensions as the original image, with background areas assigned 0 and candidate disaster areas assigned 1.

[0041] S4. Perform time series fitting based on the regional mask image to obtain a displacement rate image.

[0042] In this embodiment, time series fitting includes: Geological disasters, especially landslides and slow deformation processes, often have cumulative effects over time. Surface deformation in different regions exhibits not only spatial heterogeneity but also differs in temporal evolution trends. To reflect this trend in early warning systems, it is necessary to perform time series modeling on the pixel displacement data in the monitored time-series images and extract the displacement rate of each pixel to further assess the occurrence rate, activity level, and future trends of potential disasters.

[0043] In this step, based on the obtained regional mask map, a weighted perturbation trend fitting model is used to extract the surface displacement time series corresponding to each pixel in the target area in the multi-temporal remote sensing image, thereby calculating the average surface displacement rate per unit time and obtaining a displacement rate image, which is used as a basic indicator to measure the deformation activity of potential disaster areas.

[0044] Furthermore, the weighted disturbance trend fitting model includes: ; in: : The long-term trend displacement rate of the target pixel, which represents the average displacement increment per unit time; : represents the time index of remote sensing images; : time mean; : total time series length, indicating the number of periods of remote sensing images, usually ranging from 10 to 50; : The remote sensing displacement value of the target pixel (i, j) at the tth moment, where a negative value indicates sinking and a positive value indicates rising; : the mean displacement of the target pixel; : Perturbation weight factor, which indicates the importance of the t-th moment to the pixel trend calculation. The greater the change in shape, the stronger the disturbance and the higher the weight.

[0045] Furthermore, the disturbance weight factor includes: ; in: : The remote sensing displacement value of the target pixel (i, j) at the t-1th moment, where negative values indicate sinking and positive values indicate rising; : Perturbation amplification factor, which enhances the response sensitivity to displacement mutations and is adjusted according to the disturbance frequency. It is usually set to 1 to 5.

[0046] Furthermore, the displacement mean and time mean include: ; ; These two means are used to center the series so that the fitting results do not depend on the initial position and time starting point.

[0047] The model describes how to calculate the displacement rate at a pixel based on a region mask.

[0048] Traditional time series fitting methods mostly use polynomial fitting, Savitzky–Golay filtering or smoothing methods based on Kalman filtering. However, these methods are insensitive to local mutations or over-smooth, lack a trend description mechanism for disturbance energy concentration areas, and cannot combine regional disturbance weights to enhance the response modeling of key time periods. In contrast, this model models each pixel independently, eliminating the need to establish a full-region regression model. It is suitable for landslide areas with strong local heterogeneity, and through the weight enhancement mechanism, it can perceive and enhance the impact of short-term abnormal disturbances on trends, reflect the characteristics of the active stage of disasters, and improve adaptability. Moreover, based on linear trend construction, it is superior to high-order fitting methods in terms of computational efficiency and stability, and has improved noise resistance.

[0049] The steps for using the above model include: Data input: Get the regional mask image from step S3 and extract the remote sensing displacement value of the target pixel (i, j) in the candidate disaster area at T moments ; Parameter calculation: Based on the remote sensing displacement value of each pixel in the candidate disaster area , calculate the corresponding perturbation weight factor , displacement mean and time mean ; Displacement rate calculation: Based on the calculated parameters, calculate the displacement rate of each pixel in the candidate disaster area , and form a set to obtain the displacement rate image.

[0050] S5. Perform stability analysis based on the displacement rate image to obtain a critical coefficient graph.

[0051] In this embodiment, the stability analysis includes: The displacement rate image obtained in step S4 reflects the deformation intensity and trend of different areas of the surface. However, relying solely on the displacement rate cannot directly reflect the instability risk of potential disasters. Therefore, it is necessary to further construct an indicator image that can reflect the critical degree of system stability. This will make the obtained image more consistent with the critical instability characteristics of disasters such as landslides and collapses, thereby improving the accuracy of risk assessment.

[0052] Based on the obtained displacement rate image, the displacement rate of each pixel in the image is analyzed using the stability critical analysis model, and the stability critical coefficient of each pixel is calculated. The larger the stability critical coefficient, the greater the risk of geological disasters. The risk of geological disasters is judged based on the calculated stability critical coefficient.

[0053] Furthermore, the stability critical analysis model includes: ; in: : represents the critical coefficient of stability of pixel (i, j). The larger it is, the closer it is to instability. : The disturbance tolerance response of pixel (i, j), reflecting the stability of its historical fluctuations; : A small constant used to avoid zero denominator.

[0054] Furthermore, the disturbance tolerance response includes: ; in: : The maximum value of pixel displacement deviation in the entire time series; : A small constant used to avoid the denominator being zero during normalization.

[0055] This model describes how to perform stability analysis on each pixel in the image based on the displacement rate image to determine the risk of geological hazards.

[0056] Traditional geological hazard stability analysis mostly relies on field measurements (such as soil parameters and geological profiles), which makes data acquisition difficult and slow to update. However, this model does not rely on external geological parameters. It only infers the stability boundary based on remote sensing displacement data, and achieves criticality amplification by coupling rate and disturbance tolerance. It is more in line with the critical instability characteristics of disasters such as landslides and collapses, improving the practicality of the model.

[0057] The steps for using the model include: Data acquisition: Get the displacement rate image from step S4 and get the displacement rate data at each pixel and time series data parameters, including the total time series length , remote sensing displacement value of target pixel , the maximum value of displacement deviation wait; Tolerance calculation: Calculate the disturbance tolerance response based on the acquired data , which reflects the stability of the historical fluctuation of the pixel displacement; Stability analysis: Perform stability analysis based on the calculated data and calculate the critical stability coefficient The larger the qualitative critical coefficient is, the closer it is to instability and the higher the risk is. The smaller it is, the more stable the pixel point is and the lower the risk is. The critical coefficient of stability at each pixel point is combined to obtain the critical coefficient graph.

[0058] S6. Perform spatial clustering based on the critical coefficient map to obtain a disaster warning unit map.

[0059] In this embodiment, spatial clustering includes: In step S5, a critical coefficient map is obtained, which quantifies the risk size of geological disasters at each pixel point and obtains the critical coefficient of stability. However, disasters usually appear in planar spatial units rather than single pixel units, and the critical coefficient is a local indicator and cannot directly reflect the macroscopic spatial distribution law. What is needed in the actual early warning system is a disaster unit with clear boundaries for administrative response, risk notification and regional management. Therefore, the critical coefficient map must be organized into spatial divisions to aggregate the critical risk values of a large number of pixels into early warning units with spatial continuity that meet the actual needs of disaster management, and obtain a disaster early warning unit map.

[0060] The purpose of this step is to extract and partition the high-risk areas in the critical coefficient map through the spatial clustering method, and finally form a disaster warning unit map with closed space, clear boundaries, and can be used for subsequent hierarchical management, warning release, and on-site deployment. Each unit has relatively consistent critical risk characteristics, which can effectively identify warning areas with high disaster potential, regional concentration but irregular shape, thereby improving the practicality, operability and response efficiency of remote sensing monitoring of geological disasters, reducing the noise and false alarms of pixel-level results, and improving reliability. Compared with pixel-level warning maps, it has higher stability and robustness, can provide warning unit boundaries and structures, and provide a spatial framework for subsequent linkage response, information release, and rescue deployment.

[0061] Traditional clustering methods (such as K-Means and DBSCAN) perform poorly in remote sensing images with high noise and non-spherical structures. This step, however, integrates three dimensions: critical coefficient strength, spatial position continuity, and shape compactness. Clustering is performed through multi-scale spatial perception, balancing spatial topology and risk intensity. It offers multiple advantages, including dynamic cluster number estimation, boundary continuity control, and support for irregular regions. The steps include constructing a pixel adjacency graph, performing agglomerative partitioning clustering based on the graph structure, and constructing a disaster warning unit map.

[0062] The pixel adjacency graph is constructed based on the critical coefficient graph obtained in step S5. The graph structure is constructed so that each pixel is a node in the graph and edges are established between adjacent pixels, which are a weighted combination of the critical coefficient difference and the spatial distance.

[0063] Agglomerative partition clustering based on graph structure uses a graph-based agglomerative clustering algorithm. Initially, each pixel is treated as an independent cluster, and then the pixel pairs with the smallest adjacent weight are merged iteratively. The clustering termination condition is set as the internal consistency measure of the region (that is, the average value of the critical coefficient difference between all pixels in the region) is greater than the threshold. The critical coefficient between each pixel point in the formed cluster fluctuates very little, so that all clusters meet the requirements of continuous boundaries, internal consistency, and compact shape at termination.

[0064] Constructing a disaster warning unit map means assigning a unique number to each final clustering unit and generating a disaster warning unit map in raster or vector format. The attributes of each unit include spatial range, average critical coefficient and area, boundary, compactness, etc.

[0065] Example 2: This embodiment also provides a computer device, which is suitable for a geological disaster early warning method based on remote sensing monitoring, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a geological disaster early warning method based on remote sensing monitoring as proposed in the above embodiment.

[0066] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, a geological disaster early warning method based on remote sensing monitoring as proposed in the above embodiment is implemented.

[0067] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0068] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0069] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0070] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0071] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A geological disaster early warning method based on remote sensing monitoring, characterized in that: The following steps are involved: S1. Acquire multi-temporal remote sensing images of the target monitoring area and construct the standard deviation deformation field to obtain the standard deviation deformation field; S2. Perform regional sliding window analysis based on the standard deviation deformation field to obtain a disturbance image; S3, extracting the boundary based on the disturbed image to obtain a regional mask image; S4. performing time series fitting according to the regional mask image to obtain a displacement rate image; S5. Perform stability analysis based on the displacement rate image to obtain a critical coefficient graph; S6. Perform spatial clustering based on the critical coefficient map to obtain a disaster warning unit map.

2. A geological disaster early warning method based on remote sensing monitoring according to claim 1, characterized in that: The regional sliding window analysis in S2 includes: According to the obtained standard deviation deformation field, a sliding window area is defined, and the disturbance response model is used to convert the standard deviation deformation field into a disturbance image reflecting the degree of deformation anomaly aggregation.

3. A geological disaster early warning method based on remote sensing monitoring according to claim 2, characterized in that: The sliding window area includes: ； ； in: : Sliding window side length; : sliding window area; : The radius of the sliding window area; : A pixel in the sliding window area centered on pixel (i, j).

4. A geological disaster early warning method based on remote sensing monitoring according to claim 2, characterized in that: The disturbance response model includes: ； in: : The disturbance response value at pixel (i, j); : The structural difference between the sliding window centered on pixel point (i, j) and the eight neighboring windows; : The perturbation density of the sliding window centered at pixel (i, j).

5. A geological disaster early warning method based on remote sensing monitoring according to claim 4, characterized in that: The structural differences include: ； in: : The perturbation density of the sliding window centered at pixel (m,n); : The center point coordinate set of 8 neighborhood windows centered on pixel point (i, j).

6. A geological disaster early warning method based on remote sensing monitoring according to claim 4, characterized in that: The disturbance density includes: ； ； in: : standard deviation of the shape variable at the pixel point (x, y); : The mean of the standard deviation of the deformation variable within the sliding window centered at pixel point (i, j).

7. The geological disaster early warning method based on remote sensing monitoring according to claim 1 is characterized in that: The time series fitting in S4 includes: According to the obtained regional mask map, the weighted perturbation trend fitting model is used to extract the surface displacement time series corresponding to each pixel in the target area in the multi-temporal remote sensing image, thereby calculating the average surface displacement rate per unit time and obtaining the displacement rate image.

8. A geological disaster early warning method based on remote sensing monitoring according to claim 7, characterized in that: The weighted disturbance trend fitting model includes: ； in: : the long-term trend displacement rate of the target pixel; : represents the time index of remote sensing images; : time mean; : total time series length; : The remote sensing displacement value of the target pixel (i, j) at the tth moment; : the mean displacement of the target pixel; : Perturbation weight factor.

9. A geological disaster early warning method based on remote sensing monitoring according to claim 8, characterized in that: The disturbance weight factor includes: ； in: : The remote sensing displacement value of the target pixel (i, j) at the t-1th moment; : Perturbation amplification factor.

10. The geological disaster early warning method based on remote sensing monitoring according to claim 8, characterized in that: The displacement mean and time mean include: ； ； The displacement mean and time mean are used to center the series so that the fitting results do not depend on the initial position and time starting point.

Images