A mountain shadow removal method and device for flood monitoring

Abstract

The present application provides a mountain shadow removal method for flood monitoring, comprising: S1: acquiring synthetic aperture radar images; S2: inputting the synthetic aperture radar images into a water body recognition model to obtain water body distribution information; S3: inputting the water body distribution information into a shadow removal model to obtain de-shadowed water body distribution information; wherein the water body recognition model is trained based on synthetic aperture radar image samples and water body distribution information labels corresponding to the synthetic aperture radar image samples; and the shadow removal model is obtained based on a function model constructed by water body distribution information samples. The present application solves the defects of low precision, low efficiency and poor applicability of the prior art flood monitoring method, and improves the precision, efficiency and applicability of flood monitoring.

Description

Technical Field

[0001] This invention relates to the field of hydrological monitoring technology, and in particular to a method and apparatus for removing mountain shadows for flood monitoring. Background Technology

[0002] Floods are among the most severe natural disasters in the world, causing enormous casualties and economic losses globally every year. In recent years, with global warming and intensified human activities, the frequency and intensity of floods have been increasing. Therefore, timely and accurate monitoring of floods and analysis of their evolution trends are of great significance for disaster emergency management.

[0003] Currently, remote sensing technology, with its advantages of wide coverage and short revisit cycles, is gradually becoming the main means of flood monitoring. Depending on the detection method, it can be divided into optical remote sensing monitoring and radar remote sensing monitoring. However, since floods are often accompanied by adverse weather conditions such as clouds and rain, optical remote sensing satellites are unable to observe the Earth, making it difficult to obtain clear, cloud-free optical images. Conversely, Synthetic Aperture Radar (SAR), with its all-weather, 24 / 7 operational capability and its resistance to weather conditions, is playing an increasingly important role in flood monitoring.

[0004] Currently, flood monitoring methods based on synthetic aperture radar (SAR) imagery mainly include thresholding, object-oriented methods, active contouring, and machine learning. Among these, thresholding is the most commonly used method. Although it is fast and simple in principle, it struggles to meet accuracy requirements when dealing with uneven image grayscale distribution and large flood areas. Object-oriented methods utilize image texture and shape features, achieving good results, but the segmentation and classification scale parameters depend on experience and require multiple experiments to determine, resulting in low efficiency. Active contouring fully utilizes image color features and edge information, but speckle noise and complex calculations hinder its application in large-scale flood monitoring, making it less applicable. In recent years, machine learning methods have been increasingly applied to flood monitoring and extraction from SAR imagery, achieving higher extraction accuracy. Machine learning methods can fully utilize image feature information, and trained models can be used for multiple images of the same type, making them suitable for batch processing. Therefore, in the process of flood monitoring, how to apply machine learning methods to remove mountain shadows from SAR imagery to improve the accuracy and efficiency of subsequent monitoring is a crucial issue that urgently needs to be addressed in the industry. Summary of the Invention

[0005] This invention provides a method for removing mountain shadows for flood monitoring, which addresses the shortcomings of existing flood monitoring methods, such as low accuracy, low efficiency, and poor applicability, thereby improving the accuracy, efficiency, and applicability of flood monitoring.

[0006] This invention provides a method for removing hillside shadows for flood monitoring, comprising:

[0007] S1: Acquire synthetic aperture radar imagery;

[0008] S2: Input the synthetic aperture radar image into the water body identification model to obtain water body distribution information;

[0009] S3: Input the water distribution information into the shadow removal model to obtain the shadow-removed water distribution information;

[0010] The water body identification model is trained based on synthetic aperture radar image samples and water body distribution information labels corresponding to the synthetic aperture radar image samples.

[0011] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0012] According to the present invention, a method for removing mountain shadows for flood monitoring is provided, wherein the water body identification model is based on a support vector machine model, the support vector machine model is of type C_SVC, and the kernel function is of type Linear.

[0013] According to the present invention, a method for removing hillside shadows for flood monitoring includes training a water body identification model comprising:

[0014] A1: Perform mean filtering preprocessing on synthetic aperture radar image samples;

[0015] A2: Synthesize three-band image samples from preprocessed synthetic aperture radar image samples;

[0016] A3: Using three-band image samples as input and the water body distribution information corresponding to the three-band image samples as labels, a sample set for the water body identification model is established, which includes a training set and a validation set.

[0017] A4: The water body recognition model is iteratively trained using the training set, and the prediction results of the water body recognition model are compared with those of the validation set. If the results are less than the preset error range, the model converges, and the training of the water body recognition model is complete.

[0018] According to the present invention, a method for removing mountain shadows for flood monitoring is provided, wherein the synthetic aperture radar image is Sentine-1 dual-polarization data, and the three-band image includes VH features, VV features and SDWI features;

[0019] A2 includes:

[0020] A2-1: Calculate SDWI features from the VH and VV features in the preprocessed Sentinel-1 dual polarization data;

[0021] A2-2: Synthesizing three-band image samples using VH, VV and SDWI features;

[0022] The formula for calculating the SDWI feature is as follows:

[0023] K SDWI＝ ln(10×VV×VH)

[0024] In the formula, K SDWI The values ​​represent the results of band operations. VH and VV both represent Sentinel-1 dual-polarization data.

[0025] According to the present invention, a method for removing hillside shadows for flood monitoring is provided, wherein the construction of the shadow removal model includes:

[0026] B1: Interpret water body sample points and mountain shadow points from the water body distribution information sample;

[0027] B2: Extract the attribute values ​​of the corresponding locations of the mountain shadow points in the elevation and slope data, whereby the attribute values ​​include the height and slope values ​​of the corresponding locations;

[0028] B3: Based on the attribute values ​​of the corresponding locations of the mountain shadow points, draw the water sample points and mountain shadow points in a two-dimensional space of elevation and slope;

[0029] B4: Construct a linear function based on water sample points with the maximum threshold and maximum slope to identify mountain shadows;

[0030] B5: Construct a shadow removal model based on a linear function.

[0031] According to the present invention, a method for removing hillside shadows for flood monitoring is provided, wherein the formula of the linear function is as follows:

[0032] y = -0.0324x + 59.5059

[0033] In the formula, x represents the elevation value at any location, and y represents the result calculated by the formula. If y is greater than the slope value corresponding to the input x, the shadow removal model will identify this location as a mountain shadow; otherwise, it will not be a mountain shadow.

[0034] According to the method for removing mountain shadows for flood monitoring provided by the present invention, after the shadow removal model is constructed, the method further includes: evaluating the model performance by using the accuracy coefficient and the Kappa coefficient;

[0035] The formulas for calculating the accuracy coefficient and the Kappa coefficient are as follows:

[0036]

[0037]

[0038]

[0039] In the formula, acc represents the precision coefficient, pe represents the random consistency error, kappa represents the Kappa coefficient, TP represents the number of real water body pixels predicted by the model as water, FP represents the number of real non-water body pixels predicted by the model as water, FN represents the number of real water body pixels predicted by the model as non-water, TN represents the number of real non-water body pixels predicted by the model as non-water, k represents the number of sample classification categories, N represents the total number of samples, ai represents the number of real samples in each category, and bi represents the number of samples predicted by the model in each category.

[0040] The present invention also provides a hillside shadow removal device for flood monitoring, comprising:

[0041] The data acquisition module is used to acquire synthetic aperture radar images;

[0042] The water body identification module is used to input synthetic aperture radar images into the water body identification model to obtain water body distribution information;

[0043] The shadow removal module is used to input water distribution information into the shadow removal model to obtain the water distribution information without shadows.

[0044] The water body identification model is trained based on synthetic aperture radar image samples and water body distribution information labels corresponding to the synthetic aperture radar image samples.

[0045] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0046] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the hillside shadow removal method for flood monitoring as described above.

[0047] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the hillside shadow removal method for flood monitoring as described above.

[0048] This invention provides a method for removing mountain shadows in flood monitoring. It trains a water body recognition model using machine learning and constructs a shadow removal model using a function-based approach. Synthetic aperture radar (SAR) image samples and corresponding water body distribution information are used as input and labels, respectively, to train the water body recognition model, which extracts flood water body distribution information from SAR images. Based on the water body distribution information samples, a function model is constructed to obtain the shadow removal model, which removes the interference of mountain shadows from the extracted flood water body data. By sequentially inputting the current SAR image into the water body recognition model and the shadow removal model, the water body distribution information with mountain shadows removed is obtained. Quantitative analysis of this information achieves higher accuracy in flood monitoring. Considering the large geographical area of ​​the study area, the noise impact of SAR images, and the complex and diverse flood scenarios, this invention uses a combination of machine learning and function-based methods to remove mountain shadows for accurate monitoring of large-scale flood disasters. This addresses the shortcomings of existing flood monitoring methods, such as low accuracy, low efficiency, and poor applicability, thus improving the accuracy, efficiency, and applicability of flood monitoring. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is a schematic flowchart of the hillside shadow removal method for flood monitoring provided by the present invention;

[0051] Figure 2 This is a two-dimensional spatial distribution map of elevation and slope for the mountain shadow removal method for flood monitoring provided by the present invention;

[0052] Figure 3 This is a schematic diagram of a linear function for removing mountain shadows in the mountain shadow removal method for flood monitoring provided by the present invention;

[0053] Figure 4 This is a schematic diagram of the hillside shadow removal device for flood monitoring provided by the present invention;

[0054] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0056] The following is combined Figures 1-3 The first embodiment of the present invention describes a method for removing mountain shadows for flood monitoring.

[0057] like Figure 1 As shown, this embodiment of a method for removing mountain shadows for flood monitoring specifically includes the following steps (the numbering of each step in this invention is only for distinguishing steps and does not limit the specific execution order of each step):

[0058] S1: Acquire synthetic aperture radar imagery.

[0059] Google Earth Engine (GEE) is a cloud computing platform specifically designed for processing remote sensing imagery. It stores nearly 40 years of major open-access remote sensing imagery datasets, such as Landsat, Sentinel, and Modis series data. In this embodiment, the synthetic aperture radar (SAR) imagery is Sentinel-1 imagery, selected from the GEE cloud platform with an imaging time span of three months, covering the entire flood disaster phase, totaling 320 SAR images. For larger-scale flood disasters, combining GF-3 imagery can achieve better monitoring results.

[0060] S2: Input the synthetic aperture radar image into the water body identification model to obtain water body distribution information.

[0061] Taking the middle and lower reaches of the Yangtze River as an example, synthetic aperture radar images of the study area are input into a trained water body identification model for prediction and identification, thereby obtaining information on the distribution of water bodies during floods in the middle and lower reaches of the Yangtze River.

[0062] S3: Input the water distribution information into the shadow removal model to obtain the water distribution information after removing the shadow.

[0063] The water distribution information is then input into the shadow removal model to remove the interference of mountain shadows, thus obtaining the water distribution information after removing mountain shadows. Quantitative analysis of the shadow-removed water distribution information enables flood monitoring.

[0064] The water body identification model is trained based on synthetic aperture radar image samples and the corresponding water body distribution information labels.

[0065] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0066] This invention provides a method for removing mountain shadows in flood monitoring. It trains a water body recognition model using machine learning and constructs a shadow removal model using a function-based approach. Synthetic aperture radar (SAR) image samples and corresponding water body distribution information are used as input and labels, respectively, to train the water body recognition model, which extracts flood water body distribution information from SAR images. Based on the water body distribution information samples, a function model is constructed to obtain the shadow removal model, which removes the interference of mountain shadows from the extracted flood water body data. By sequentially inputting the current SAR image into the water body recognition model and the shadow removal model, the water body distribution information with mountain shadows removed is obtained. Quantitative analysis of this information achieves higher accuracy in flood monitoring. Considering the large geographical area of ​​the study area, the noise impact of SAR images, and the complex and diverse flood scenarios, this invention uses a combination of machine learning and function-based methods to remove mountain shadows for accurate monitoring of large-scale flood disasters. This addresses the shortcomings of existing flood monitoring methods, such as low accuracy, low efficiency, and poor applicability, thus improving the accuracy, efficiency, and applicability of flood monitoring.

[0067] In this embodiment, the water body identification model is based on the support vector machine model, the support vector machine model is of type C_SVC, and the kernel function type is Linear.

[0068] Support Vector Machine (SVM) is a binary classification model suitable for handling the water body identification classification problem in this invention. The SVM model type is C_SVC (C_SVM for Classification), where parameter C is the penalty coefficient of the objective function, used to balance the classification margin and misclassified samples. The kernel function of the SVM model is a linear kernel function, primarily used for linearly separable cases. In this invention, the three-band training samples are linearly separable, and the linear kernel function provides high model efficiency while maintaining high model accuracy.

[0069] In this embodiment, the training of the water body recognition model includes:

[0070] A1: Perform mean filtering preprocessing on the synthetic aperture radar image samples.

[0071] Mean filtering was performed on the Sentinel-1 imagery, with the filter window size set to 3×3 pixels to suppress speckle noise. After preprocessing, a backscatter map of the study area at a resolution of 10 meters was obtained.

[0072] A2: Synthesize three-band image samples from preprocessed synthetic aperture radar image samples.

[0073] A3: Using three-band image samples as input and the water body distribution information corresponding to the three-band image samples as labels, a sample set for establishing a water body identification model is established. The sample set includes a training set and a validation set.

[0074] This invention employs a Support Vector Machine (SVM) model to extract water body distribution during floods. The accuracy of the SVM model's predictions depends on the construction of the sample dataset. Therefore, this embodiment uses three-band image samples as input and the corresponding water body distribution information as labels, establishing a 70,000-sample set, including 50,000 training samples and 20,000 test samples. The training set includes 14,423 water body samples and 35,577 non-water body samples, while the test set includes 8,842 water body samples and 11,158 non-water body samples. The inclusion of both water and non-water body samples in the sample set helps improve the model's recognition accuracy during training. The sample set comes from six geographical regions and different time periods, ensuring the model's robustness and generalization ability.

[0075] A4: The water body recognition model is iteratively trained using the training set, and the prediction results of the water body recognition model are compared with those of the validation set. If the results are less than the preset error range, the model converges, and the training of the water body recognition model is complete.

[0076] The three-band image samples are used as input, and the corresponding water body distribution information is used as labels. During the image recognition process of model training, it is determined whether the corresponding sample is a water body. If it is a water body, the label is 1; if it is not a water body, the label is 0. By having the trained model determine whether a pixel in the image is a water body, predictions are made.

[0077] The training set and sample set were input into the support vector machine model for training. Through multiple experiments, suitable training parameters were finally obtained, namely, the support vector machine model type was C_SVC and the kernel function type was Linear. After iterative training, the model converged when the error was less than the preset error range, and the water body recognition model training was completed.

[0078] In this embodiment, the synthetic aperture radar image is Sentine-1 dual-polarization data, and the three-band image includes VH features, VV features, and SDWI features.

[0079] After preprocessing the synthetic aperture radar (SAR) image, the Sentinel-1 dual-polarization data and the derived SDWI features are then superimposed to synthesize a new three-band image. VH and VV are the polarization modes of the SAR image; VH is cross-polarization, and VV is co-polarization. In radar imaging, utilizing different polarization modes of different electromagnetic waves can acquire richer remote sensing information. Since training samples are difficult to distinguish in the two-dimensional space composed of VH and VV polarization modes, and considering the limited polarization data of Sentinel-1, this invention introduces SDWI features to enrich the data features.

[0080] A2 includes:

[0081] A2-1: Calculate SDWI features from the VH and VV features in the preprocessed Sentine-1 dual-polarization data.

[0082] A2-2: Synthesize three-band image samples using VH features, VV features, and SDWI features.

[0083] First, SDWI features are constructed using preprocessed Sentinel-1 dual-polarization data. Then, three-band image samples are synthesized using VH, VV, and SDWI features. Finally, training samples with VH-VV-SDWI feature information are input into the support vector machine model for training.

[0084] The formula for calculating the SDWI feature is as follows:

[0085] K SDWI＝ ln(10×VV×VH)

[0086] In the formula, K SDWI The values ​​represent the results of band operations, with VH and VV both representing Sentinel-1 dual-polarization data. SDWI feature references the vegetation index NDWI, which utilizes band operations between Sentinel-1 dual-polarization data to enhance water body information, achieving good water body information extraction results.

[0087] In this embodiment, the construction of the shadow removal model includes:

[0088] B1: Interpret water sample points and mountain shadow points from the water body distribution information sample.

[0089] After obtaining water distribution information during floods in the middle and lower reaches of the Yangtze River, this invention constructs a two-dimensional spatial distribution of mountain shadows using elevation data from the Digital Elevation Model (DEM) of the study area and derived slope data, thereby removing the influence of mountain shadows.

[0090] First, with the aid of Sentinel-1 imagery, several water sample points and mountain shadow points were extracted from the water body distribution map and interpreted.

[0091] B2: Extract the attribute values ​​of the corresponding locations of the mountain shadow points in the elevation and slope data. The attribute values ​​include the height and slope values ​​of the corresponding locations.

[0092] Then, the attribute values ​​corresponding to the mountain shadow points are extracted from the elevation and slope data.

[0093] B3: Based on the attribute values ​​of the corresponding locations of the mountain shadow points, draw the water sample points and mountain shadow points in a two-dimensional space of elevation and slope.

[0094] Subsequently, the mountain shadow points and water sample points were plotted in a two-dimensional space based on elevation and slope, such as... Figure 2 As shown, it can be seen that the water body and mountain shadow samples are separable in the two-dimensional space of elevation and slope.

[0095] B4: Construct a linear function based on water sample points with the maximum threshold and maximum slope to identify mountain shadows.

[0096] B5: Constructing a shadow removal model based on a linear function

[0097] This invention is based on the function method. A linear function is constructed based on the water sample points with the maximum threshold and the maximum slope. A shadow removal model is built based on the linear function. The linear function model is applied to all water distribution maps to obtain water distribution maps after removing mountain shadows. Compared with existing methods, the shadow recognition rate is significantly improved.

[0098] In this embodiment, the formula for the linear function is as follows:

[0099] y = -0.0324x + 59.5059

[0100] In the formula, x represents the elevation value at any location, and y represents the result calculated by the formula. If y is greater than the slope value corresponding to the input x, the shadow removal model will identify this location as a mountain shadow; otherwise, it will not be a mountain shadow.

[0101] like Figure 3 As shown, based on the water body sample points with the maximum threshold and maximum slope, a linear function y = -0.0324x + 59.5059 is constructed in the two-dimensional spatial distribution map of elevation and slope. The function method is used to remove mountain shadows, thus obtaining the water body distribution map after removing mountain shadows. Through this method, compared with the commonly used threshold method in existing methods, the shadow recognition rate is significantly improved in this embodiment, resulting in a better shadow removal effect. The effect of the function method used in the shadow removal model of this invention compared with other methods is shown in Table 1:

[0102] Table 1 Comparison of the effects of methods for removing mountain shadows

[0103]

[0104]

[0105] Methods 1-3 are threshold methods. In Method 1, the threshold is determined based on the maximum elevation and slope of the water samples. Applying this maximum threshold to the mountain shadow samples removes 0.706% of the mountain shadow samples. In Method 2, the threshold is determined based on the elevation and slope values ​​of the water samples, arranged from low to high up to 99%, achieving a shadow recognition rate of 70.54%. Similarly, Method 3 has a threshold ratio of 99.5%, with a shadow recognition rate of 55.94%. Method 4, using a function method, achieves a shadow recognition rate of 75.46%. Therefore, this invention selects the linear function model constructed using Method 4 to remove mountain shadows.

[0106] In this embodiment, after the shadow removal model is constructed, the method further includes: evaluating the model performance using the accuracy coefficient and the Kappa coefficient.

[0107] The formulas for calculating the accuracy coefficient and the Kappa coefficient are as follows:

[0108]

[0109]

[0110]

[0111] In the formula, acc represents the precision coefficient, pe represents the random consistency error, kappa represents the Kappa coefficient, TP represents the number of real water body pixels predicted by the model as water, FP represents the number of real non-water body pixels predicted by the model as water, FN represents the number of real water body pixels predicted by the model as non-water, TN represents the number of real non-water body pixels predicted by the model as non-water, k represents the number of sample classification categories, N represents the total number of samples, ai represents the number of real samples in each category, and bi represents the number of samples predicted by the model in each category.

[0112] To examine the extent to which the shadow removal model of this invention improves flood monitoring capabilities, this paper analyzes the method from a quantitative perspective. This invention selects two indicators—accuracy (acc) and Kappa coefficient—to quantitatively evaluate the performance of the shadow removal model. Accuracy represents the proportion of correctly classified pixels out of the total number of pixels. Its calculation is simple and is usually expressed as a percentage; a higher value indicates better model performance. However, due to the imbalance in the number of samples across classes during classification, a high accuracy rate does not necessarily indicate the accuracy of classification for each class. The Kappa coefficient, on the other hand, is a comprehensive evaluation indicator of model performance and is more reliable. Its range is between -1 and 1; a higher value indicates better predictive performance. Here, pe represents the random consistency error. Because the number of samples in each class is imbalanced during actual classification, a high accuracy coefficient may not result in good actual classification performance. Therefore, pe is introduced to address the accuracy "bias" problem caused by sample imbalance.

[0113] This embodiment selects a synthetic aperture radar image of a geographic area with a pixel size of 1319×2058 to quantitatively evaluate the effect of the shadow removal model. The selected area is located in the western part of the middle and lower reaches of the Yangtze River. 5000 points were randomly generated in the image. The shadow removal model for the selected area was quantitatively evaluated, and the water body results before and after removing the mountain shadows were compared with the actual water body. Before removing the mountain shadows, the accuracy coefficient and Kappa coefficient of the selected area were 93.06% and 0.9173, respectively. After removing the mountain shadows using the shadow removal model, the water body distribution information showed an accuracy coefficient and Kappa coefficient of 95% and 0.9315, respectively, representing improvements of 1.94% and 0.0142. It can be seen that this function model is useful for improving flood monitoring capabilities.

[0114] The following describes the hillside shadow removal device for flood monitoring provided by the present invention. The hillside shadow removal device for flood monitoring described below can be referred to in correspondence with the hillside shadow removal method for flood monitoring described above.

[0115] like Figure 4 As shown, the second embodiment of the present invention also provides a hillside shadow removal device for flood monitoring, comprising:

[0116] The data acquisition module 410 is used to acquire synthetic aperture radar images.

[0117] The water body identification module 420 is used to input synthetic aperture radar images into the water body identification model to obtain water body distribution information.

[0118] The shadow removal module 430 is used to input water distribution information into the shadow removal model to obtain the water distribution information after shadow removal.

[0119] The water body identification model is trained based on synthetic aperture radar image samples and the corresponding water body distribution information labels.

[0120] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0121] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a hillside shadow removal method for flood monitoring, the method including:

[0122] S1: Acquire synthetic aperture radar imagery.

[0123] S2: Input the synthetic aperture radar image into the water body identification model to obtain water body distribution information.

[0124] S3: Input the water distribution information into the shadow removal model to obtain the water distribution information after removing the shadow.

[0125] The water body identification model is trained based on synthetic aperture radar image samples and the corresponding water body distribution information labels.

[0126] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0127] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part 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 to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the hillside shadow removal method for flood monitoring provided by the above methods, the method comprising:

[0129] S1: Acquire synthetic aperture radar imagery.

[0130] S2: Input the synthetic aperture radar image into the water body identification model to obtain water body distribution information.

[0131] S3: Input the water distribution information into the shadow removal model to obtain the water distribution information after removing the shadow.

[0132] The water body identification model is trained based on synthetic aperture radar image samples and the corresponding water body distribution information labels.

[0133] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0134] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the hillside shadow removal method for flood monitoring provided by the methods described above, the method comprising:

[0135] S1: Acquire synthetic aperture radar imagery.

[0136] S2: Input the synthetic aperture radar image into the water body identification model to obtain water body distribution information.

[0137] S3: Input the water distribution information into the shadow removal model to obtain the water distribution information after removing the shadow.

[0138] The water body identification model is trained based on synthetic aperture radar image samples and the corresponding water body distribution information labels.

[0139] The shadow removal model is derived from a function model constructed based on water distribution information samples.

[0140] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A mountain shadow removal method for flood monitoring, characterized by, The method comprises the following steps: S1: acquiring synthetic aperture radar images; S2: inputting the synthetic aperture radar images into a water body recognition model to obtain water body distribution information; S3: inputting the water body distribution information into a shadow removal model to obtain shadow-removed water body distribution information; The water body recognition model is trained based on synthetic aperture radar image samples and water body distribution information labels corresponding to the synthetic aperture radar image samples; The shadow removal model is obtained by constructing a function model based on water body distribution information samples; The water body recognition model is based on a support vector machine model, the type of the support vector machine model is C_SVC, and the kernel function type is Linear; The training of the water body recognition model comprises the following steps: A1: performing mean filtering preprocessing on the synthetic aperture radar image samples; A2: synthesizing three-band image samples from the preprocessed synthetic aperture radar image samples; A3: taking the three-band image samples as input and the water body distribution information corresponding to the three-band image samples as labels to establish a sample set of the water body recognition model, wherein the sample set comprises a training set and a verification set; A4: iteratively training the water body recognition model through the training set, comparing the prediction results of the water body recognition model through the verification set, and converging the model if the error is less than a preset error range, thereby completing the training of the water body recognition model; The synthetic aperture radar images are Sentine-1 dual-polarization data, and the three-band images comprise VH features, VV features and SDWI features; The A2 comprises the following steps: A2-1: calculating and generating SDWI features from the VH features and VV features in the preprocessed Sentine-1 dual-polarization data; A2-2: synthesizing three-band image samples from the VH features, VV features and SDWI features; The calculation formula of the SDWI features is as follows: K SDWI= ln(10 x VV x VH) In the formula, K SDWI The results of the band operation are represented by V, and both VH and VV represent Sentinel-1 dual-polarization data.

2. The mountain shadow removal method for flood monitoring according to claim 1, wherein, The construction of the shadow removal model comprises the following steps: B1: interpreting water sample points and mountain shadow points from the water body distribution information samples; B2: extracting attribute values of the corresponding positions of the mountain shadow points in the elevation data and slope data, wherein the attribute values comprise height values and slope values of the corresponding positions; B3: according to the attribute values of the corresponding positions of the mountain shadow points, drawing the water sample points and the mountain shadow points in the two-dimensional space of elevation and slope; B4: constructing a linear function according to the water sample points with the maximum threshold value and the maximum slope to identify mountain shadows; B5: constructing a shadow removal model based on the linear function.

3. The mountain shadow removal method for flood monitoring according to claim 2, wherein, The formula of the linear function is as follows: y = -0.0324x + 59.5059 In the formula, x represents the elevation value of any position, and y represents the result value calculated by the formula. If y is greater than the slope value corresponding to the input x, the shadow removal model identifies this position as a mountain shadow; otherwise, it is not a mountain shadow.

4. The mountain shadow removal method for flood monitoring according to claim 2, wherein, After the construction of the shadow removal model is completed, the method further comprises evaluating the model performance by using a precision coefficient and a Kappa coefficient; The calculation formula of the precision coefficient and the Kappa coefficient is as follows: ； ； ； In the formula, acc represents an accuracy coefficient, pe represents a chance agreement error, kappa represents a Kappa coefficient, TP represents the number of real water body pixels predicted by the model as water body, FP represents the number of real non-water body pixels predicted by the model as water body, FN represents the number of real water body pixels predicted by the model as non-water body, TN represents the number of real non-water body pixels predicted by the model as non-water body, k represents the number of sample classification categories, N represents the total number of samples, ai represents the number of real samples of each category, and bi represents the number of samples of each category predicted by the model.

5. A mountain shadow removal device for flood monitoring, characterized by, Comprise: a data acquisition module configured to acquire synthetic aperture radar images; a water body identification module configured to input the synthetic aperture radar images into a water body identification model to obtain water body distribution information; a shadow removal module configured to input the water body distribution information into a shadow removal model to obtain shadow-removed water body distribution information; wherein the water body identification model is trained based on synthetic aperture radar image samples and water body distribution information labels corresponding to the synthetic aperture radar image samples; the shadow removal model is obtained based on a function model constructed based on water body distribution information samples; the water body identification model is based on a support vector machine model, the type of the support vector machine model is C_SVC, and the kernel function type is Linear; the training of the water body identification model comprises: A1: performing mean filtering preprocessing on the synthetic aperture radar image samples; A2: synthesizing three-band image samples from the preprocessed synthetic aperture radar image samples; A3: taking the three-band image samples as input and the water body distribution information corresponding to the three-band image samples as labels to establish a sample set of the water body identification model, the sample set comprising a training set and a validation set; A4: iteratively training the water body identification model through the training set and comparing the prediction results of the water body identification model through the validation set, the model converges when the prediction results are less than a preset error range, and the training of the water body identification model is completed; the synthetic aperture radar images are Sentine-1 dual-polarization data, and the three-band images comprise VH features, VV features and SDWI features; the A2 comprises: A2-1: calculating the VH features and the VV features in the preprocessed Sentine-1 dual-polarization data to generate SDWI features; A2-2: synthesizing three-band image samples from the VH features, the VV features and the SDWI features; the calculation formula of the SDWI features is as follows: K SDWI= ln(10 x VV x VH) In the formula, K SDWI The results of the band operation are represented by V, and both VH and VV represent Sentinel-1 dual-polarization data.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, the processor executes the program to implement the method for removing mountain shadows for flood monitoring according to any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, the computer program is executed by the processor to implement the method for removing mountain shadows for flood monitoring according to any one of claims 1 to 4.

Images