Urban impervious surface extraction method based on SAR image and visible light image
By processing visible light and SAR images, constructing a polarization prediction model and performing information fusion, the problems of low resolution and poor accuracy caused by noise interference in SAR images are solved, and high-precision extraction of urban impervious surfaces is achieved.
Patent Information
- Application Number
- CN202410977365.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-07-22
AI Technical Summary
SAR images are susceptible to speckle noise, which can cause drastic changes in grayscale values and result in low spatial resolution, affecting the accuracy of urban impervious surface extraction.
By acquiring visible light and SAR images from the same time period, performing resolution downscaling and aggregation averaging calculations, a SAR image VV polarization prediction model is constructed. High-resolution SAR image polarization information is obtained using convolutional neural networks and the XGBoost algorithm. Threshold segmentation is performed by combining the urban building index and the OTU method, and finally, the impermeable surface extraction result is comprehensively judged.
It improved the spatial resolution of SAR images and the accuracy of extracting impermeable surfaces in cities, thus enhancing image quality and extraction accuracy.
Smart Images

Figure CN118506207B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of urban impervious surface extraction, and in particular to an urban impervious surface extraction method based on SAR images and visible light images. Background Art
[0002] SAR data is not affected by weather conditions such as clouds and rain, and can be observed all day and all weather. Currently, SAR satellites such as Lacrosse, ERS, JERS, Radarsat, SRTM, TerraSAR, Cosmo-Skymed, Gaofen-3, ALOS, Sentinel-1, etc. are being used more and more widely, forming a multi-band, multi-polarization, multi-view, and multi-mode spaceborne SAR system, which provides a large amount of data for various remote sensing applications. However, SAR is easily interfered by coherent speckle noise, which causes drastic changes in the grayscale value of SAR images, thereby reducing the quality of SAR images. Therefore, compared with visible light images, SAR images have relatively low spatial resolution and relatively low accuracy in extracting urban impervious surfaces. Summary of the invention
[0003] In order to solve the above technical problems, the present application provides a method for extracting urban impervious surfaces based on SAR images and visible light images over a large range with medium and low spatial resolution (such as above 10m), which can effectively improve the spatial resolution of SAR images and improve the extraction accuracy of urban impervious surfaces.
[0004] The present application provides a method for extracting urban impervious surfaces based on SAR images and visible light images, comprising the following steps:
[0005] Acquire a first visible light image and a first SAR image of the target city in the same time period; wherein the spatial resolution of the visible light image and the SAR image is greater than 10m;
[0006] Performing aggregate average calculation on the first visible light image to obtain a second visible light image; wherein the spatial resolution of the second visible light image is lower than that of the first visible light image;
[0007] Constructing a SAR image VV polarization prediction model, inputting the second visible light image into the SAR image VV polarization prediction model, and obtaining second SAR image VV polarization prediction information;
[0008] Inputting the first visible light image into the SAR image VV polarization prediction model to obtain first SAR image VV polarization prediction information;
[0009] Based on the first SAR image, the first SAR image, the VV polarization prediction information of the first SAR image, and the VV polarization prediction information of the second SAR image are input into a preset second SAR image generation formula to perform information fusion calculation to obtain a second SAR image;
[0010] The first visible light image and the second SAR image are respectively subjected to the urban building index NDBI calculation to obtain the first pixel value and the second pixel value, and the first visible light image and the second SAR image are respectively subjected to the threshold segmentation using the OSTU method to obtain the first segmentation threshold σ vis and the second segmentation threshold σ SAR ;
[0011] The first pixel value and the first segmentation threshold σ vis Performing comparison to obtain a first comparison value;
[0012] The second pixel value and the second segmentation threshold σ SAR Performing comparison to obtain a second comparison value;
[0013] The first comparison value and the second comparison value are combined to determine the extraction result of the urban impervious surface.
[0014] Compared with the prior art, the present invention first obtains the first visible light image and the first SAR image of the target city in the same time period, obtains the second visible light image by reducing the resolution of the first visible light image, obtains the pixel spatial feature information of the high-resolution and low-resolution visible light images respectively, and then performs machine learning calculation to obtain the polarization prediction information of the SAR image, combines the first SAR image with the SAR image polarization prediction information, and generates a high-resolution second SAR image. The first visible light image and the second SAR image are threshold-compared and then combined, and finally the extraction result of the urban impervious surface image is obtained. The SAR image quality within the target city is improved, and the accuracy of urban impervious surface extraction is improved by threshold-comparing the SAR image and the visible light image respectively, and then combining them for comprehensive judgment. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0016] Figure 1A flowchart of a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application;
[0017] Figure 2 A flowchart of the steps of obtaining VV polarization prediction information of a second SAR image in a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application;
[0018] Figure 3 A flowchart of the steps of obtaining a second SAR image in a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application;
[0019] Figure 4 A flowchart of the steps of determining the urban impervious surface image extraction result of a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application;
[0020] Figure 5 A technical flow chart of a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application. DETAILED DESCRIPTION
[0021] The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention. The flowchart used in this application shows the operations implemented in some embodiments of the embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical contextual relationships can be reversed in order or implemented simultaneously. In addition, under the guidance of the content of this application, those skilled in the art can add at least one other operation to the flowchart, or remove at least one operation from the flowchart.
[0022] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.
[0023] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly precise meanings unless specifically defined as herein.
[0024] This application provides a method for extracting urban impervious surfaces based on SAR images and visible light images. Figure 1 , Figure 1 A flowchart of a method for extracting urban impervious surfaces based on SAR images and visible light images is provided in an embodiment of the present application. The method comprises the following steps:
[0025] S101. Acquire a first visible light image and a first SAR image of a target city in the same time period; wherein the spatial resolution of the visible light image and the SAR image is greater than 10 m.
[0026] S102 . Perform aggregation and averaging calculation on the first visible light image to obtain a second visible light image; wherein the spatial resolution of the second visible light image is lower than that of the first visible light image.
[0027] S103: construct a SAR image VV polarization prediction model, input the second visible light image into the SAR image VV polarization prediction model, and obtain second SAR image VV polarization prediction information.
[0028] S104: Input the first visible light image into the SAR image VV polarization prediction model to obtain first SAR image VV polarization prediction information.
[0029] S105. Based on the first SAR image, input the first SAR image, the VV polarization prediction information of the first SAR image, and the VV polarization prediction information of the second SAR image into a preset second SAR image generation formula to perform information fusion calculation to obtain a second SAR image.
[0030] S106, respectively calculating the urban building index NDBI for the first visible light image and the second SAR image to obtain a first pixel value and a second pixel value, and respectively performing threshold segmentation on the first visible light image and the second SAR image using the OSTU method to obtain a first segmentation threshold σ vis and the second segmentation threshold σ SAR .
[0031] S107, comparing the first pixel value with the first segmentation threshold σvis A comparison is performed to obtain a first comparison value.
[0032] S108, comparing the second pixel value with the second segmentation threshold σ SAR A comparison is performed to obtain a second comparison value.
[0033] S109: Combine the first comparison value and the second comparison value to determine the extraction result of the urban impervious surface.
[0034] SAR image, namely Synthetic Aperture Radar image, is a remote sensing image obtained by radar technology, which obtains surface information by actively transmitting microwave signals and receiving signals reflected by ground objects. The processing of SAR images involves multiple key steps, among which radiation calibration is an important link to improve the quality and accuracy of SAR images, mainly including transmitter power calibration, receiver gain calibration, radiation imaging geometry calibration and other processes to eliminate the influence of system and external factors on imaging effects. However, SAR is easily interfered by coherent speckle noise, which causes drastic changes in the grayscale value of SAR images, thereby reducing the quality of SAR images and resulting in relatively low spatial resolution of SAR images compared to visible light images.
[0035] The steps of this application are described in detail below.
[0036] Step S101, obtaining a first visible light image and a first SAR image of a target city in the same time period, includes the following steps:
[0037] S1011, acquiring a first visible light image and a first SAR image of a target city within the same time period.
[0038] S1012: Perform data preprocessing on the first visible light image and the first SAR image respectively.
[0039] The data preprocessing process of the visible light image includes radiation calibration, atmospheric correction and orthorectification. The radiation calibration is a step that must be performed when the user needs to calculate the spectral reflectance or spectral radiation brightness of the object, or needs to compare the images acquired at different times and by different sensors. This process mainly converts the brightness grayscale value (Digital Number, DN) of the image into absolute radiation brightness; the atmospheric correction refers to the correction of the pixel value of the remote sensing image by calculating and analyzing the atmospheric influence of the remote sensing image, so as to obtain more accurate surface coverage information and remote sensing data; the orthorectification is the process of forming an orthorectified projection by digitally correcting the influence of the central projection. It uses the digital elevation model (DEM) data to simultaneously perform tilt correction and projection difference correction on the image, and resample the image into an orthorectified image.
[0040] The data preprocessing process of the SAR image includes registration, filtering, geocoding and radiation calibration. In one embodiment of the present application, in order to solve the problem that the SAR image contains some coherent speckle noise and has low spatial resolution, and the edge information of the building is relatively blurred, in the present application, the filtering process of the first SAR image adopts a bilateral filter, and a bilateral filtering algorithm is used to suppress noise and enhance edge information. The bilateral filter (Bilateral Filter, BF) consists of two parts: the spatial distance of the image and the similarity of the pixel grayscale value of the image.
[0041] The step of filtering the first SAR image comprises:
[0042] The first SAR image is input into the following formula for calculating the spatial distance weight function ws of the sliding window to obtain the spatial distance weight of the first SAR image:
[0043]
[0044] Among them, the coordinate (k, l) is the neighborhood coordinate centered at the coordinate (i, j), is the standard deviation in Gaussian filtering;
[0045] The first SAR image is input into the following weight function wr formula for calculating the similarity of the pixel grayscale values of the sliding window to obtain the similarity of the pixel grayscale values of the first SAR image:
[0046]
[0047] Wherein, f(i,j) is the grayscale value of the pixel of the first SAR image at the coordinate (i,j), f(k,l) is the grayscale value of the pixel of the first SAR image at the coordinate (k,l), and the coordinate (k,l) is the neighborhood coordinate centered at the coordinate (i,j). is the standard deviation of pixel grayscale values;
[0048] Since bilateral filtering takes into account both the influence of image spatial distance and the similarity of image grayscale values, the weight function of bilateral filtering is obtained by multiplying the image spatial distance weight function and the pixel grayscale similarity function.
[0049] The pixel gray value similarity of the first SAR image and the spatial distance weight of the first SAR image are input into the following bilateral filtering weight function formula to obtain the bilateral filtering weight function:
[0050]
[0051] The weight function of the bilateral filter is input into the following bilateral filter formula to obtain the first SAR image after noise reduction:
[0052]
[0053] Among them, (k,l)∈S(i,j) means that the coordinate (k,l) is the coordinate with (i,j) as the neighborhood, f'(i,j) is the pixel grayscale value of the first SAR image at the coordinate (i,j) after filtering, and w(i,j,k,l) is the weight function of bilateral filtering.
[0054] Using a bilateral filter to reduce noise on the first SAR image can enable the first SAR image to maintain better edge features.
[0055] In step S102 , aggregation and averaging calculation is performed on the first visible light image to obtain a second visible light image.
[0056] The aggregate average calculation refers to a commonly used data processing and statistical analysis method, which is mainly used to aggregate a group of values and calculate their average value. In this embodiment, the first visible light image obtained is aggregated and averaged, and then the pixel values in the aggregated area or group are averaged to obtain the second visible light image. In this application, this process is a process of upscaling the first visible light image, and the spatial resolution of the obtained second visible light image is lower than that of the first visible light image.
[0057] See also Figure 2 , Figure 2 A flowchart of the steps of obtaining the VV polarization prediction information of the second SAR image in a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application. For step S103, a SAR image VV polarization prediction model is constructed, and the second visible light image is input into the SAR image VV polarization prediction model to obtain the VV polarization prediction information of the second SAR image, including the following steps:
[0058] S1031, constructing a convolutional neural network CNN, inputting the second visible light image into the convolutional neural network CNN, and obtaining second pixel space feature set data:
[0059]
[0060] i is the i-th pixel position, b is the b-th visible light band, is the image of the bth band of visible light, CNN is a convolutional neural network, Represents the spatial feature information of the i-th pixel in the b-th band image of visible light; wherein the second pixel spatial feature set data includes the spatial feature information of each pixel of the second visible light image.
[0061] S1032, inputting the second visible light image and the second pixel spatial feature set data into the following XGBoost machine learning algorithm formula to obtain VV polarization prediction information of the second SAR image;
[0062]
[0063] Among them, low is low resolution, j is the jth band, is the second visible light image of the jth band with low resolution, is the second pixel spatial feature set data corresponding to the second visible light image of the jth band of low resolution, is the polarization prediction information of the second SAR image VV.
[0064] For step S104, inputting the first visible light image into the SAR image VV polarization prediction model to obtain the first SAR image VV polarization prediction information includes the following steps:
[0065] S1041, inputting the first visible light image into the neural network CNN to obtain first pixel spatial feature set data; wherein the first pixel spatial feature set data includes spatial feature information of each pixel of the first visible light image.
[0066] S1042, inputting the first visible light image and the first pixel spatial feature set data into the following XGBoost machine learning algorithm formula to obtain VV polarization prediction information of the first SAR image;
[0067]
[0068] Among them, high is high resolution, j is the jth band, is the first visible light image of the jth band with high resolution, is the first pixel spatial feature set data corresponding to the first visible light image of the jth band of high resolution, is the polarization prediction information of the first SAR image VV.
[0069] Among them, the convolutional neural network CNN is a type of feedforward neural network that includes convolution calculations and has a deep structure. CNN extracts local features of the input image through convolution operations. The convolution operation usually uses a filter (Filter) or a convolution kernel (Kernel) to scan the input image and generate a corresponding feature map (Feature Map). In this application, by constructing a convolutional neural network CNN, a high-resolution first visible light image and a low-resolution second visible light are respectively input into the convolutional neural network CNN, so that the convolutional neural network CNN automatically learns the features in the image, so that it can well capture and describe the details of the image, and obtain the spatial feature information of each pixel in the image.
[0070] The XGBoost is an optimized distributed gradient boosting library designed to be efficient, flexible and mobile. In the present application, by inputting the first / second visible light image and its corresponding first / second pixel spatial feature set data into a preset XGBoost algorithm learning formula for calculation, the polarization prediction information of the SAR image corresponding to the visible light image in the same time period of the target area can be obtained.
[0071] See also Figure 3 , Figure 3 A flowchart of the steps of obtaining a second SAR image in a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application, wherein step S105 specifically includes the following steps:
[0072] S1051. According to the VV polarization prediction information of the second SAR image, subtract the VV polarization prediction information of the second SAR image from the VV polarization information of the first SAR image to obtain a VV polarization information residual of the second SAR image; wherein the VV polarization information of the first SAR image is obtained by a SAR sensor.
[0073] S1052. Perform spatial interpolation on the second SAR image VV polarization information residual according to a thin plate spline function to obtain the first SAR image VV polarization information residual; wherein the spatial resolution of the first SAR image VV polarization information residual is higher than that of the second SAR image VV polarization information residual.
[0074] S1053. Add the VV polarization information residual of the first SAR image to the VV polarization prediction information of the first SAR image to obtain the second SAR image; wherein the spatial resolution of the second SAR image is higher than that of the first SAR image.
[0075] Among them, the polarization information of the SAR image is one of the important features of the SAR image. The SAR sensor can receive echo signals of multiple polarization modes, mainly including horizontal polarization (HH), vertical polarization (VV), horizontal-vertical polarization (HV), etc. These polarization modes reflect different physical processes when radar waves interact with ground targets. The present application uses a high-resolution first visible light image and a low-resolution second visible light image to perform convolutional neural network CNN calculations and XGBoost machine algorithm learning formula calculations to obtain the polarization information residual of the SAR image, and combines the low-resolution first SAR image with the calculated SAR image polarization information residual to generate a second SAR image with high resolution. In other embodiments of the present application, a model for obtaining other polarization information of the SAR image can also be constructed, and the first visible light image and the second visible light image are input into the model to obtain other polarization information of the SAR image, such as HH polarization information, HV polarization information, etc., and these polarization information are combined to generate a high-resolution SAR image.
[0076] Step S106 specifically includes the following steps:
[0077] S1061, respectively inputting the first visible light image and the second SAR image into the following urban building index NDBI calculation formula to obtain a first pixel value corresponding to the first visible light image and a second pixel value corresponding to the second SAR image:
[0078]
[0079] The SWIR is the reflectivity of the short-wave infrared band, the NIR is the reflectivity of the near-infrared band, and the NDBI value is the first / second pixel value.
[0080] The Normalized Difference Built-up Index is a remote sensing index used to measure the density of surface buildings. It is mainly used to evaluate and quantify the distribution of buildings in urban surface cover types. It effectively distinguishes between buildings and non-buildings by calculating the difference in reflectivity of different bands. The principle of NDBI is that buildings and other types of land objects such as vegetation and water bodies have different reflection characteristics in different bands of remote sensing images. By calculating the difference in reflectivity of these bands, building information can be highlighted, thereby realizing the evaluation of urban building density.
[0081] S1062: Use the OSTU method to perform threshold segmentation on the first visible light image and the second SAR image respectively to obtain a first segmentation threshold σ corresponding to the first visible light image. visand the second segmentation threshold σ corresponding to the second SAR image SAR .
[0082] The OSTU method is an image denoising algorithm based on threshold segmentation. Its basic principle is to select a threshold on the original image, regard pixels with pixel values below the threshold as noise, and regard pixels with pixel values above the threshold as clear image parts. The OSTU method can quickly remove a large amount of noise while maintaining the basic features of the image.
[0083] For step S107, the first pixel value and the first segmentation threshold σ vis Comparing to obtain a first comparison value, the specific steps include:
[0084] The first pixel value and the first segmentation threshold σ vis For comparison:
[0085] If the first pixel value is greater than or equal to the first segmentation threshold σ vis , then the first comparison value is 1;
[0086] If the first pixel value is less than the first segmentation threshold σ vis , then the first comparison value is 0;
[0087] When the first comparison value is 1, the first visible light image is judged to be an urban impervious surface image, and when the first comparison value is 0, the first visible light image is judged to be a non-impervious surface image.
[0088] For step S108, the second pixel value and the second segmentation threshold σ SAR Performing comparison to obtain a second comparison value, the specific steps include:
[0089] The second pixel value and the second segmentation threshold σ SAR For comparison:
[0090] If the second pixel value is greater than or equal to the second segmentation threshold σ SAR , then the second comparison value is 1;
[0091] If the second pixel value is less than the second segmentation threshold σ SAR , then the second comparison value is 0;
[0092] When the second comparison value is 1, the second SAR image is judged to be an urban impervious surface image, and when the second comparison value is 0, the second SAR image is judged to be a non-impervious surface image.
[0093] See also Figure 4, Figure 4 A flowchart of the steps of determining the urban impervious surface image extraction result of a method for extracting urban impervious surfaces based on SAR images and visible light images provided in an embodiment of the present application; for step S109, the first comparison value and the second comparison value are combined to determine the extraction result of the urban impervious surface, and the specific steps include:
[0094] S1091: Combine the first visible light image and the second SAR image to form a city remote sensing image.
[0095] S1092, adding the first comparison value and the second comparison value to obtain an impervious surface value; if the impervious surface value is 2, the urban remote sensing image is judged to be an urban impervious surface image; if the impervious surface value is 1, the urban remote sensing image is manually judged; if the impervious surface value is 0, the urban remote sensing image is judged to be a non-urban impervious surface image.
[0096] Compared with the prior art, the present invention first obtains the first visible light image and the first SAR image of the target city in the same time period, obtains the second visible light image by reducing the resolution of the first visible light image, obtains the pixel spatial feature information of the high-resolution and low-resolution visible light images respectively, and then performs machine learning calculation to obtain the polarization prediction information of the SAR image, combines the first SAR image with the SAR image polarization prediction information, and generates a high-resolution second SAR image. The first visible light image and the second SAR image are threshold-compared and then combined, and finally the extraction result of the urban impervious surface image is obtained. The SAR image quality within the target city is improved, and the accuracy of urban impervious surface extraction is improved by threshold-comparing the SAR image and the visible light image respectively, and then combining them for comprehensive judgment.
[0097] The above embodiments are only used to introduce the technical solution of the present application in detail, but the description of the above embodiments is only used to help understand the method of the embodiments of the present application and should not be understood as limiting the present application. Any changes or substitutions that can be easily thought of by technicians in this field should be included in the protection scope of the embodiments of the present application.
Claims
1. A method for extracting urban impervious surfaces based on SAR images and visible light images, characterized in that: include: Acquire a first visible light image and a first SAR image of the target city in the same time period; wherein the spatial resolution of the visible light image and the first SAR image is greater than 10m; Performing aggregate average calculation on the first visible light image to obtain a second visible light image; wherein the spatial resolution of the second visible light image is lower than that of the first visible light image; Constructing a SAR image VV polarization prediction model, inputting the second visible light image into the SAR image VV polarization prediction model, and obtaining second SAR image VV polarization prediction information; Inputting the first visible light image into the SAR image VV polarization prediction model to obtain first SAR image VV polarization prediction information; Based on the first SAR image, the first SAR image, the VV polarization prediction information of the first SAR image, and the VV polarization prediction information of the second SAR image are input into a preset second SAR image generation formula for information fusion calculation to obtain a second SAR image, including: According to the VV polarization prediction information of the second SAR image, subtract the VV polarization prediction information of the second SAR image from the VV polarization information of the first SAR image to obtain a VV polarization information residual of the second SAR image; wherein the VV polarization information of the first SAR image is obtained by a SAR sensor; According to the thin plate spline function, spatially interpolate the polarization information residual of the second SAR image VV to obtain the polarization information residual of the first SAR image VV; wherein the spatial resolution of the polarization information residual of the first SAR image VV is higher than that of the polarization information residual of the second SAR image VV; Adding the VV polarization information residual of the first SAR image to the VV polarization prediction information of the first SAR image to obtain the second SAR image; wherein the spatial resolution of the second SAR image is higher than that of the first SAR image; The first visible light image and the second SAR image are respectively subjected to the urban building index NDBI calculation to obtain the first pixel value and the second pixel value, and the first visible light image and the second SAR image are respectively subjected to the threshold segmentation using the OSTU method to obtain the first segmentation threshold σ vis and the second segmentation threshold σ SAR ; The first pixel value and the first segmentation threshold σ vis Performing comparison to obtain a first comparison value; The second pixel value and the second segmentation threshold σ SAR Performing comparison to obtain a second comparison value; The first comparison value and the second comparison value are combined to determine the extraction result of the urban impervious surface.
2. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 1, characterized in that: The step of acquiring the first visible light image and the first SAR image of the target city within the same time period includes: Acquire the first visible light image and the first SAR image of the target city within the same time period; The first visible light image and the first SAR image are respectively subjected to data preprocessing; wherein the data preprocessing process of the visible light image includes radiation calibration, atmospheric correction and orthorectification; the data preprocessing process of the SAR image includes registration, filtering, geocoding and radiation calibration.
3. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 2 is characterized in that: The filtering process of the first SAR image adopts a bilateral filter, and the steps of filtering the first SAR image include: The first SAR image is input into the following formula for calculating the spatial distance weight function ws of the sliding window to obtain the spatial distance weight of the first SAR image: Among them, the coordinate (k, l) is the neighborhood coordinate centered at the coordinate (i, j), is the standard deviation in Gaussian filtering; The first SAR image is input into the following weight function wr formula for calculating the similarity of the pixel grayscale values of the sliding window to obtain the similarity of the pixel grayscale values of the first SAR image: Wherein, f(i,j) is the grayscale value of the pixel of the first SAR image at the coordinate (i,j), f(k,l) is the grayscale value of the pixel of the first SAR image at the coordinate (k,l), and the coordinate (k,l) is the neighborhood coordinate centered at the coordinate (i,j). is the standard deviation of pixel grayscale values; The pixel gray value similarity of the first SAR image and the spatial distance weight of the first SAR image are input into the following bilateral filtering weight function formula to obtain the bilateral filtering weight function: The weight function of the bilateral filter is input into the following bilateral filter formula to obtain the first SAR image after noise reduction: Among them, (k,l)∈S(i,j) means that the coordinate (k,l) is the coordinate with (i,j) as the neighborhood, f'(i,j) is the pixel grayscale value of the first SAR image at the coordinate (i,j) after filtering, and w(i,j,k,l) is the weight function of bilateral filtering.
4. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 1, characterized in that: The step of constructing a SAR image VV polarization prediction model, inputting the first visible light image into the SAR image VV polarization prediction model, and obtaining the first SAR image VV polarization prediction information comprises: Construct a convolutional neural network CNN, input the second visible light image into the convolutional neural network CNN, and obtain the second pixel space feature set data: i is the i-th pixel position, b is the b-th visible light band, is the image of the bth band of visible light, CNN is a convolutional neural network, Represents the spatial feature information of the i-th pixel in the b-th band image of visible light; wherein the second pixel spatial feature set data includes the spatial feature information of each pixel of the second visible light image; Inputting the second visible light image and the second pixel spatial feature set data into the following XGBoost machine learning algorithm formula to obtain VV polarization prediction information of the second SAR image; Among them, low is low resolution, j is the jth band, is the second visible light image of the jth band with low resolution, is the second pixel spatial feature set data corresponding to the second visible light image of the jth band of low resolution, is the polarization prediction information of the second SAR image VV.
5. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 4 is characterized in that: The step of inputting the first visible light image into the SAR image VV polarization prediction model to obtain the first SAR image VV polarization prediction information comprises: Inputting the first visible light image into the convolutional neural network CNN to obtain first pixel spatial feature set data; wherein the first pixel spatial feature set data includes spatial feature information of each pixel of the first visible light image; Inputting the first visible light image and the first pixel spatial feature set data into the following XGBoost machine learning algorithm formula to obtain VV polarization prediction information of the first SAR image; Among them, high is high resolution, j is the jth band, is the first visible light image of the jth band with high resolution, is the first pixel spatial feature set data corresponding to the first visible light image of the jth band of high resolution, is the polarization prediction information of the first SAR image VV.
6. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 1, characterized in that: The first visible light image and the second SAR image are respectively subjected to the urban building index NDBI calculation to obtain the first pixel value and the second pixel value, and the first visible light image and the second SAR image are respectively subjected to the threshold segmentation by using the OSTU method to obtain the first segmentation threshold σ vis and the second segmentation threshold σ SAR The steps include: The first visible light image and the second SAR image are respectively input into the following urban building index NDBI calculation formula to obtain a first pixel value corresponding to the first visible light image and a second pixel value corresponding to the second SAR image: Wherein, the SWIR is the reflectivity of the short-wave infrared band, the NIR is the reflectivity of the near-infrared band, and the NDBI value is the first / second pixel value; The OSTU method is used to perform threshold segmentation on the first visible light image and the second SAR image, respectively, to obtain a first segmentation threshold σ corresponding to the first visible light image. vis and the second segmentation threshold σ corresponding to the second SAR image SAR .
7. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 1, characterized in that: The first pixel value and the first segmentation threshold σ vis The step of performing comparison to obtain a first comparison value comprises: The first pixel value and the first segmentation threshold σ vis For comparison: If the first pixel value is greater than or equal to the first segmentation threshold σ vis , then the first comparison value is 1; If the first pixel value is less than the first segmentation threshold σ vis , then the first comparison value is 0; When the first comparison value is 1, the first visible light image is judged to be an urban impervious surface image, and when the first comparison value is 0, the first visible light image is judged to be a non-impervious surface image.
8. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 7, characterized in that: The second pixel value and the second segmentation threshold σ SAR The step of performing comparison to obtain a second comparison value comprises: The second pixel value and the second segmentation threshold σ SAR For comparison: If the second pixel value is greater than or equal to the second segmentation threshold σ SAR , then the second comparison value is 1; If the second pixel value is less than the second segmentation threshold σ SAR , then the second comparison value is 0; When the second comparison value is 1, the second SAR image is judged to be an urban impervious surface image, and when the second comparison value is 0, the second SAR image is judged to be a non-impervious surface image.
9. The method for extracting urban impervious surfaces based on SAR images and visible light images according to claim 8, characterized in that: The step of combining the first comparison value and the second comparison value to determine the extraction result of the urban impervious surface includes: combining the first visible light image with the second SAR image to form a city remote sensing image; The first comparison value and the second comparison value are added to obtain an impervious surface value; if the impervious surface value is 2, the urban remote sensing image is judged to be an urban impervious surface image; if the impervious surface value is 1, the urban remote sensing image is manually judged; if the impervious surface value is 0, the urban remote sensing image is judged to be a non-urban impervious surface image.
