A method for identifying glacier seasonal variation by remote sensing
By preprocessing and registering remote sensing images of glaciers in different seasons, and combining STL decomposition and LSTM neural networks, the problem of lacking seasonal analysis in glacier analysis was solved, the accuracy and stability of glacier monitoring were achieved, and detailed information on the seasonal changes of glaciers was provided.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID SICHUAN ELECTRIC POWER CORP ELECTRIC POWER RES INST
- Filing Date
- 2026-03-05
- Publication Date
- 2026-07-07
AI Technical Summary
Existing glacier analysis methods lack seasonality analysis, resulting in inaccurate glacier monitoring and difficulties in obtaining high-quality remote sensing data and distinguishing glaciers from surrounding land features in glacier areas.
By acquiring remote sensing image sets of glaciers in different seasons, preprocessing, data registration, and image adjustment are performed. STL decomposition and LSTM neural network are used to analyze glacier boundary parameters. Combined with improved single-channel algorithm and small baseline time series analysis, the seasonal changes of glaciers are identified.
It improves the accuracy and stability of identifying seasonal changes in glaciers, eliminates the influence of weather factors, accurately extracts glacier boundary parameters, and provides information on the seasonal changes in glacier area, length, and temperature.
Smart Images
Figure CN122347735A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of radar remote sensing image processing technology, and in particular to a remote sensing identification method for seasonal changes in glaciers. Background Technology
[0002] Glaciers are the product of a combination of factors including climate and topography, and they can record information about environmental changes, thus often being considered sensitive indicators of climate change. Furthermore, glacial meltwater is a vital water source for many rivers and lakes, especially in arid and semi-arid regions. Therefore, by studying the seasonal changes in glaciers, we can monitor and record trends and patterns of climate change, particularly in the context of global warming, where glacial retreat and growth can provide important information about changes in the climate system. Understanding the seasonal melting patterns of glaciers helps predict water availability and its impacts on agriculture, industry, and ecosystems.
[0003] The area and volume of glaciers fluctuate seasonally due to climate change, which can lead to difficulties in identification. Furthermore, due to the special characteristics of glacier regions, such as high altitude and cloud cover, obtaining high-quality remote sensing data is also a challenge. In addition, distinguishing glaciers from surrounding land features (such as snow, moraine, and shadows) is a difficult point in remote sensing identification, and existing glacier analysis methods lack seasonal analysis, resulting in insufficient accuracy in glacier monitoring. Summary of the Invention
[0004] In view of this, the purpose of this invention is to provide a remote sensing identification method for seasonal changes in glaciers, which solves the problem that existing glacier analysis methods lack seasonal analysis, resulting in inaccurate glacier monitoring.
[0005] This invention provides a remote sensing method for identifying seasonal changes in glaciers, comprising: S1: Obtain remote sensing images of glaciers in the study area in each season, and construct a remote sensing image set from the remote sensing images of each season; S2: Preprocess the remote sensing image set to obtain the preprocessed remote sensing image set, and perform data registration on the preprocessed remote sensing image set to obtain the registered remote sensing image set; S3: Perform image adjustment, boundary extraction, and seasonal inversion on the registered remote sensing image set to obtain the glacier boundary parameter set; S4: The set of glacier boundary parameters is analyzed by STL decomposition to obtain the results of seasonal changes in glaciers.
[0006] Preferably, step S2 specifically includes: S21: Perform radiometric calibration on the remote sensing image set to obtain the remote sensing image set for the first processing stage; S22: Perform atmospheric correction on the remote sensing image set of the first processing stage to obtain the remote sensing image set of the second processing stage; S23: Select remote sensing images with panchromatic and multispectral data from the remote sensing image set in the second processing stage to obtain the preprocessed remote sensing image set; S24: Register the remote sensing images in the preprocessed remote sensing image set in the same geospatial coordinate system to obtain the registered remote sensing image set.
[0007] Preferably, step S3 specifically includes: S31: Set the window size to... In the domain, extract the remote sensing image of the i-th season from the registered remote sensing image set. ; S32: Calculate and obtain remote sensing images Medium pixel grayscale value and in pixels Average gray value within the center area ; S33: Through and The ratio R is calculated using the following formula:
[0008] Where E() represents the expectation; Set ratio threshold ;like Then the pixels Assign to the glacier region, otherwise use pixels Assign to the background area; S34: Repeat steps S32-S33 to process the remote sensing image. All pixels are allocated to obtain the adjusted remote sensing image. ; S35: Repeat steps S31-S34 to obtain all the adjusted remote sensing images; S36: Calculate the rate of change of total glacier area, the length-area weighted value of glacier, and the glacier surface temperature for each season using the adjusted remote sensing images; S37: By improving the single-channel algorithm, seasonal inversion and small baseline time series analysis of glacier surface temperature in each season are performed to obtain the glacier surface deformation in each season. S38: A set of glacier boundary parameters consisting of the rate of change of total glacier area in each season, the weighted value of glacier length and area, glacier surface temperature, and glacier surface deformation.
[0009] Preferred: The formula for calculating the rate of change of the total glacier area is:
[0010] Where i is the season number, For the first The rate of change of total glacier area within each season For the first The change in total glacier area within a season For the first The total area of glaciers in each season For the first The time interval between seasons.
[0011] Preferred: The formula for calculating the length-area weighted value of a glacier is:
[0012] in, For the first The weighted average value of glacier length and area within each season. For the first The rate of change of total glacier area within a given season, where L is the glacier length obtained through visual interpretation. and All of these are impact factors.
[0013] Preferred: The formula for calculating glacier surface temperature is:
[0014] in, For surface emissivity, For band variation parameters, This represents the top radiance value of the atmospheric spectrum. The brightness temperature of band 10 of the TIRS sensor. atmospheric water vapor content The function.
[0015] Preferred: The expression is:
[0016] in, , and for of Correlation coefficient; The expression is:
[0017] in, , and for of Correlation coefficient; The expression is:
[0018] in, , and for of Correlation coefficient; All are atmospheric water vapor content The coefficient.
[0019] Preferably, step S4 specifically includes: S41: Construct the inner and outer loops of the STL decomposition; S42: Seasonal and trend smoothing of the glacier boundary parameter set are performed through an inner loop to obtain the seasonal series. and trend sequence ; S43: Seasonal sequence through outer loop and trend sequence Perform residual calculations to obtain the residual series. ; S44: Through residual series Train the LSTM neural network to obtain a trained LSTM neural network; S45: Input the set of glacier boundary parameters into the trained LSTM neural network to obtain the results of seasonal changes in the glacier.
[0020] A storage medium storing instructions and data for implementing the remote sensing identification method for seasonal glacier changes.
[0021] A remote sensing device for identifying seasonal glacier changes includes: a processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement the remote sensing method for identifying seasonal glacier changes.
[0022] The present invention has the following beneficial effects: By preprocessing the remote sensing image set to eliminate the influence of factors such as weather, and by registering the remote sensing image set to improve accuracy, and by adjusting the image, extracting the boundary, and retrieving the seasonality of the registered remote sensing image set, a set of glacier boundary parameters is obtained. Seasonal analysis is introduced into glacier analysis, taking into account the influence of many factors on the data, and improving the stability of glacier seasonal change identification. Attached Figure Description
[0023] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 Flowchart for Fourier-Melin transform registration; Figure 3 This is a schematic diagram of STL decomposition. Figure 4 This is a diagram of the LSTM neural network structure. Figure 5 This is a system device diagram according to an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0024] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0025] Reference Figure 1 This application provides a remote sensing method for identifying seasonal changes in glaciers, including: S1: Obtain remote sensing images of glaciers in the study area in each season, and construct a remote sensing image set from the remote sensing images of each season; Specifically, LandSat satellite imagery data from multiple years and ASTER GDEM digital elevation model data were selected to obtain land cover information at different time points. The remote sensing image set was divided, with 70% used as the training set and 30% as the validation set.
[0026] S2: Preprocess the remote sensing image set to obtain the preprocessed remote sensing image set, and perform data registration on the preprocessed remote sensing image set to obtain the registered remote sensing image set; Furthermore, step S2 specifically involves: S21: Perform radiometric calibration on the remote sensing image set to obtain the remote sensing image set for the first processing stage; Specifically, the data is first processed using radiometric calibration. The main method of radiometric calibration is to measure the sensor's response to a known radiating target and then correct the sensor's radiometric error based on the actual radiative value of the target and the response value. The correction mainly includes the intensity of the radiative response, corresponding to the spectral amplitude and wavelength of the radiating target. Radiation intensity correction refers to measuring the corresponding sensor response value, such as power and frequency, within the effective spatial range. Spectral calibration of ground objects involves measuring the sensor's spectral response as a function of incident wavelength and frequency. Currently, linear models are generally used for radiometric calibration. ( Gain : offset)(1); S22: Perform atmospheric correction on the remote sensing image set of the first processing stage to obtain the remote sensing image set of the second processing stage; Specifically, the data is then atmospherically corrected using the FLAASH atmospheric correction module. The specific correction formula is as follows: (2) in, The radiance value of a single pixel on a satellite sensor; The surface reflectance representing a pixel; The surface reflectance of a single pixel and its surrounding mixed pixels; The reflectivity of the balloon surface; This represents the radiance value of atmospheric radiation entering the sensor; These are atmospheric condition coefficients. They are used in the model. The above processing ensures data quality and eliminates the influence of sensors and atmospheric conditions; S23: Select remote sensing images with panchromatic and multispectral data from the remote sensing image set in the second processing stage to obtain the preprocessed remote sensing image set; Specifically, select remote sensing images with panchromatic and multispectral data, ensuring they were acquired at the same or similar times to reduce errors caused by time differences; S24: Register the remote sensing images in the preprocessed remote sensing image set in the same geospatial coordinate system to obtain the registered remote sensing image set; For details, please refer to Figure 2 The image is registered using Fourier Merlin transform. Step S24 is as follows: S241. Fill in the original image to create two images of the same size, denoted as... .
[0027] S242, Regarding the image and Perform Bartlett triangle window processing and record it as... and The specific formula is as follows: (3) The corresponding window spectrum function is: (4) Its amplitude spectrum function is: (5) S243. Processed image as well as Perform a Fourier transform and center the frequency spectrum after the transform, which is... and .
[0028] S244 and After high-pass filtering, a logarithmic polar coordinate transformation is performed, the cross-power spectrum is obtained using phase correlation, and then a Fourier transform is performed to obtain the image. and Rotation and scaling parameters.
[0029] S245. The image to be registered is inversely transformed using the rotation and scaling parameters obtained above, resulting in the image. .
[0030] S246, The restored image Compared with reference image There is only a translation transformation between them, and the translation variable is obtained again using the frequency domain phase correlation method.
[0031] S247. Standardize and enhance the input image to improve image contrast.
[0032] S248. Extract image texture and edge features. Texture and edge feature extraction is implemented using the LBP algorithm and the Sobel operator, namely: (6) in It performs an image expansion operation. This refers to performing an erosion operation on an image.
[0033] S249. Using PCA decomposition and reconstruction, the extracted features are fused with grayscale features to obtain a multi-feature fused pixel image. This involves reading the equalized remote sensing image and obtaining the corresponding matrix, denoted as... Size is ;Read the marker as Texture feature matrix Read-in is recorded as Edge feature matrix Normalize between 0 and 1, respectively, using the following formula: (7) Later Matrix conversion The matrix, denoted as Find The correlation matrix, denoted as Find the eigenvalues and eigenvectors The corresponding principal component matrix is obtained, denoted as . The obtained principal component transformation matrix is transformed into The matrix, denoted as .in: (8) (9) The result obtained in the previous step Arrange from smallest to largest, then flip left and right. The first corresponding channel is the first principal component, normalized texture feature. and edge features Histogram matching is performed with the second and third principal components respectively. After matching, the corresponding principal component matrices are replaced with... and Then the matrix Transform into The matrix is then multiplied by the transpose of the eigenvectors to obtain the reconstructed fused image: (10).
[0034] S3: Perform image adjustment, boundary extraction, and seasonal inversion on the registered remote sensing image set to obtain the glacier boundary parameter set; Step S3 is as follows: S31: Set the window size to... In the domain, extract the remote sensing image of the i-th season from the registered remote sensing image set. ; Specifically, the size of the domain can be set to ; S32: Calculate and obtain remote sensing images Medium pixel grayscale value and in pixels Average gray value within the center area ; S33: Through and The ratio R is calculated using the following formula: (11) Where E() represents the expectation; Set ratio threshold ;like Then the pixels Assign to the glacier region, otherwise use pixels Assign to the background area; S34: Repeat steps S32-S33 to process the remote sensing image. All pixels are allocated to obtain the adjusted remote sensing image. ; S35: Repeat steps S31-S34 to obtain all the adjusted remote sensing images; S36: Calculate the rate of change of total glacier area, the length-area weighted value of glacier, and the glacier surface temperature for each season using the adjusted remote sensing images; Furthermore, the formula for calculating the rate of change of the total glacier area is: (12) Where i is the season number, For the first The rate of change of total glacier area within each season For the first The change in total glacier area within a season For the first The total area of glaciers in each season For the first The time interval between seasons; The formula for calculating the length-area weighted value of a glacier is: (13) in, For the first The weighted average value of glacier length and area within each season. For the first The rate of change of total glacier area within a given season, where L is the glacier length obtained through visual interpretation. and All are impact factors; The formula for calculating glacier surface temperature is: (14) in, For surface emissivity, For band variation parameters, This represents the top radiance value of the atmospheric spectrum. The brightness temperature of band 10 of the TIRS sensor. atmospheric water vapor content The function.
[0035] The expression is: (15) in, , and for of Correlation coefficient; The expression is: (16) in, , and for of Correlation coefficient; The expression is: (17) in, , and for of Correlation coefficient; All are atmospheric water vapor content The coefficient; S37: By improving the single-channel algorithm, seasonal inversion and small baseline time series analysis of glacier surface temperature in each season are performed to obtain the glacier surface deformation in each season. Specifically, the small baseline time series analysis technique first generates multiple master image sequence interferograms by combining short spatiotemporal baselines, performs spatial filtering on the differential interferometric phase, and identifies slow loss-correlation filtered phase pixels based on average spatial coherence; then, it performs three-dimensional phase unwrapping and singular value decomposition to solve the single master image phase sequence; finally, it uses spatiotemporal filtering to estimate and remove atmospheric delay phase to obtain topographic elevation error and deformation sequence information.
[0036] S38: A set of glacier boundary parameters consisting of the rate of change of total glacier area in each season, the weighted value of glacier length and area, glacier surface temperature, and glacier surface deformation.
[0037] S4: The set of glacier boundary parameters is analyzed by STL decomposition to obtain the results of seasonal changes in glaciers.
[0038] Furthermore, the STL decomposition consists of two processes, including an inner loop and an outer loop. The inner loop is nested within the outer loop. The main steps of the inner loop are seasonal smoothing and trend smoothing.
[0039] refer to Figure 3 Step S4 is as follows: S41: Construct the inner and outer loops of the STL decomposition; S42: Seasonal and trend smoothing of the glacier boundary parameter set are performed through an inner loop to obtain the seasonal series. and trend sequence ; Furthermore, S42 specifically refers to: S421. Eliminate the trend by reverting from the initial value. Subtract the trend value from the middle to obtain a new sequence. ; S422. Perform Cycle-subseries smoothing. A locally weighted scatter plot smoothing regression method is applied to each obtained Cycle-subseries data. The result is expressed as follows: ; S423, Low-pass filtering, consists of three steps. First, a moving average of length n is applied to the original data. Then, another moving average of length n is applied, and finally, a moving average of length 3 is applied. Afterwards, locally weighted scatter plot smoothing regression is applied to the result of the low-pass filtering, denoted as... ; S424. Eliminate the trend, from Subtract Get seasonal sequence ,Right now: (18) S45. Eliminate seasonality, through... minus Obtain the non-seasonal data. Then, smooth the trend by applying locally weighted scatter plot smoothing regression to the seasonalized series to obtain the trend series. ; S43: Seasonal sequence through outer loop and trend sequence Perform residual calculations to obtain the residual series. ; Specifically, the following is the outer loop, firstly and Obtained from the inner loop. The residual series can then be calculated. The specific formula is as follows: (19) Robustness weight parameters Used to estimate robustness, It is moment v The robustness weights are given by the following formula: (20) The formula for the bivariate weight function B is as follows: (twenty one) S44: Through residual series Train the LSTM neural network to obtain a trained LSTM neural network; Specifically, the prediction results of the subsequences are added to the prediction results of the original data, and then the calculation is performed. The mean absolute error (MAE) and root mean square error (RMSE) are calculated using the following formulas: (twenty two) in For predicted values, It is the actual value. yes The average value, The number of data points in the test set; Adjust the parameters to find the optimal decomposition frequency and LSTM time step until the metrics can no longer be significantly improved. The model with the best metrics in the test set can be used to fit the seasonal variations of various glacier parameters.
[0040] refer to Figure 4 The LSTM neural network modeling process is as follows: LSTM neural networks consist of three state gates. LSTM can use these three gates to delete or add information to the cell state. The forget gate determines how much information obtained in the previous time step can be retained in the current time step. The specific formula is as follows: (twenty three) In the formula, Here is the weight matrix for the forget gate. For the current input, For the output before the memory block, For the bias term of the forget gate, for Type function.
[0041] The input gate determines the input from the current input. How much of the information obtained can be stored in the cellular state? middle: (twenty four) Output gate and output value Closely related, the specific formula is as follows: (25) Old cells To the new state The entire transformation process is as follows: It is an active function. The specific formula is as follows: (26) S45: Input the set of glacier boundary parameters into the trained LSTM neural network to obtain the results of seasonal changes in the glacier.
[0042] Please see Figure 5 , Figure 5 This is a schematic diagram of the hardware device in operation according to an embodiment of the present invention. The hardware device specifically includes: a remote sensing identification device 401 for seasonal changes in glaciers, a processor 402, and a storage medium 403.
[0043] A remote sensing identification device 401 for glacier seasonality changes: The remote sensing identification device 401 for glacier seasonality changes implements the remote sensing identification method for glacier seasonality changes.
[0044] Processor 402: The processor 402 loads and executes the instructions and data in the storage medium 403 to implement the remote sensing identification method for seasonal changes in glaciers.
[0045] Storage medium 403: The storage medium 403 stores instructions and data; the storage medium 403 is used to implement the remote sensing identification method for seasonal changes in glaciers.
[0046] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.
[0047] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as identifiers.
[0048] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A remote sensing method for identifying seasonal changes in glaciers, characterized in that, include: S1: Obtain remote sensing images of glaciers in the study area in each season, and construct a remote sensing image set from the remote sensing images of each season; S2: Preprocess the remote sensing image set to obtain the preprocessed remote sensing image set, and perform data registration on the preprocessed remote sensing image set to obtain the registered remote sensing image set; S3: Perform image adjustment, boundary extraction, and seasonal inversion on the registered remote sensing image set to obtain the glacier boundary parameter set; S4: The set of glacier boundary parameters is analyzed by STL decomposition to obtain the results of seasonal changes in glaciers.
2. The remote sensing identification method for seasonal glacier changes according to claim 1, characterized in that, Step S2 is as follows: S21: Perform radiometric calibration on the remote sensing image set to obtain the remote sensing image set for the first processing stage; S22: Perform atmospheric correction on the remote sensing image set of the first processing stage to obtain the remote sensing image set of the second processing stage; S23: Select remote sensing images with panchromatic and multispectral data from the remote sensing image set in the second processing stage to obtain the preprocessed remote sensing image set; S24: Register the remote sensing images in the preprocessed remote sensing image set in the same geospatial coordinate system to obtain the registered remote sensing image set.
3. The remote sensing identification method for seasonal glacier changes according to claim 1, characterized in that, Step S3 is as follows: S31: Set the window size to... In the domain, extract the remote sensing image of the i-th season from the registered remote sensing image set. ; S32: Calculate and obtain remote sensing images Medium pixel grayscale value and in pixels Average gray value within the center area ; S33: Through and The ratio R is calculated using the following formula: Where E() represents the expectation; Set ratio threshold ;like Then the pixels Assign to the glacier region, otherwise use pixels Assign to the background area; S34: Repeat steps S32-S33 to process the remote sensing image. All pixels are allocated to obtain the adjusted remote sensing image. ; S35: Repeat steps S31-S34 to obtain all the adjusted remote sensing images; S36: Calculate the rate of change of total glacier area, the length-area weighted value of glacier, and the glacier surface temperature for each season using the adjusted remote sensing images; S37: By improving the single-channel algorithm, seasonal inversion and small baseline time series analysis of glacier surface temperature in each season are performed to obtain the glacier surface deformation in each season; S38: A set of glacier boundary parameters consisting of the rate of change of total glacier area in each season, the weighted value of glacier length and area, glacier surface temperature, and glacier surface deformation.
4. The remote sensing identification method for seasonal glacier changes according to claim 3, characterized in that: The formula for calculating the rate of change of total glacier area is: Where i is the season number, For the first The rate of change of total glacier area within each season For the first The change in total glacier area within a season For the first The total area of glaciers in each season For the first The time interval between seasons.
5. The remote sensing identification method for seasonal glacier changes according to claim 3, characterized in that: The formula for calculating the length-area weighted value of a glacier is: in, For the first The weighted average value of glacier length and area within each season. For the first The rate of change of total glacier area within a season, where L is the glacier length obtained through visual interpretation. and All of these are impact factors.
6. The remote sensing identification method for seasonal glacier changes according to claim 3, characterized in that: The formula for calculating glacier surface temperature is: in, For surface emissivity, For band variation parameters, This represents the top radiance value of the atmospheric spectrum. The brightness temperature of band 10 of the TIRS sensor. atmospheric water vapor content The function.
7. The remote sensing identification method for seasonal glacier changes according to claim 6, characterized in that: The expression is: in, , and for of Correlation coefficient; The expression is: in, , and for of Correlation coefficient; The expression is: in, , and for of Correlation coefficient; All are atmospheric water vapor content The coefficient.
8. The remote sensing identification method for seasonal glacier changes according to claim 1, characterized in that, Step S4 is as follows: S41: Construct the inner and outer loops of the STL decomposition; S42: Seasonal and trend smoothing of the glacier boundary parameter set are performed through an inner loop to obtain the seasonal series. and trend sequence ; S43: Seasonal sequence through outer loop and trend sequence Perform residual calculations to obtain the residual series. ; S44: Through residual series Train the LSTM neural network to obtain a trained LSTM neural network; S45: Input the set of glacier boundary parameters into the trained LSTM neural network to obtain the results of the seasonal changes of the glacier.
9. A storage medium, characterized in that: The storage medium stores instructions and data to implement the remote sensing identification method for glacier seasonal changes as described in any one of claims 1 to 6.
10. A remote sensing identification device for seasonal changes in glaciers, characterized in that: include: A processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement the remote sensing identification method for glacier seasonal changes as described in any one of claims 1 to 6.