Remote sensing extraction method of land-water transition zone terrain information based on classification rules

Abstract

The present invention discloses a classification rule-based remote sensing method for extracting terrain information from a hydro-terrestrial ecotone, including: preprocessing collected time-series radar data to obtain a time-series radar image set, constructing an NDWI index image based on the collected time-series optical remote sensing data; using the OTSU algorithm to binary-classify the time-series radar image set and the NDWI index image into water bodies and land, respectively, to obtain a time-series water range dataset; using the midline of a permanent water body as the longitudinal section of the water area to obtain a time-series water level dataset; based on the time-series water range dataset, eliminating noise points within temporary water areas, and using the ROC curve method to obtain the optimal water level demarcation threshold to obtain elevation. The present invention determines elevation using information about the dynamic process of flooding, enabling the acquisition of a detailed DEM for a large-scale hydro-terrestrial ecotone in an environment with frequent water level fluctuations. The method is simple to operate and improves accuracy and adaptability.

Description

Technical Field

[0001] The invention belongs to the field of geographic information technology, and in particular relates to a remote sensing extraction method for land-water transition zone terrain information based on classification rules. Background Art

[0002] The ecotone, an ecological transition zone between inland water ecosystems and terrestrial ecosystems, encompasses riparian zones, lakeshore zones, drawdown zones, and coastal zones. It is characterized by ecological fragility, heterogeneity, dynamism, and high biological complexity. Due to interfacial effects, ecotones perform important ecological functions within complex ecosystems, including maintaining species diversity, conserving water, stabilizing adjacent ecosystems, and intercepting the flow of matter and energy. Topography, a key factor influencing the local environment, plays a crucial role in quantifying water-level fluctuations, water storage, and sedimentation rates within the ecotone. It is also highly correlated with vegetation growth and provides crucial information for environmental protection and research. Accurate and timely topographic information is urgently needed to enable accurate hydrological dynamics simulations. Among these, the accuracy of digital elevation models (DEMs) is considered the most important factor influencing the accuracy of hydrodynamic simulations.

[0003] With the rapid development of remote sensing technology, numerous global-scale DEM datasets, such as SRTM, GMTED2010, and ASTER GDEM, have emerged and are widely used in various fields. However, due to the frequent water level fluctuations and inundation in ecotones, traditional methods are difficult to obtain DEMs. Furthermore, the topography of ecotones has undergone significant changes driven by changes in land cover and hydrological dynamics. Therefore, a feasible method for timely updating DEM information in ecotones is urgently needed.

[0004] Optical imagery and LiDAR are commonly used data for obtaining fine DEMs, but their application over large areas is time-consuming and labor-intensive, and flooding can also hinder the application of these methods. SAR data can be used to obtain DEMs through interference methods, but this method is sensitive to land cover changes and is not suitable for land-water transition zones where water levels fluctuate frequently. The waterline method uses the ever-changing waterline as a tool for measuring elevation and is used to obtain topographic information of large-scale tidal flats. Due to the presence of noise in SAR images and images of ships in rivers or lakes, there will be many misclassifications, which will affect the accuracy of the waterline and thus reduce the accuracy of the DEM. At the same time, a lot of work is required to interpolate the null value areas. Therefore, there is an urgent need for a more convenient and feasible method to avoid these effects and obtain accurate DEM information. Summary of the Invention

[0005] The purpose of the embodiment of the present invention is to provide a remote sensing extraction method for terrain information of the terrestrial and hydrological transition zone based on classification rules, which determines the elevation by information on the dynamic process of flooding, and realizes the acquisition of fine DEM of a large-scale terrestrial and hydrological transition zone in an environment with frequent water level fluctuations. The operation is simple and the accuracy and adaptability are improved.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is a remote sensing method for extracting terrain information of the terrestrial and hydrozonal zone based on classification rules, comprising the following steps:

[0007] S1. Preprocess the collected time series radar data to obtain a time series radar image set, and construct an NDWI index image based on the collected time series optical remote sensing data;

[0008] S2. Use the OTSU algorithm to classify the time series radar image set and the NDWI index image into water bodies and land, respectively, to obtain the time series water range dataset;

[0009] S3. Use the middle line of the permanent water body as the longitudinal section line of the water area, obtain the line distance from each position on the longitudinal section line to the starting point of the longitudinal section line, and assign the line distance value on the longitudinal section line to each pixel in the study area according to the principle of closest distance. Obtain the section line distance data for the entire study area. Based on the water level and section line distance of the hydrological station, calculate the water level of each pixel point using the IDW algorithm, and then obtain the time series water level dataset.

[0010] S4. Based on the time series water range dataset, pixels that are always in water or land state are identified, and the ALOSDEM data corrected based on the water level data of the hydrological station is used as their terrain information; other pixels are identified as temporary water areas, and the noise points in the temporary water areas are eliminated. The ROC curve method is used to obtain the optimal water level division threshold and elevation, thereby obtaining the terrain information of the temporary water areas.

[0011] Furthermore, in S3, the water level formula for each pixel is calculated using the IDW algorithm:

[0012]

[0013] Where Hs(t) represents the water level of pixel s at time t, H A (t) and H B (t) are the water levels of the two hydrological stations A and B at the beginning and end of the profile at time t, respectively. A and D B are the profile distances of hydrological stations A and B, respectively, and Ds represents the profile distance of pixel s.

[0014] Furthermore, in said S4, the method for removing noise points in the temporary water area is:

[0015] S41, based on the JM distance separability index, the separability of flooding and exposure at each location in the time series water level dataset of the study area is determined. If the JM distance is greater than the separability threshold, it is determined that there is no noise point at that location and it can be directly classified. The optimal water level division threshold, i.e., the elevation value, for distinguishing flooding and exposure is obtained by the ROC curve method. If the JM distance of a pixel in the time series water level dataset is less than or equal to the separability threshold, then there is a noise point at that location.

[0016] S42: For data with noise or anomalies, sort the exposed and flooded water levels from low to high, and calculate the water level difference between adjacent water levels. If the water level difference exceeds the noise point discrimination threshold, it is considered a noise point and needs to be removed; if the water level difference is below the threshold, it is not a noise point.

[0017] S43, after removing the noise points, returns to S41, repeats the iterative discrimination until all the data can be separated, and finally obtains the optimal water level division threshold, i.e., the elevation value, for distinguishing flooding and exposure through the ROC curve method.

[0018] Furthermore, the method for obtaining the optimal water level division threshold by using the ROC curve method includes the following steps:

[0019] The actual time series flooding information of each location is used as the true classification value. If the pixel is covered by water, it is marked as a water body; otherwise, if the pixel is not covered by water, it is marked as land.

[0020] According to the time information of the flooding process, the water level information at the corresponding moment in the time series water level dataset obtained by S3 is extracted and the water levels are arranged in order from low to high;

[0021] Starting from the lowest water level, gradually increase the water level as the classification threshold; when the water level of a pixel is lower than the set threshold, the point is predicted to be land; conversely, when the water level is higher than or equal to the set threshold, the point is predicted to be a water body; in this way, based on the water level of each location time series, the prediction classification result of the time series is generated;

[0022] For each set threshold and classification result, the relationship between the true positive rate and false positive rate of pixels in the temporary water area at different thresholds is plotted through the ROC curve, and the optimal classification threshold, that is, the elevation value, is selected.

[0023] Furthermore, the calculation formula of the JM distance separability index is:

[0024]

[0025] JM=2×(1-e -B )

[0026] Where, σ W , σ L Represent the water level variance of water body and land respectively; m W and m L They represent the average water levels of water bodies and land respectively; e represents the base of natural logarithm.

[0027] Furthermore, in S41, the separability threshold of the JM distance is obtained by statistically calculating the JM index of the noise-free data.

[0028] Furthermore, in S42, the noise point discrimination threshold is obtained by statistically calculating the water level difference of the hydrological stations.

[0029] The beneficial effects of the present invention are:

[0030] The present invention uses the elevation of each pixel to determine whether it is submerged under water under water level fluctuations. That is, if a location is flooded, it means that its elevation is lower than the water level of the water surface at that location. If the location is exposed, it means that the elevation of the location is higher than the potential water level at that location. The elevation is the threshold that determines whether it can be flooded. Therefore, by using time series water level fluctuations and flooding process information, the ROC curve (Receiver Operating Characteristic) two-class optimal threshold acquisition algorithm is adopted to determine the optimal water level division threshold (i.e., elevation). In this way, the elevation acquisition problem can be converted into a two-class classification problem. At the same time, the JM separability index is used to eliminate noise and outliers, thereby accurately obtaining seamless DEM information of the water-land transition zone.

[0031] Compared with the prior art, the present invention also has the following advantages:

[0032] 1. Based on whether each pixel is covered by water and the water level at the corresponding moment, the present invention uses information about the dynamic process of flooding to determine elevation. This successfully converts elevation acquisition into a two-class classification problem, enabling the acquisition of fine DEMs for large-scale terrestrial and hydrozonal zones in environments with frequent water level fluctuations. This method can fully transform the adverse factors of the flooding process into effective information for DEM acquisition. Terrestrial and hydrozonal zones are prone to drastic water level fluctuations, with alternating flooding and exposure, and rapid changes in water conditions, which seriously hinder terrain acquisition methods such as ground surveying, lidar, and SAR. The present invention uses the flooding process as effective information about the terrain of the measured area. The higher the flooding frequency, the more accurate the DEM obtained by this method.

[0033] 2. The present invention removes noise through JM separability and the distance to adjacent water levels, thereby reducing the impact of noise (such as ship interference). Finally, the ROC method is used to determine the optimal water level threshold to obtain the elevation, which has high adaptability and fault tolerance.

[0034] 3. The present invention obtains the threshold value by calculating the statistical information of water level and water range. The steps are simple and the method has strong universality. It provides a simple, efficient and timely method for obtaining the terrain of the terrestrial and hydrozonal zone. It has broad application prospects in terrestrial and hydrozonal areas where the water level fluctuates violently. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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.

[0036] Figure 1 Schematic diagram of the principle of an embodiment of the present invention.

[0037] Figure 2 Flowchart of an embodiment of the present invention.

[0038] Figure 3 This is a section line distance diagram of Dongting Lake in an embodiment of the present invention.

[0039] Figure 4 2 is an example diagram of the relationship between water level and water range in an embodiment of the present invention.

[0040] Figure 5 This is the DEM result diagram in the embodiment of the present invention. The left side is the DEM of the entire study area (named ROC DEM) obtained in the embodiment of the present invention, and the right side is the local area DEM and the corresponding Sentinel-2 remote sensing image.

[0041] Figure 6 This is a comparison chart of the results obtained by an embodiment of the present invention; A, B, C, and D in the figure are the DEM, ALOSDEM, SRTM DEM, and Sentinel-2 remote sensing images obtained by the embodiment of the present invention, respectively. Image 1 on the right shows the actual road photographed at location 1, and Image 2 shows the remote sensing image of the low-lying area at location 2.

[0042] Figure 7 A comparison chart of elevation values ​​generated by the embodiment of the present invention and other methods. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] A remote sensing method for extracting terrain information from terrestrial and hydrozonal zones based on classification rules, see Figures 1 and 2 , including the following steps:

[0045] S1. Preprocess time series radar and optical remote sensing data.

[0046] S11. In this embodiment of the present invention, the Sentinel-1 radar dataset (i.e., time series radar data) of Dongting Lake from 2017 to 2020 in IW mode and VH polarization, provided by the European Space Agency's official website, is used as a long-term remote sensing image series. Simultaneously, Sentinel-2 optical remote sensing data is obtained as supplementary data for obtaining flooding information. Sentinel-1 is a Sentinel radar remote sensing satellite, and Sentinel-2 is a Sentinel optical remote sensing satellite.

[0047] The S12 and Sentinel-1 radar datasets were preprocessed using the process creation tool in the SNAP software. The preprocessing process was then batch processed using the General Processing Tool (GPT). The preprocessing steps included thermal noise removal, radiometric correction, speckle filtering, terrain correction, and decibel conversion to obtain the radar backscatter coefficients.

[0048] Sentinel-2 optical remote sensing data preprocessing uses the Sen2Cor module in the SNAP software to perform atmospheric correction and index calculation to obtain the Normalized Difference Water Index (NDWI). The NDWI index can effectively distinguish between water bodies and non-water areas and is used to extract flooding information from optical remote sensing images.

[0049] S2. Based on the OTSU algorithm, the Sentienl-1 image set and NDWI are used to classify water bodies and land, and a high-temporal-resolution time-series water extent dataset is constructed. To obtain water extent data with higher spatial resolution, optical and radar data are combined to jointly construct a water extent time-series dataset (i.e., time-series flooding extent information).

[0050] S21. In Sentinel-1 images, the specular reflection of radar signals reaching the water surface causes the water surface to exhibit a lower backreflection coefficient than other land features. The OTSU algorithm is used to automatically classify radar images based on the backscattering rate, thereby obtaining a water range dataset for the radar image time series.

[0051] In S22 and NDWI data, the index value of water bodies is higher, while the index value of non-water bodies is lower. The OTSU algorithm is also used to automatically classify the NDWI index images to obtain the water range dataset of the optical image time series.

[0052] S3. Generate a profile distance map based on the river water profile, combine it with the water level data of the hydrological station, and use the IDW algorithm to obtain the time series water level dataset.

[0053] S31, obtaining data; In the embodiment of the present invention, water level data from two water level stations A and B (Chenglingji and Xiaohezui) are used;

[0054] S32. Describe the middle line of the permanent water body of lakes and rivers as the longitudinal section line of the lakes and rivers. In the embodiment of the present invention, the middle line of the permanent water body of Dongting Lake is drawn as the longitudinal section line in the GIS software, such as Figure 3 The line shown is the longitudinal section of Dongting Lake. Calculate the distance from each location along the longitudinal section line to the starting point of the longitudinal section line, hydrological station A (Chenglingji), i.e., the section distance.

[0055] S33. For a certain pixel, calculate the distance from each position on the longitudinal section line to the pixel using GIS software, and find the position on the longitudinal section line closest to the pixel; use the section line distance value of the nearest position on the searched longitudinal section line as the section line distance value of this pixel; traverse each pixel in the study area to generate a section line distance map for the entire study area. Figure 3 Shown is the section distance map of Dongting Lake.

[0056] S34. To obtain water level information for each pixel, this embodiment interpolates the water level using the inverse distance weighted (IDW) method based on the profile distance map. If the location is flooded, the actual water level information is obtained; if the location is not flooded, the potential water level information is obtained. Based on the time-series water level data from the hydrological station, a time-series water level dataset can be obtained for each pixel.

[0057] At time t, the water level Hs(t) of a pixel s is calculated as:

[0058]

[0059] Where Hs(t) represents the water level of pixel s at time t, H A (t) and H B (t) are the water levels of the two hydrological stations A and B at the beginning and end of the profile at time t, respectively. A and D B are the profile distances of hydrological stations A and B, respectively, and Ds represents the profile distance of pixel s.

[0060] S4. Use the separability JM index and water level distance to filter and eliminate noise points. Use the ROC method to obtain the optimal water level division threshold, thereby obtaining the elevation information of the hydro-land transition zone. The specific method is as follows:

[0061] S41. Based on the time series water extent dataset obtained from S21 and S22, identify permanent water bodies, permanent land, and temporary water areas.

[0062] See also Figure 4 In A, within the water level range, the pixels whose corresponding positions are always in the land state are permanent land, and the pixels whose corresponding positions are always in the water state in B are permanent water bodies; pixels other than these two states are identified as temporary water bodies, which are further divided into noiseless water point locations such as C, G, and noisy water point locations such as E, F, H, I, J, K, and L.

[0063] S42. The elevation of each pixel in permanent water bodies and land areas is determined by calibrated ALOSDEM (Advanced Land Observing Satellite) data.

[0064] By constructing a linear relationship between the ALOSDEM elevation values ​​at the water-land boundary of hydrological stations A and B and the water levels of the hydrological stations, the conversion relationship between water level and ALOSDEM was obtained, thereby correcting the ALOSDEM data. To ensure the continuity of the digital elevation model (DEM) information, the elevations of permanent water bodies / land areas were filled in with the corrected ALOS DEM data.

[0065] S43. Determine the elevation of temporary water areas.

[0066] S431. Use the JM distance criterion to determine the classification separability of the water level time series set and determine the separability threshold.

[0067] In this embodiment, the JM distance criterion is used to determine whether the water level time series set can be directly classified. In order to balance high separability and ensure that the introduced noise is minimized, according to the histogram statistics of the JM distance index of the noise-free data in this implementation case, it is found that most JM values ​​are greater than 1.5. Therefore, in this implementation case, the separability threshold is set to 1.5, and the JM distance calculation formula is:

[0068]

[0069] JM=2×(1-e -B )

[0070] Where, σ W , σ L Represent the water level variance of water body and land respectively; m W and m L denote the average water level of water bodies and land, respectively. JM denotes the JM distance; and e denotes the base of the natural logarithm.

[0071] S432. When the separability is greater than the threshold, it indicates that the separability of the point is good and there are no noise points (abnormal water points). The ROC curve method can be used to determine the optimal classification threshold to obtain the elevation. The specific process is as follows:

[0072] S4321. The actual time series flooding information of each location is used as the classification truth value. If the pixel is covered by water, it is marked as a water body; otherwise, if the pixel is not covered by water, it is marked as land.

[0073] S4322. Extract the water level information at the corresponding moment in the time series water level dataset obtained in S3 according to the time information of the time series flooding range dataset, and arrange the water levels in order from low to high to form a time series water level dataset.

[0074] S4323. Starting from the lowest water level, gradually increase the water level as the classification threshold. When the water level at the corresponding moment extracted from each pixel S35 is lower than the set threshold, the point is predicted to be land; the reason is that this threshold is the set elevation value. If the elevation of the location is higher than the water level, that is, when the water level is lower than the threshold, it means that the point will be exposed and is land); conversely, when the water level is higher than or equal to the set threshold, the point is predicted to be a water body. In this way, for each threshold, there will be a prediction result at each water level, and the time series water level will obtain a time series prediction result. According to different classification thresholds and time series water levels, a series of prediction and classification results can be generated.

[0075] S4324. For each set threshold and classification result, combined with its classification true value, the number of water bodies whose true value is water and predicted to be water bodies is counted as the true positive number TP, the number of water bodies whose true value is water bodies and predicted to be land is counted as the false negative number FN, the number of water bodies whose true value is land and predicted to be water bodies is counted as the false positive number FP, and the number of water bodies whose true value is land and predicted to be land is counted as the true negative number TN. Then calculate the proportion of samples predicted to be positive and actually positive to all positive samples, that is, the true positive rate (TPR), and the proportion of samples predicted to be positive but actually negative to all negative samples, that is, the false positive rate (FPR). The calculation formula is:

[0076] TPR=TP / (TP+FN)

[0077] FPR=FP / (FP+TN)

[0078] Where TP, FN, FP, and TN represent the number of true positives, false negatives, false positives, and true negatives, respectively.

[0079] S4325. Plot the ROC curve with FPR as the horizontal axis and TPR as the vertical axis. The ROC curve shows the performance of the classifier at different thresholds. The closer the curve is to the upper left corner, the better the classifier performance. Ideally, we want to find a threshold that maximizes TPR while minimizing FPR.

[0080] S4326. Determine the optimal classification threshold, i.e., the elevation. The value corresponding to the upper left corner is the threshold with the most ideal classification effect, because this point indicates that the true positive rate is high while the false positive rate is low, that is, the classifier has the best performance. The calculation formula is:

[0081] q(c)=TPR

[0082] p(c)=1-FPR

[0083]

[0084] Where T represents the optimal threshold, TPR and FPR represent the true positive rate and false positive rate under the threshold c, respectively, and q(c) and p(c) represent the accuracy and specificity under the threshold c, respectively.

[0085] S433: When the separability is less than the threshold (1.5), it means that the separability of the water point at this location is poor, and there are many noise points and abnormal points. It is necessary to identify and remove abnormal points based on the adjacent water level difference. The specific steps for removing noise are:

[0086] S4331. Under normal circumstances, flooded water points will cluster in one area, while non-flooded water points will also cluster in another area. If some water points are too far away from the clustered area, they are considered to be noise abnormal water points. In order to obtain this distance threshold, the water level difference of the hydrological station consistent with the time information of the water range data is calculated (i.e., in the embodiment, the water level difference of the two adjacent water levels of the hydrological station in S4322 is calculated). Based on the statistically obtained water level difference histogram, the upper limit of the 95% confidence interval is used as the threshold for anomaly detection, thereby determining the maximum threshold of the water level difference to be 0.9. If this value is exceeded, it indicates that these points are potential abnormal points and need to be removed.

[0087] The calculation formula of adjacent water level difference LD is:

[0088] LD=neigborhood_dis(Sort(water_level))

[0089] In the formula, Sort means sorting the water level of this point in chronological order, and neigborhood_dis() means calculating the water level difference between adjacent moments of this point.

[0090] Step 4332: Calculate the adjacent water level differences of other pixels.

[0091] In step 4333, all points where the water level difference exceeds the threshold are considered noise points and removed from the time series data. Points where the water level difference does not exceed the threshold are retained.

[0092] Repeat S41 to S4333 until the water points of all pixels are separable. Use the ROC curve method in S432 to determine the optimal classification threshold, that is, the elevation value, and finally obtain the DEM data of the study area, as shown in Figure 5 shown.

[0093] Accuracy assessment:

[0094] After realizing remote sensing extraction of land-water ecotone terrain information based on classification rules, the ALOSDEM, SRTM DEM and field-collected elevation data in the study area were collected and compared qualitatively and quantitatively with the elevation data obtained by the embodiment of the present invention. The quantitative comparison indicators include: R 2 Coefficient of determination, MAE mean absolute error, and RMSE root mean square error.

[0095] Figure 5 On the left is the DEM of the entire study area (named ROC DEM) obtained by the embodiment of the present invention, and on the right is the DEM of the local area and the corresponding remote sensing image. It can be seen that the obtained DEM has a high degree of consistency with the actual terrain.

[0096] Figure 6 This is a qualitative comparison of the visual effects of the DEM (A) obtained by the embodiment of the present invention with the ALOSDEM (B), SRTM DEM (C), and Sentinel-2 remote sensing image (D) in a local area of ​​the study area. According to the comparison results, the DEM obtained by the embodiment of the present invention is quite close to the actual terrain reflected by the Sentinel-2 remote sensing image, especially the middle road in the local area ( Figure 6 Red circle 1) and near the center of the lake ( Figure 6 In the area of ​​the red circle 2), the DEM obtained by the embodiment of the present invention can well reflect the roads and low-lying areas.

[0097] Figure 7 is the DEM obtained by the embodiment of the present invention ( Figure 6 The scatter plots of ROC DEM), ALOSDEM, and SRTM DEM were made with the elevation data collected in the field (as the real value) to conduct quantitative comparison. Among the three data products, the DEM obtained by the embodiment of the present invention is the most relevant and close to the elevation data collected in the field. The R 2 It reaches 0.9, and both MAE and RMSE are minimal, indicating that compared with ALOSDEM and SRTM DEM products, the DEM obtained by the embodiment of the present invention has the highest accuracy in this area and is closest to the true value;

[0098] Comparison results show that the method provided by the present invention uses readily available site water level data and Sentinel-1 and Sentinel-2 remote sensing data, and determines elevation based on whether each pixel is covered by water under water level fluctuations. This strategy converts elevation acquisition into a two-class classification problem, effectively transforming dynamic flooding information that hinders terrain monitoring into effective information for terrain acquisition. In summary, the method provided by the present invention has high fault tolerance and reliability, reduces the impact of noise and vessel interference, and achieves high-precision DEM information extraction in the terrestrial and hydrozonal zone.

[0099] Each embodiment in this specification is described in a related manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiment is generally similar to the method embodiment, so the description is relatively simple. For related parts, refer to the description of the method embodiment.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A remote sensing method for extracting terrain information of a hydro-terrestrial transition zone based on classification rules, characterized in that: The following steps are involved: S1. Preprocess the collected time series radar data to obtain a time series radar image set, and construct an NDWI index image based on the collected time series optical remote sensing data; S2. Use the OTSU algorithm to classify the time series radar image set and the NDWI index image into water bodies and land, respectively, to obtain the time series water range dataset; S3. Use the middle line of the permanent water body as the longitudinal section line of the water area, obtain the line distance from each position on the longitudinal section line to the starting point of the longitudinal section line, and assign the line distance value on the longitudinal section line to each pixel in the study area according to the principle of closest distance. Obtain the section line distance data for the entire study area. Based on the water level and section line distance of the hydrological station, calculate the water level of each pixel point using the IDW algorithm, and then obtain the time series water level dataset. S4. Based on the time series water range dataset, pixels that are always in water or land state are identified, and the ALOSDEM data corrected based on the water level data of the hydrological station is used as their terrain information; other pixels are identified as temporary water areas, and the noise points in the temporary water areas are eliminated. The ROC curve method is used to obtain the optimal water level division threshold and elevation, thereby obtaining the terrain information of the temporary water areas.

2. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 1 is characterized in that: In S3, the water level formula for each pixel is calculated using the IDW algorithm: Where Hs(t) represents the water level of pixel s at time t, H A (t) and H B (t) are the water levels of the two hydrological stations A and B at the beginning and end of the profile at time t, respectively. A and D B are the profile distances of hydrological stations A and B, respectively, and Ds represents the profile distance of pixel s.

3. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 1 is characterized in that: In said S4, the method for removing noise points in the temporary water area is as follows: S41, based on the JM distance separability index, the separability of flooding and exposure at each location in the time series water level dataset of the study area is determined. If the JM distance is greater than the separability threshold, it is determined that there is no noise point at that location and it can be directly classified. The optimal water level division threshold, i.e., the elevation value, for distinguishing flooding and exposure is obtained by the ROC curve method. If the JM distance of a pixel in the time series water level dataset is less than or equal to the separability threshold, then there is a noise point at that location. S42: For data with noise or anomalies, sort the exposed and flooded water levels from low to high, calculate the water level difference between adjacent water levels, and if the water level difference exceeds the noise point discrimination threshold, it is considered a noise point and needs to be removed; If the water level difference is lower than the threshold, it means it is not a noise point; S43, after removing the noise points, returns to S41, repeats the iterative discrimination until all the data can be separated, and finally obtains the optimal water level division threshold, i.e., the elevation value, for distinguishing flooding and exposure through the ROC curve method.

4. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 3 is characterized in that: The method for obtaining the optimal water level division threshold by the ROC curve method includes the following steps: The actual time series flooding information of each location is used as the true classification value. If the pixel is covered by water, it is marked as a water body; otherwise, if the pixel is not covered by water, it is marked as land. According to the time information of the flooding process, the water level information at the corresponding moment in the time series water level dataset obtained by S3 is extracted and the water levels are arranged in order from low to high; Starting from the lowest water level, gradually increase the water level as the classification threshold; when the water level of a pixel is lower than the set threshold, the point is predicted to be land; conversely, when the water level is higher than or equal to the set threshold, the point is predicted to be a water body; in this way, based on the water level of each location time series, the prediction classification result of the time series is generated; For each set threshold and classification result, the relationship between the true positive rate and false positive rate of pixels in the temporary water area at different thresholds is plotted through the ROC curve, and the optimal classification threshold, that is, the elevation value, is selected.

5. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 3 is characterized in that: The calculation formula of the JM distance separability index is: JM=2×(1-e -B ) Where, σ W , σ L Represent the water level variance of water body and land respectively; m W and m L They represent the average water levels of water bodies and land respectively; e represents the base of natural logarithm.

6. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 3 is characterized in that: In the above S41, the separability threshold of the JM distance is obtained by statistically analyzing the JM index of the noise-free data.

7. The remote sensing extraction method for land-water transition zone terrain information based on classification rules according to claim 3 is characterized in that: In the above S42, the noise point discrimination threshold is obtained by counting the water level differences of the hydrological stations.

Images