Tidal flat change information extraction method, device and electronic equipment

Abstract

The present application provides a kind of beach change information extraction method, device and electronic equipment, it is related to remote sensing image information extraction technical field, method includes: obtaining the remote sensing image sequence of target time resolution acquisition in target period corresponding to target area;Respectively based on each first remote sensing image in remote sensing image sequence, obtain and each first remote sensing image respectively corresponding first ground water body vector;Respectively based on each first ground water body vector, and the first remote sensing image corresponding to each first ground water body vector, using region growing algorithm, obtain and each first remote sensing image respectively corresponding second ground water body vector;Topological relationship operation is carried out to first ground water body vector and second ground water body vector, obtains and each first remote sensing image respectively corresponding beach vector;Based on each beach vector, determine the beach change information corresponding to target area of target area.The present application can realize the high space-time precision accurate extraction of beach change information of target area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing image information extraction, and in particular to a tidal flat change information extraction method and device and electronic equipment. BACKGROUND

[0002] Tidal flat is the general term of beach, river beach and lake beach, and is referred to as "tidal zone" in geomorphology, which is an important land resource and space resource. Therefore, it is of great significance to monitor the dynamic change of tidal flat.

[0003] In recent years, remote sensing technology has become an effective method for monitoring the dynamic change of tidal flat, because it has the characteristics of macroscopic, rapid, high frequency observation and low cost, and can be used to monitor a variety of ground objects and land cover types in a large range. However, in the monitoring of the dynamic change of tidal flat, due to the combined influence of imaging mode, atmospheric conditions and tidal conditions, the related technology cannot accurately extract the change information of tidal flat based on the obtained remote sensing image.

[0004] Therefore, how to accurately extract the change information of tidal flat with high precision has become a problem to be solved in the industry. SUMMARY

[0005] In view of the problems in the prior art, the present application provides a tidal flat change information extraction method, device and electronic equipment.

[0006] In a first aspect, the present application provides a tidal flat change information extraction method, comprising:

[0007] obtaining a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target area;

[0008] obtaining a first ground water body vector corresponding to each first remote sensing image in the remote sensing image sequence based on the first remote sensing image respectively;

[0009] obtaining a second ground water body vector corresponding to each first remote sensing image based on each first ground water body vector and the first remote sensing image corresponding to each first ground water body vector using a region growing algorithm;

[0010] topological relationship operation is performed on the first ground water body vector and the second ground water body vector corresponding to each first remote sensing image respectively to obtain a tidal flat vector corresponding to each first remote sensing image;

[0011] determining the tidal flat change information corresponding to the target area based on each tidal flat vector.

[0012] Optionally, the present application provides a tidal flat change information extraction method, wherein the first ground water body vector corresponding to each first remote sensing image is obtained based on the respective first remote sensing image.

[0013] The polarization band, terrain feature and texture feature corresponding to each first remote sensing image are obtained based on the respective first remote sensing image.

[0014] The multi-dimensional feature space corresponding to each first remote sensing image is constructed based on the polarization band, terrain feature and texture feature corresponding to the respective first remote sensing image.

[0015] The first ground water body vector corresponding to each first remote sensing image is obtained based on the respective multi-dimensional feature space.

[0016] Optionally, the present application provides a tidal flat change information extraction method, wherein the first ground water body vector corresponding to each first remote sensing image is obtained based on the respective multi-dimensional feature space.

[0017] The multi-dimensional feature space is input into a support vector machine classification model to obtain the first ground water body vector corresponding to each first remote sensing image output by the support vector machine classification model.

[0018] Optionally, the present application provides a tidal flat change information extraction method, wherein after the multi-dimensional feature space is input into a support vector machine classification model to obtain the first ground water body vector corresponding to each first remote sensing image output by the support vector machine classification model, the method further comprises:

[0019] The first ground water body vector is denoised to obtain a denoised first ground water body vector.

[0020] Optionally, the present application provides a tidal flat change information extraction method, wherein before the multi-dimensional feature space is input into a support vector machine classification model to obtain the first ground water body vector corresponding to each first remote sensing image output by the support vector machine classification model, the method further comprises:

[0021] A segmentation threshold for segmenting water body and non-water body is determined.

[0022] The sample remote sensing image is threshold segmented based on the segmentation threshold to obtain the target vector data corresponding to the sample remote sensing image, wherein the target vector data includes water body vector data and non-water body vector data.

[0023] Respectively selecting a plurality of sample points from each of the target vector data, to obtain a plurality of sample point sets;

[0024] Training an initial support vector machine classification model based on the plurality of sample point sets and a cross-validation method, to obtain the support vector machine classification model after training.

[0025] Optionally, according to the beach change information extraction method provided by the present application, the second ground water body vector corresponding to each of the first remote sensing images is obtained by using a region growing algorithm based on each of the first ground water body vector and the first remote sensing image corresponding to each of the first ground water body vector, and the method comprises the following steps:

[0026] Respectively determining a target remote sensing image corresponding to a VH polarization band in a polarization band corresponding to each of the first remote sensing images;

[0027] Respectively taking each of the first ground water body vector as a mask, extracting a pixel in each of the target remote sensing images as a seed point set, and respectively determining a threshold range meeting a growing condition corresponding to each of the seed point sets based on each of the seed point sets;

[0028] Respectively based on each of the target remote sensing images, each of the seed point sets and each of the threshold ranges, performing region growing, and in a case where each of the seed point sets is determined to be empty, obtaining the second ground water body vector corresponding to each of the first remote sensing images.

[0029] Optionally, according to the beach change information extraction method provided by the present application, the threshold range meeting the growing condition corresponding to each of the seed point sets is determined based on each of the seed point sets, and the method comprises the following steps:

[0030] Respectively determining a mean value and a standard deviation value of a seed point in each of the seed point sets;

[0031] Respectively based on the mean value and the standard deviation value corresponding to each of the seed point sets, determining the threshold range meeting the growing condition corresponding to each of the seed point sets.

[0032] Optionally, according to the beach change information extraction method provided by the present application, the polarization band, the terrain feature and the texture feature corresponding to each of the first remote sensing images are obtained based on each of the first remote sensing images, and the method comprises the following steps:

[0033] Respectively pre-processing each of the first remote sensing images, to obtain a polarization band corresponding to each of the first remote sensing images, and a second remote sensing image corresponding to the polarization band;

[0034] obtain the terrain feature and the texture feature corresponding to each of the first remote sensing images based on the second remote sensing image;

[0035] The preprocessing includes thermal noise removal, track correction, radiation scaling, coherent spot filtering, terrain correction and decibel processing.

[0036] In a second aspect, the present application provides a beach change information extraction device, comprising:

[0037] An acquisition module is configured to acquire a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target region;

[0038] A first obtaining module is configured to obtain a first surface water body vector corresponding to each of the first remote sensing images based on each of the first remote sensing images in the remote sensing image sequence;

[0039] A second obtaining module is configured to obtain a second surface water body vector corresponding to each of the first remote sensing images by using a region growing algorithm based on each of the first surface water body vectors and the first remote sensing image corresponding to each of the first surface water body vectors;

[0040] A third obtaining module is configured to perform a topological relationship operation on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images to obtain a beach vector corresponding to each of the first remote sensing images;

[0041] A determination module is configured to determine beach change information corresponding to the target region based on each of the beach vectors.

[0042] In a third aspect, the present application provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the beach change information extraction method of the first aspect when executing the program.

[0043] The present application provides a beach change information extraction method, device and electronic equipment, which comprises the following steps: acquiring a remote sensing image sequence collected in a target period corresponding to a target region with a target time resolution; obtaining a plurality of first ground water body vectors based on each first remote sensing image in the remote sensing image sequence; obtaining a plurality of second ground water body vectors based on each first ground water body vector and the first remote sensing image corresponding to each first ground water body vector by using a region growing algorithm; performing a topological relationship operation on the plurality of first ground water body vectors and the plurality of second ground water body vectors to obtain a plurality of beach vectors corresponding to the target region; and determining the beach change information corresponding to the target region based on the plurality of beach vectors. Since the change of the tidal level of the target region is considered when selecting the remote sensing image, the remote sensing image sequence collected in the target period corresponding to the target region with the target time resolution is selected to extract the beach change information. The second ground water body vector is extracted based on the first ground water body vector and the first remote sensing image by using the region growing algorithm, so that the extraction accuracy of the ground water body vector is improved. The topological relationship operation is performed on the first ground water body vector and the second ground water body vector to obtain the plurality of beach vectors corresponding to the target region. Finally, the beach change information corresponding to the target region is determined based on the plurality of beach vectors, so that the high spatio-temporal accuracy of the beach change information of the target region is realized. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0045] Figure 1 It is a flowchart of the beach change information extraction method provided by the present application;

[0046] Figure 2 It is a schematic diagram of the optimal classification line of the support vector machine provided by the present application;

[0047] Figure 3 It is a schematic diagram of the classification accuracy contour determined according to C and Gamma provided by the present application;

[0048] Figure 4 It is a flowchart of the preprocessing of the Sentinel-1A GRD image provided by the present application;

[0049] Figure 5 It is a flowchart of the beach vector extraction provided by the present application;

[0050] Figure 6is a structural schematic diagram of a beach change information extraction device provided by the present application.

[0051] Figure 7 is a physical structure schematic diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0052] For the purpose, technical solutions and advantages of the present application to be more clear, the technical solutions in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] It should be noted that, in the description of the present application, the terms "first", "second" and the like are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second" and the like are generally of a kind and do not limit the number of objects, for example, the first object can be one or more.

[0054] In order to facilitate a clearer understanding of the embodiments of the present application, some related background knowledge is first introduced as follows.

[0055] Beach refers to the tidal zone between high tide and low tide, the beach between the normal water level and flood level of rivers and lakes, the beach below the seasonal lake and river flood level, and the beach area between the normal water level and the maximum flood level of reservoirs and ponds. According to the material composition of the beach, it can be divided into three types: rock beach, sand beach and mud beach; according to the tide level, width and slope, it can be divided into three types: high tide beach, medium tide beach and low tide beach. Due to the diversity of the types of coast, the effects of water flow and the sediment content of rivers, some are eroded by water and the beach retreats towards the land; some accumulate strongly and the beach extends towards the water; some are relatively stable and the range of the beach is also relatively stable. With the increasing influence of season, climate change and environment, some natural and semi-artificial surface water bodies rapidly decrease in area in a short period of time, eventually forming a beach with a certain area between the normal water level and the low water level.

[0056] The microwave scattering characteristics of the beach are related to geological characteristics (such as fine sand, coarse sand, silt and gravel), surface roughness and water content, etc. Generally, the greater the water content and surface roughness, the stronger the scattering ability of the beach to the microwave, thereby causing different backscattering intensities of the same beach on the synthetic aperture radar (SAR) images in different periods. Therefore, it is difficult to directly classify and extract the beach features on the SAR images, and certain techniques need to be developed to realize the dynamic monitoring of the beach.

[0057] However, due to the periodic submergence of the beach by tidal water, the muddy beach surface and poor accessibility, and frequent changes, the traditional investigation method is difficult to meet the change monitoring needs of this highly dynamic environment. Due to the macroscopic, rapid, high-frequency observation and low-cost characteristics of remote sensing technology, it can be used to monitor a variety of ground objects and land cover types in a large area, and provides the possibility for long-term large-area sea, river, lake and beach resource investigation and timely monitoring.

[0058] However, in the dynamic change monitoring of the beach, due to the combined influence of imaging methods, atmospheric conditions and tidal conditions, there are still many difficulties in the application of remote sensing image data, such as tidal level correction problems, differences in large-scale beach information acquisition, and low probability of image acquisition at low tide, making it difficult to accurately obtain the change information of the beach.

[0059] The beach change information extraction method, device and electronic equipment provided by the present application will be exemplarily introduced below in combination with the accompanying drawings.

[0060] Figure 1 is a flowchart of the beach change information extraction method provided by the present application, as shown in Figure 1 The method comprises the following steps.

[0061] Step 100, acquiring a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target area;

[0062] Step 110, respectively based on each first remote sensing image in the remote sensing image sequence, obtaining a first ground water body vector corresponding to each first remote sensing image;

[0063] Step 120, respectively based on each first ground water body vector and the first remote sensing image corresponding to each first ground water body vector, using a region growing algorithm to obtain a second ground water body vector corresponding to each first remote sensing image;

[0064] In step 130, topological relationship operation is performed on the first surface water body vector and the second surface water body vector corresponding to each first remote sensing image respectively, to obtain a tidal flat vector corresponding to each first remote sensing image respectively.

[0065] In step 140, based on each tidal flat vector, tidal flat change information corresponding to the target region is determined.

[0066] It should be noted that the execution subject of the tidal flat change information extraction method provided by the embodiments of the present application can be an electronic device, a component in the electronic device, an integrated circuit, or a chip. The electronic device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a wearable device, an Ultra-mobile Personal Computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), etc., and the non-mobile electronic device can be a server, a Network Attached Storage (NAS), a Personal Computer (PC), a Television (TV), a teller machine, or a self-service machine, etc., and the embodiments of the present application do not make a specific limitation in this regard.

[0067] In the following, taking the computer executing the tidal flat change information extraction method provided by the present application as an example, the technical solutions of the embodiments of the present application are described in detail.

[0068] Specifically, in order to overcome the defect that the prior art is difficult to accurately obtain the change information of the mudflat based on the obtained remote sensing image, the present application obtains a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target region, then obtains a plurality of first ground water body vectors based on each first remote sensing image in the remote sensing image sequence, and then obtains a plurality of second ground water body vectors based on each first ground water body vector and the first remote sensing image corresponding to each first ground water body vector by using a region growing algorithm, and further performs a topological relationship operation on the plurality of first ground water body vectors and the plurality of second ground water body vectors to determine a plurality of mudflat vectors corresponding to the target region, and finally determines the mudflat change information corresponding to the target region based on the plurality of mudflat vectors; since the present application considers the change of the tidal level of the target region when selecting the remote sensing image, the remote sensing image sequence collected at the target time resolution in the target period corresponding to the target region is selected for extracting the mudflat change information, and the second ground water body vector is extracted by using the region growing algorithm based on the extracted first ground water body vector and the first remote sensing image, thereby improving the extraction accuracy of the ground water body vector, and then the topological relationship operation is performed on the first ground water body vector and the second ground water body vector to determine the plurality of mudflat vectors corresponding to the target region, and finally the mudflat change information corresponding to the target region is determined based on the plurality of mudflat vectors, thereby realizing the high spatio-temporal accuracy extraction of the mudflat change information of the target region.

[0069] It should be noted that in the embodiments of the present application, the target region refers to a region for which the mudflat change information needs to be extracted. For example, the mudflat between the normal water level and the flood water level of river A.

[0070] It should be noted that the mudflat in the embodiments of the present application can be the mudflat associated with a lake or reservoir, which refers to the region formed by the lake water rise in the lake or reservoir and the drought of the weather, or refers to the overwater region of the lake or reservoir without vegetation.

[0071] Optionally, the remote sensing image sequence collected at a target time resolution in a target period corresponding to a target region can be obtained, wherein the target period can be adaptively set based on actual application, and the embodiments of the present application do not make specific limitation thereon, for example, the target period is one year; the range of the target time resolution can be 5-12 days, for example, the target time resolution can be 5 days, 10 days or 12 days, and the embodiments of the present application do not make specific limitation thereon.

[0072] For example, the target period is one year, and the target time resolution is 12 days, then the first remote sensing image corresponding to the target region every 12 days in the year is obtained, and then the remote sensing image sequence finally obtained should include 30 first remote sensing images.

[0073] Optionally, the first remote sensing image can be a radar image observed by a Sentinel-1 satellite.

[0074] It should be noted that the Sentinel-1 is an active microwave remote sensing satellite, which carries a C-band synthetic aperture radar sensor and can observe the ground and perform radar imaging at all times and in all weather. The Sentinel-1 satellite is an Earth observation satellite of the Copernicus program of the European Space Agency (ESA), which is composed of two satellites A and B. The revisit period of a single satellite is 12 days, and the revisit period of the two satellites is 6 days. The Sentinel-1 has four imaging modes: strip imaging mode (SM), interferometric wide swath mode (IW), extra wide swath mode (EW), and wave mode (WV).

[0075] In the embodiment of the present application, the ground range detected (GRD) product of the Sentinel-1A IW imaging mode is used, the revisit period is 12 days (the time resolution is 12 days), the spatial resolution is 10 meters, and there are two polarization modes VV and VH (where VV refers to a same direction polarization and VH refers to a cross polarization), which can realize high-frequency monitoring of changes in surface water bodies.

[0076] Optionally, after the sequence of remote sensing images is obtained, a first surface water body vector corresponding to each first remote sensing image in the sequence of remote sensing images can be obtained based on the first remote sensing image.

[0077] Optionally, after the first surface water body vector corresponding to each first remote sensing image is obtained, for each first surface water body vector, a second surface water body vector corresponding to the first remote sensing image can be obtained based on the first surface water body vector and the first remote sensing image corresponding to the first surface water body vector by using a region growing algorithm.

[0078] It should be noted that in the process of extracting the surface water body vector, the prior art is limited by the selection of sample points, resulting in deviation of the extracted surface water body contour from the actual situation. In the embodiment of the present application, the second surface water body vector is extracted based on the first surface water body vector and the first remote sensing image by using a region growing algorithm. Compared with the image gray threshold segmentation technology, the region growing considers the connectivity of the image pixel space, which can improve the extraction accuracy of the surface water body vector.

[0079] Optionally, after obtaining the second surface water body vectors corresponding to each first remote sensing image respectively, the mudflat vectors corresponding to each first remote sensing image respectively can be obtained by performing topological relationship operation on the first surface water body vectors and the second surface water body vectors corresponding to each first remote sensing image respectively.

[0080] Optionally, after obtaining the mudflat vectors corresponding to each first remote sensing image respectively, the mudflat change information corresponding to the target region can be determined based on the obtained mudflat vectors of the target region in different periods, so as to realize dynamic change monitoring of the mudflat of the target region.

[0081] The mudflat change information extraction method provided by the application extracts the mudflat change information by obtaining the remote sensing image sequence collected in the target period of the target region with the target time resolution, then obtaining a plurality of first surface water body vectors based on each first remote sensing image in the remote sensing image sequence, and then obtaining a plurality of second surface water body vectors based on each first surface water body vector and the first remote sensing image corresponding to each first surface water body vector by using a region growing algorithm, further performing topological relationship operation on the plurality of first surface water body vectors and the plurality of second surface water body vectors to obtain a plurality of mudflat vectors corresponding to the target region, and finally determining the mudflat change information corresponding to the target region based on the plurality of mudflat vectors. Since the change of the tide level of the target region is considered when selecting the remote sensing image, the remote sensing image sequence collected in the target period of the target region with the target time resolution is selected for extracting the mudflat change information, and the second surface water body vector is extracted by using the region growing algorithm based on the extracted first surface water body vector and the first remote sensing image, thereby improving the extraction accuracy of the surface water body vector. Then, the topological relationship operation is performed on the first surface water body vector and the second surface water body vector to obtain a plurality of mudflat vectors corresponding to the target region, and finally the mudflat change information corresponding to the target region is determined based on the plurality of mudflat vectors, so as to realize high spatio-temporal accuracy extraction of the mudflat change information of the target region.

[0082] Optionally, the first surface water body vector corresponding to each first remote sensing image is obtained based on each first remote sensing image in the remote sensing image sequence, comprising:

[0083] The polarization band, the terrain feature and the texture feature corresponding to each first remote sensing image are obtained based on each first remote sensing image respectively;

[0084] A multi-dimensional feature space corresponding to each first remote sensing image is constructed based on the polarization band, the terrain feature and the texture feature corresponding to each first remote sensing image respectively;

[0085] The first surface water body vector corresponding to each first remote sensing image is obtained based on each multi-dimensional feature space.

[0086] Specifically, in the embodiments of the present application, in order to obtain the first surface water body vector corresponding to each first remote sensing image, the polarized wave band, the terrain feature and the texture feature corresponding to each first remote sensing image can be obtained based on each first remote sensing image respectively, and then the multi-dimensional feature space corresponding to each first remote sensing image can be constructed based on the polarized wave band, the terrain feature and the texture feature corresponding to each first remote sensing image respectively, and then the first surface water body vector corresponding to each first remote sensing image can be obtained based on each multi-dimensional feature space respectively.

[0087] It should be noted that due to the influence of the radar side-looking imaging mode, the electromagnetic wave cannot reach the back of the mountain, thereby forming a mountain shadow on the SAR image (the first remote sensing image). Because the electromagnetic wave is specularly reflected on the water surface, the echo power received by the water surface is extremely small, so that the backscattering intensity of the water body is similar to that of the mountain shadow, and it is difficult to separate the water body and the mountain shadow by using single polarized wave band information. However, the position of the mountain shadow has obvious terrain features, and the slope is an index commonly used to represent the terrain features. For example, the surface of the water body is often calm, and the slope is 0, while the place where the mountain shadow appears has a large slope value.

[0088] Therefore, in the embodiments of the present application, in order to eliminate the influence of the mountain shadow on the extraction of the water body, the digital elevation model (Digital Elevation Model, DEM) and the slope and other terrain features corresponding to the SAR image are introduced.

[0089] Furthermore, because the paddy fields and roads and other surface features in the plain area show weak backscattering characteristics on the SAR image, it is difficult to distinguish them from the water body in terms of gray value, and finally the accuracy of water body extraction will be affected. However, in general, as farmland, the paddy fields are mostly regular grid-shaped, and the surface shows obvious regularity and roughness, while the surface of the water body is usually smooth and the boundary is irregular. Texture is a basic feature of an image, and can express the surface and structure properties of a surface feature, such as the roughness of the surface and the uniformity of the structure.

[0090] Therefore, in the embodiments of the present application, in order to more intuitively distinguish different surface features such as the water body, the paddy field and the road in the SAR image, three texture features are introduced, which are:

[0091] Angular Second Moment (ASM), which reflects the degree of uniformity of the image gray scale distribution and the texture roughness, and the value is larger when the local uniformity is higher;

[0092] Entropy (ENT), which reflects the information amount of the image and the randomness of the pixel value, and the entropy value is larger when the local change is larger;

[0093] Homogeneity (HOM) reflects the homogeneity of image texture, the more uniform the local of different areas, the greater the value, so the water body has greater homogeneity.

[0094] Therefore, based on the polarization band, terrain features and texture features corresponding to the first remote sensing image, the multi-dimensional feature space constructed includes the VV polarization band, the VH polarization band, the DEM, the slope, the aspect, the angular second moment, the entropy and the homogeneity.

[0095] Optionally, after constructing the multi-dimensional feature space corresponding to each first remote sensing image, the first surface water body vector corresponding to each first remote sensing image can be obtained based on each multi-dimensional feature space respectively.

[0096] It can be understood that, by constructing a multi-dimensional feature space and then obtaining the first surface water body vector corresponding to the first remote sensing image based on the constructed multi-dimensional feature space, the present embodiment can eliminate the interference of water fields and mountain shadows and the like, and improve the extraction accuracy of surface water bodies in complex terrain and cloudy and rainy environments.

[0097] Optionally, the obtaining of the first surface water body vector corresponding to each first remote sensing image based on each multi-dimensional feature space respectively comprises:

[0098] Each multi-dimensional feature space is input into a support vector machine (SVM) classification model to obtain the first surface water body vector corresponding to each first remote sensing image output by the support vector machine classification model.

[0099] Specifically, in the present embodiment, each multi-dimensional feature space can be classified based on a support vector machine classification model to obtain the first surface water body vector corresponding to each first remote sensing image respectively.

[0100] It should be noted that, since the support vector machine belongs to a supervised classification method, and the supervised classification is based on pixels, some small polygons (noise) will inevitably be produced in the classification result, therefore, from the perspective of practical application, these small polygons need to be removed or reclassified.

[0101] Optionally, after inputting each multi-dimensional feature space into a support vector machine classification model to obtain the first surface water body vector corresponding to each first remote sensing image output by the support vector machine classification model, the method further comprises:

[0102] The first surface water body vector is denoised to obtain a denoised first surface water body vector.

[0103] Optionally, in the embodiment of the present application, the classification result output by the support vector machine classification model can be processed by using methods including but not limited to Majority / Minority analysis, clustering processing and filtering processing, etc., to remove salt and pepper noise, and then the first surface water body vector corresponding to each first remote sensing image is obtained.

[0104] It should be noted that the Majority analysis can use a method similar to convolution filtering to classify false pixels in a larger class into the class, and a transform kernel size is defined first, and the pixel class occupying the main position (the most pixels) in the transform kernel is used to replace the class of the center pixel. For the Minority analysis, the class of the pixel occupying the secondary position in the transform kernel is used to replace the class of the center pixel.

[0105] Preferably, in the embodiment of the present application, the first surface water body vector corresponding to each first remote sensing image output by the support vector machine classification model is denoised by using the Majority analysis method, and the denoised first surface water body vector is obtained.

[0106] It can be understood that, by denoising the first surface water body vector, the embodiment of the present application is beneficial to more accurately obtaining the mudflat vector corresponding to the target region by using the denoised first surface water body vector subsequently.

[0107] Optionally, before the step of inputting each multi-dimensional feature space into the support vector machine classification model to obtain the first surface water body vector corresponding to each first remote sensing image output by the support vector machine classification model, the method further comprises:

[0108] determining a segmentation threshold for segmenting water body and non-water body;

[0109] based on the segmentation threshold, performing threshold segmentation processing on each sample remote sensing image respectively to obtain target vector data corresponding to each sample remote sensing image respectively, wherein the target vector data includes water body vector data and non-water body vector data;

[0110] selecting a plurality of sample points from each target vector data respectively to obtain a plurality of sample point sets;

[0111] training an initial support vector machine classification model based on the plurality of sample point sets and a cross-validation method to obtain the support vector machine classification model trained.

[0112] Specifically, in the embodiment of the present application, before classifying each multi-dimensional feature space by using the support vector machine classification model, it is necessary to first construct and train the support vector machine classification model. Specifically, first, a segmentation threshold for segmenting water bodies and non-water bodies is determined, then based on the segmentation threshold, the threshold segmentation processing is performed on each sample remote sensing image respectively to obtain target vector data corresponding to each sample remote sensing image respectively, the target vector data includes water body vector data and non-water body vector data, then a plurality of sample point sets are obtained by selecting a plurality of sample points from each target vector data respectively, and finally, the initial support vector machine classification model is trained based on the obtained plurality of sample point sets and the cross-validation method, and the trained support vector machine classification model is finally obtained.

[0113] It should be noted that SVM is an optimal boundary classification method based on Vapnik-Chervonenkis Dimension (VC) theory and structural risk minimization criterion, which is a maximum interval classifier defined in a feature space. The purpose of SVM is to find the compromise value of minimizing empirical risk and confidence interval, which considers both training error and model complexity, and can obtain good classification effect when the number of samples is small. That is, SVM finds the optimal segmentation hyperplane by maximizing the interval of the data set through support vectors.

[0114] Figure 2 is the optimal classification line of the support vector machine provided by the present application, as shown in Figure 2 which shows the optimal classification line in linearly separable samples, H represents the maximum classification interval between data sets, which is called the optimal hyperplane, H1 and H2 are two mutually parallel hyperplanes on both sides of the optimal hyperplane H, and the maximum classification interval is Y=+1 and Y=-1 are classification labels.

[0115] It should be noted that the traditional SVM classification method extracts water bodies mainly based on the band data of SAR images to construct a feature space, and trains a model by selecting sample points, thereby realizing the extraction of water bodies. However, due to the complex scene of SAR images, it is difficult to accurately describe the difference between water bodies and non-water bodies by using only one type of feature.

[0116] Therefore, the embodiment of the present application considers constructing the feature space of the sample by combining multiple auxiliary data, mapping the non-linear sample to a high-dimensional space, constructing the optimal classification hyperplane, and solving the non-linear problem.

[0117] It should be noted that for non-linearly separable samples, a kernel function needs to be introduced. The kernel function can map the sample data in the low-dimensional space to the high-dimensional space, so that the sample is linearly separable, and the calculation formula is as follows:

[0118]

[0119] where K(X,Z) represents a kernel function, is a mapping function, is and inner product of represents a nonlinear feature mapping process from the input space to a high-dimensional space.

[0120] Therefore, after adding the constraint condition, the process of finding the optimal classification hyperplane can be regarded as a process of solving a convex quadratic programming problem. By solving the problem, the optimal classification function is obtained, and the expression is as follows:

[0121]

[0122] where, and b * are parameters determined only by support vectors, which can be used to regulate the optimal classification plane, l represents a dimension, and sign represents a sign function.

[0123] Optionally, the sample remote sensing image in the embodiment of the present application can be a 10-meter spatial resolution European Space Agency (ESA) land cover type product, two types of grid data of cultivated land and permanent water body can be extracted based on the product, and then the two types of grid data are converted into vector data, 1000 points and 500 points are randomly generated in the vector range, and the positions of the random points are regarded as samples of cultivated land and water body. Combined with the optical image of Google Earth (GE), 500 sample points of mountains are selected, and finally three sample point sets including water body, cultivated land and mountain shadow are constructed, wherein the three sample point sets include 500 points, 1000 points and 500 points respectively.

[0124] It should be noted that for the SVM classification model, the penalty coefficient C and the Gamma coefficient are two important parameters. C is called a penalty factor (or a penalty coefficient), and if the value is larger, the tolerance to incorrect classification is lower, and overfitting phenomenon (regarding non-water body pixels as water body) may occur; otherwise, the tolerance to incorrect classification is higher, thereby reducing the accuracy of the classification result. When running supervised classification (SVM) on an ENVI (Environment for Visualizing Images) or other remote sensing image processing platform, the C and Gamma parameters are default values. However, under the classification model corresponding to the default parameters, the accuracy of the final classification result is often not the highest. In order to improve the accuracy of the classification result, the embodiments of the present application determine the optimal C and Gamma parameters through cross-validation method.

[0125] It should be noted that the cross-validation method essentially adopts a grid search algorithm, that is, within a certain range of parameters, different parameters are arranged and combined according to a specified step size, each set of parameter combinations is tested, and the set of parameters with the optimal performance index is taken as the value of the final parameter.

[0126] Specifically, the process of determining the optimal C and Gamma parameters by the cross-validation method includes: combining the C and Gamma parameters in a certain value range according to a certain step size, dividing the training data set into k groups under different C and Gamma combinations, one group is used as validation data (to verify the accuracy of the model classification), and the remaining k-1 groups are used as training data. In each group (C, Gamma) combination, each group of data is used as a validation data sample in turn, so k times of calculation are required, and the average classification accuracy obtained by the k times of cross-validation is taken as the average classification accuracy under the (C, Gamma) model. The average classification accuracy of the model under different (C, Gamma) combinations is obtained by the above calculation, and the (C, Gamma) combination with the highest average classification accuracy is taken.

[0127] It should be noted that when different C and Gamma parameter combinations have the same average classification accuracy, the combination with the smaller C value can be taken, because in the case of ensuring the classification accuracy of the model, the smaller the C value, the greater the fault tolerance of the classification model, thereby avoiding the overfitting phenomenon. If the C value is also the same, the first (C, Gamma) combination is taken.

[0128] Figure 3 The classification accuracy contour line determined according to C and Gamma provided by the application is shown in FIG. 1, which shows a contour line composed of C, Gamma coefficient and average classification accuracy, wherein the horizontal coordinate represents different values of C, the vertical coordinate represents different values of the Gamma coefficient, and the numbers on different contour lines represent the average classification accuracy, for example, from FIG. 1, the value range of C and Gamma corresponding to the highest average classification accuracy (82%) can be determined. Figure 3 Figure 3

[0129] Optionally, the second surface water body vector corresponding to each of the first remote sensing images is obtained by using a region growing algorithm based on each of the first surface water body vectors and the first remote sensing image corresponding to each of the first surface water body vectors, and the region growing algorithm includes:

[0130] The target remote sensing image corresponding to the VH polarization band in the polarization band corresponding to each of the first remote sensing images is determined respectively.

[0131] ​​respectively as a mask, extract pixels in each of the target remote sensing images as a seed point set, and respectively determine a threshold range meeting a growth condition corresponding to each of the seed point sets based on each of the seed point sets;

[0132] respectively based on each of the target remote sensing images, each of the seed point sets and each of the threshold ranges, perform region growing, and in a case where each of the seed point sets is determined to be empty, obtain second surface water body vectors corresponding to each of the first remote sensing images respectively.

[0133] Specifically, in the embodiments of the present application, in order to obtain second surface water body vectors corresponding to each of the first remote sensing images by using the region growing algorithm, a target remote sensing image corresponding to a VH polarized wave band in the polarized wave bands corresponding to each of the first remote sensing images can be determined first, then each of the first surface water body vectors is respectively taken as a mask, pixels in each of the target remote sensing images are extracted as a seed point set, and a threshold range meeting a growth condition corresponding to each of the seed point sets is respectively determined based on each of the seed point sets, and then region growing is performed based on each of the seed point sets and each of the threshold ranges, and in a case where each of the seed point sets is determined to be empty, second surface water body vectors corresponding to each of the first remote sensing images are obtained respectively.

[0134] It should be noted that the basic idea of the region growing algorithm is: one or a group of pixels are selected as seed points; the characteristics and similarity decision criteria are determined; starting from the first seed point, growing outward, first judging whether the pixels in the seed neighborhood meet the similarity condition, and if so, merging with the seed into the same region, and then taking the merged pixels as growth points and using a similar method for growth; repeat the above operation until there are no more pixels meeting the condition to be merged into the region.

[0135] Therefore, before performing region growing, the input image y i (i = 1, 2, 3,..., n) of the region growing algorithm is determined first, the seed point set x j (j = 1, 2, 3,..., m), x j ∈ y i , where m < n, the threshold range I meeting the growth condition is [P min , P max ], P min represents the minimum pixel value, and P max represents the maximum pixel value; and then the region growing process is performed.

[0136] The flow of the region growing algorithm includes: starting from the first seed point x1, judging whether the pixel values in the eight neighborhoods of x1 meet the growth condition I. If there is a pixel x m+1 in the eight neighborhoods of x1 meeting the growth condition I, then the pixel xm+1 merge into the seed point set x j x1 is removed from x j Meanwhile, the pixel at the coordinate of x m+1 is set to 1, and the next seed point x2 is judged until x j ∈φ, and finally the binary-valued raster data is obtained.

[0137] It should be noted that under different polarization modes, the echo power of the SAR received by the beach and the water body is different, thereby causing the two ground objects to have different backscattering intensities on different polarization images. In order to select the input image required for region growing, the present embodiment selects a plurality of pixels on the VV and VH polarization bands for analysis for the two ground objects of the lake water body and the beach along the shore, and obtains a gray level histogram of the two different ground objects of the lake shore and the water body on the VV polarization band, which has a more obvious "double peak" effect, and represents that the two ground objects have higher distinguishability under VV polarization, which is helpful to improve the accuracy of water body extraction in SVM classification. On the contrary, the backscattering characteristics of the two ground objects are not much different on the VH polarization image, and the region growing based on the water body pixel set on the VH polarization image will maximize the extraction of the beach along the shore. Therefore, the present embodiment selects the target remote sensing image corresponding to the VH polarization band as the input image for region growing.

[0138] It can be understood that after the classification by the support vector machine classification model and the post-processing of removing the salt and pepper noise, the first ground water body vector corresponding to each first remote sensing image is obtained, which can maximize the representation of the contour of the ground water body and eliminate the interference of the farmland and the mountain shadow. The first ground water body vector can be used as a mask to extract the pixels on the input image (the target remote sensing image corresponding to the VH polarization band) as a seed point set.

[0139] It should be noted that the specific determination method of the threshold range satisfying the growth condition is described below.

[0140] Optionally, region growing can be performed based on each target remote sensing image, each seed point set and each threshold range, and a second ground water body vector corresponding to each first remote sensing image is obtained in a case where each seed point set is determined to be empty.

[0141] Optionally, the determination of the threshold range satisfying the growth condition corresponding to each seed point set based on each seed point set comprises:

[0142] determining the mean value and the standard deviation value of each seed point in each seed point set, respectively;

[0143] The threshold range satisfying the growth condition corresponding to each seed point set is determined based on the mean value and the standard deviation value corresponding to each seed point set respectively.

[0144] Specifically, in the embodiments of the present application, in order to determine the threshold range satisfying the growth condition, the threshold range satisfying the growth condition corresponding to each seed point set can be determined based on the mean value and the standard deviation value corresponding to each seed point set.

[0145] The mean value of the seed point set is calculated by the following formula:

[0146]

[0147] The standard deviation value of the seed point set is calculated by the following formula:

[0148]

[0149] Optionally, in the embodiments of the present application, the threshold range satisfying the growth condition can be [mean-3×std, mean+3×std].

[0150] Specifically, the first seed point in the seed point set can be taken as a starting point, and it is determined whether the value of the eight-neighborhood pixel of the seed point meets the threshold range. If no pixel meets the threshold range, the neighborhood pixel of the next seed point is determined. If there is a pixel meeting the condition, the neighborhood pixel is added to the seed point set as a new seed point. In this way, the seed point set is empty.

[0151] Optionally, the polarized wave band, the terrain feature and the texture feature corresponding to each first remote sensing image are obtained based on the first remote sensing image respectively, and the method comprises the following steps:

[0152] The first remote sensing image is preprocessed to obtain the polarized wave band corresponding to the first remote sensing image and the second remote sensing image corresponding to the polarized wave band;

[0153] The terrain feature and the texture feature corresponding to each first remote sensing image are obtained based on the second remote sensing image;

[0154] The preprocessing comprises thermal noise removal, orbit correction, radiation scaling, speckle filtering, terrain correction and decibel processing.

[0155] Specifically, in the embodiments of the present application, in order to obtain the polarized wave band, the terrain feature and the texture feature corresponding to each first remote sensing image based on each first remote sensing image, the first remote sensing image can be preprocessed first to obtain the polarized wave band corresponding to each first remote sensing image and the second remote sensing image corresponding to the polarized wave band, and then based on the second remote sensing image, the terrain feature and the texture feature corresponding to each first remote sensing image are obtained, wherein the preprocessing of each first remote sensing image includes thermal noise removal, orbit correction, radiation calibration, speckle filtering, terrain correction and decibel processing.

[0156] For example, Figure 4 is a flowchart of the preprocessing of the Sentinel-1A GRD image provided by the present application, as Figure 4 shown, the Sentinel-1A GRD image is sequentially subjected to thermal noise removal, orbit correction, radiation calibration, speckle filtering, terrain correction and decibel processing, and finally the polarized wave band, the terrain feature and the texture feature corresponding to the Sentinel-1A GRD image are obtained.

[0157] It should be noted that thermal noise is caused by the backscattering energy of the SAR receiver, although thermal noise rarely appears in SAR images, but for calm water, rivers and areas with low backscattering values, thermal noise will have a certain influence on the quality of SAR images. Therefore, the thermal noise removal processing of the first remote sensing image in the embodiments of the present application is beneficial to the subsequent effective extraction of the surface water body vector.

[0158] It should be noted that due to the influence of external conditions and internal factors, the flight attitude of the satellite is slightly dithered or deviated, which will cause a certain system error. Therefore, in order to reduce the system error caused by the orbit motion, the orbit is accurately calibrated by using the precise orbit information data, so as to obtain accurate orbit position and speed information.

[0159] It should be noted that since the radar system does not carry out radiation calibration processing on the Level-1 data, the SAR intensity image needs to be converted into a backscattering coefficient image before using the data. Radiation calibration is to generate the radar backscattering coefficient value corresponding to the ground target, so as to facilitate the quantitative analysis of the radar image. In the embodiments of the present application, the image can be calibrated as a Beta-naught product, and the Beta-naught product is taken as the input of the next step.

[0160] It should be noted that the enhancement and weakening of the SAR coherent wave will cause the radar image to form periodic speckle noise, and the influence of the coherent noise is far greater than other noises, and the existence of the speckle noise will reduce the signal-to-noise ratio of the SAR image, and in severe cases, the image features will disappear. In the embodiment of the application, in order to keep as much image detail information as possible and suppress the influence of noise on remote sensing image interpretation, the remote sensing image is subjected to coherent speckle filtering processing.

[0161] Optionally, Lee Sigma filtering can be used to perform coherent speckle filtering processing on the remote sensing image. Lee Sigma filtering is based on the Sigma probability of Gaussian distribution, and image noise is filtered out by averaging the pixels within the filtering window that fall within the two Sigma ranges of the central pixel. The edge retention capability of this algorithm is good, and it can effectively reduce the bright and dark spots caused by water surface fluctuations, and has good noise suppression effect in extracting surface water bodies using SAR images.

[0162] It should be noted that the terrain correction includes geocoding (assigning actual coordinate information to the remote sensing image) and terrain radiation correction processing. In SAR imaging, due to the terrain changes and the incidence angle of the satellite sensor, the distance in the SAR image is distorted, and the SAR image is distorted to a certain extent.

[0163] Therefore, in the embodiment of the application, the distance Doppler algorithm can be used to perform terrain correction processing on the remote sensing image, to reduce the influence of terrain distortion on the remote sensing image, and the projection coordinate system of the output image is set to UTM Zone 49 / World Geodetic System 1984 (WGS84: World Geodetic System 1984, which is a coordinate system established for the Global Positioning System (GPS)).

[0164] It should be noted that after the remote sensing image is subjected to processing including thermal noise removal, orbit correction, radiation calibration, coherent speckle filtering and terrain correction, the linear proportional unit of the backscattering coefficient is obtained, and the value is usually a very small positive value (reason: long transmission distance, and the receiver finally receives very little energy / power).

[0165] Therefore, after the decibel processing in the embodiment of the application, the radar backscattering coefficient range is approximately normally distributed, and the storage bit number of the data is stored as float floating point type data from double type data, which is convenient for visualization and data analysis.

[0166] Optionally, in the embodiment of the present application, the SRTM (Shuttle Radar Topography Mission) 1 Sec 30-meter DEM of the target area can be acquired, and the resolution of the DEM is resampled to 10 meters, and then the terrain features are calculated based on the DEM data.

[0167] Optionally, in the embodiment of the present application, the texture feature calculation can be based on second-order probability statistics.

[0168] It should be noted that the second-order probability statistics is to calculate the texture value by using a gray space correlation matrix, which is a relative frequency matrix, i.e., the frequency of the pixel value in two adjacent processing windows (separated by a certain distance and direction), which shows the number of occurrences of the relationship between a pixel and its specific neighborhood. After the decibel processing is completed, two band images of Beta0_VV_db and Beta0_VH_db are output, and the two band images can be quantized to improve the calculation rate.

[0169] For example, the step length of the texture feature calculation is set to 1, the window size is set to 5x5, and the direction is set to 0°, 45°, 90° and 135°, and finally the texture features of the VV and VH different polarization bands are obtained.

[0170] Figure 5 is a flowchart of the beach vector extraction provided by the present application, as shown in Figure 5 After the SAR image of the studied area is acquired, the SAR image is preprocessed to obtain the polarization band, terrain feature and texture feature corresponding to the SAR image, then a multi-dimensional feature space is constructed based on the polarization band, terrain feature and texture feature, and then the multi-dimensional feature space is classified by the SVM classification method, and after the classification result is classified and processed, a first ground water body vector is obtained, further, the first ground water body vector is used as a mask to extract the seed point set in the VH polarization band image, and based on the average value and standard deviation value of the seed points in the seed point set, a threshold condition is determined, and then the region growing is performed, until the seed point set is empty, and then a second ground water body vector is obtained, finally, the first ground water body vector and the second ground water body vector are topologically related to obtain the beach vector.

[0171] It should be noted that in the embodiment of the present application, the topological relationship operation of the first ground water body vector and the second ground water body vector can include: performing intersection analysis on the first ground water body vector and the second ground water body vector, and obtaining the face vector of the growing beach part based on the analysis result. This process can be understood as deleting (erasing) the water body part, thereby obtaining the beach vector.

[0172] It can be understood that in the embodiments of the present application, if the time resolution is 12 days, the SAR images collected every 12 days can be subjected to the extraction of the beach vector according to the beach vector extraction process shown in Figure 5 As shown in the above embodiment, the beach vector of the target region in the target period can be obtained, and the beach change information of the target region can be determined based on the beach vector.

[0173] Alternatively, in the embodiments of the present application, based on the SAR images with a time resolution of 12 days and a spatial resolution of 10 meters, the spatial change characteristics of the surface water body in the target region every 12 days can be obtained by using the support vector machine classification method, and then based on the characteristics that the backscattering characteristics of the surface water body and the coastal beach are not much different under the VH polarization, the region growth through the surface water body pixel set is carried out on the VH polarization image, and then the topological relationship operation is carried out between the vector results after the region growth processing and the surface water body vector obtained initially, so as to finally obtain the lake and reservoir coastal beach vector results, thereby realizing the monitoring of the space-time change of the lake and reservoir coastal beach in the target region.

[0174] The beach change information extraction method provided by the present application comprises the following steps: obtaining a remote sensing image sequence collected in a target period corresponding to a target region with a target time resolution; obtaining a plurality of first surface water body vectors based on each first remote sensing image in the remote sensing image sequence; obtaining a plurality of second surface water body vectors based on each first surface water body vector and the first remote sensing image corresponding to the first surface water body vector by using a region growth algorithm; performing topological relationship operation on the plurality of first surface water body vectors and the plurality of second surface water body vectors to obtain a plurality of beach vectors corresponding to the target region; and determining the beach change information of the target region based on the plurality of beach vectors.

[0175] The beach change information extraction device provided by the present application will be described below, and the beach change information extraction device described below can be correspondingly referred to the beach change information extraction method described above.

[0176] Figure 6is a structural schematic diagram of the beach change information extraction device provided by the present application, as shown in the figure, the device comprises: an acquisition module 610, a first obtaining module 620, a second obtaining module 630, a third obtaining module 640 and a determination module 650; wherein: Figure 6

[0177] The acquisition module 610 is used for acquiring a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target area;

[0178] The first obtaining module 620 is used for obtaining a first surface water body vector corresponding to each first remote sensing image in the remote sensing image sequence respectively based on the first remote sensing image;

[0179] The second obtaining module 630 is used for obtaining a second surface water body vector corresponding to each first remote sensing image respectively based on each first surface water body vector and the first remote sensing image corresponding to each first surface water body vector by using a region growing algorithm;

[0180] The third obtaining module 640 is used for performing a topological relationship operation on the first surface water body vector and the second surface water body vector corresponding to each first remote sensing image respectively to obtain a beach vector corresponding to each first remote sensing image;

[0181] The determination module 650 is used for determining beach change information corresponding to the target area based on each beach vector.

[0182] ​The beach change information extraction device provided by the present application can realize all the method steps achieved by the beach change information extraction method embodiment, and can achieve the same technical effects, and the same parts and beneficial effects in the method embodiment will not be described in detail.

[0183] It should be noted that the beach change information extraction device provided by the embodiment of the present application can realize all the method steps achieved by the beach change information extraction method embodiment, and can achieve the same technical effects, and the same parts and beneficial effects in the method embodiment will not be described in detail.

[0184] Figure 7 The present application provides an electronic device entity structure schematic diagram, as shown in Figure 7 The electronic device can include a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communications bus 740. The processor 710 can call the logic instructions in the memory 730 to execute the beach change information extraction method provided by each method, which includes:

[0185] Obtain a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target area;

[0186] Obtain a first ground water body vector corresponding to each first remote sensing image in the remote sensing image sequence based on each first remote sensing image;

[0187] respectively based on each of the first surface water body vectors and the first remote sensing image corresponding to each of the first surface water body vectors, a second surface water body vector corresponding to each of the first remote sensing images is obtained by using a region growing algorithm;

[0188] topological relationship operations are respectively performed on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images to obtain a tidal flat vector corresponding to each of the first remote sensing images;

[0189] Based on each of the tidal flat vectors, the tidal flat change information corresponding to the target region is determined.

[0190] In addition, the logical instructions in the memory 730 described above can be implemented in the form of a software function unit and sold or used as a separate product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part of the prior art that contributes essentially or the part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0191] On the other hand, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the computer can execute the tidal flat change information extraction method provided by the above-mentioned method, which comprises:

[0192] Obtaining a remote sensing image sequence collected at a target time resolution in a target period corresponding to a target region;

[0193] Based on each of the first remote sensing images in the remote sensing image sequence, a first surface water body vector corresponding to each of the first remote sensing images is obtained;

[0194] respectively based on each of the first surface water body vectors and the first remote sensing image corresponding to each of the first surface water body vectors, a second surface water body vector corresponding to each of the first remote sensing images is obtained by using a region growing algorithm;

[0195] perform topological relation operation on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images respectively, to obtain a tidal flat vector corresponding to each of the first remote sensing images respectively;

[0196] Based on each of the tidal flat vectors, determine the tidal flat change information corresponding to the target region.

[0197] In another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned tidal flat change information extraction method, which comprises:

[0198] Obtain a remote sensing image sequence of a target region corresponding to a target period, which is collected at a target time resolution;

[0199] Based on each of the first remote sensing images in the remote sensing image sequence, obtain a first surface water body vector corresponding to each of the first remote sensing images respectively;

[0200] Based on each of the first surface water body vectors and the first remote sensing image corresponding to each of the first surface water body vectors, use a region growing algorithm to obtain a second surface water body vector corresponding to each of the first remote sensing images respectively;

[0201] Perform topological relation operation on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images respectively, to obtain a tidal flat vector corresponding to each of the first remote sensing images respectively;

[0202] Based on each of the tidal flat vectors, determine the tidal flat change information corresponding to the target region.

[0203] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme. Those skilled in the art can understand and implement it without creative labor.

[0204] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0205] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for extracting information on tidal flat changes, characterized in that, include: Acquire the remote sensing image sequence of the target area within the target period at the target time resolution; Based on each first remote sensing image in the remote sensing image sequence, obtain the first surface water body vector corresponding to each first remote sensing image; Based on each of the first surface water body vectors and the first remote sensing images corresponding to each of the first surface water body vectors, a region growing algorithm is used to obtain the second surface water body vectors corresponding to each of the first remote sensing images. Topological relationship operations are performed on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images to obtain the tidal flat vector corresponding to each of the first remote sensing images. Based on the aforementioned tidal flat vectors, the tidal flat change information corresponding to the target area is determined.

2. The method for extracting information on tidal flat changes according to claim 1, characterized in that, The step of obtaining the first surface water body vector corresponding to each of the first remote sensing images in the remote sensing image sequence includes: Based on each of the first remote sensing images, obtain the polarization band, terrain features and texture features corresponding to each of the first remote sensing images respectively; Based on the polarization band, topographic features, and texture features corresponding to each of the first remote sensing images, a multidimensional feature space corresponding to each of the first remote sensing images is constructed. Based on each of the multidimensional feature spaces, a first surface water vector corresponding to each of the first remote sensing images is obtained.

3. The method for extracting information on tidal flat changes according to claim 2, characterized in that, The step of obtaining the first surface water body vector corresponding to each of the first remote sensing images based on each of the multidimensional feature spaces includes: Each of the multidimensional feature spaces is input into a support vector machine classification model to obtain the first surface water body vector output by the support vector machine classification model, which corresponds to each of the first remote sensing images.

4. The method for extracting information on tidal flat changes according to claim 3, characterized in that, After inputting each of the multidimensional feature spaces into a support vector machine classification model to obtain the first surface water body vectors output by the support vector machine classification model corresponding to each of the first remote sensing images, the method further includes: The first surface water vector is denoised to obtain the denoised first surface water vector.

5. The method for extracting information on tidal flat changes according to claim 3, characterized in that, Before inputting each of the multidimensional feature spaces into the support vector machine classification model to obtain the first surface water body vectors output by the support vector machine classification model corresponding to each of the first remote sensing images, the method further includes: Determine the segmentation threshold used to separate water bodies from non-water bodies; Based on the segmentation threshold, threshold segmentation processing is performed on each sample remote sensing image to obtain target vector data corresponding to each sample remote sensing image. The target vector data includes water body vector data and non-water body vector data. Several sample points are selected from each of the target vector data to obtain multiple sample point sets; The initial support vector machine classification model is trained based on the multiple sample point sets and cross-validation method to obtain the trained support vector machine classification model.

6. The method for extracting tidal flat change information according to any one of claims 2-5, characterized in that, The step of obtaining second surface water vectors corresponding to each of the first remote sensing images based on each of the first surface water vectors and the first remote sensing images corresponding to each of the first surface water vectors, using a region growing algorithm, includes: The target remote sensing images corresponding to the VH polarization band in the polarization band of each of the first remote sensing images are determined respectively. Each of the first surface water body vectors is used as a mask to extract the pixels in each of the target remote sensing images as seed point sets, and based on each of the seed point sets, a threshold range that satisfies the growth condition is determined for each of the seed point sets. Based on each of the target remote sensing images, each of the seed point sets, and each of the threshold ranges, a region growing operation is performed, and if each of the seed point sets is determined to be empty, a second surface water vector corresponding to each of the first remote sensing images is obtained.

7. The method for extracting information on tidal flat changes according to claim 6, characterized in that, The step of determining the threshold range that satisfies the growth condition for each seed point set, based on each of the seed point sets, includes: Determine the mean and standard deviation of the seed points in each of the seed point sets; Based on the average value and the standard deviation corresponding to each of the seed point sets, the threshold range that satisfies the growth condition for each of the seed point sets is determined.

8. The method for extracting information on tidal flat changes according to claim 2, characterized in that, The step of obtaining the polarization band, terrain features, and texture features corresponding to each of the first remote sensing images, respectively, includes: Each of the first remote sensing images is preprocessed to obtain the polarization bands corresponding to each of the first remote sensing images, and the second remote sensing images corresponding to the polarization bands. Based on the second remote sensing image, the terrain features and texture features corresponding to each of the first remote sensing images are obtained; The preprocessing includes thermal noise removal, orbit correction, radiometric calibration, speckle filtering, terrain correction, and decibel reduction.

9. A device for extracting information on tidal flat changes, characterized in that, include: The acquisition module is used to acquire remote sensing image sequences acquired at the target time resolution within the target period corresponding to the target area; The first acquisition module is used to obtain a first surface water body vector corresponding to each first remote sensing image in the remote sensing image sequence. The second acquisition module is used to obtain, based on each of the first surface water body vectors and the first remote sensing images corresponding to each of the first surface water body vectors, a region growing algorithm is used to obtain the second surface water body vectors corresponding to each of the first remote sensing images. The third obtaining module is used to perform topological relationship calculations on the first surface water body vector and the second surface water body vector corresponding to each of the first remote sensing images, respectively, to obtain the tidal flat vector corresponding to each of the first remote sensing images; The determination module is used to determine the tidal flat change information corresponding to the target area based on each of the tidal flat vectors.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for extracting tidal flat change information as described in any one of claims 1 to 8.

Images