A DOM-assisted method and system for automatic extraction of stereo SAR control points

Abstract

This invention discloses a method and system for automatic extraction of stereo SAR control points using DOM-assisted methods. The method involves: reading sub-meter-level DOM image data; reading DEM image data; sharpening the DOM image; extracting streetlight templates and obtaining the streetlight base positions; template matching to extract streetlight base positions from the DOM image; reading SAR image data and RPC parameters to construct an RPC model; registering the DOM and SAR images to obtain the streetlight coordinates in the SAR image; repeating the processing of other SAR images; refining the streetlight coordinates through point target analysis; reading SAR image metadata files to construct an RD positioning model; and using stereo SAR processing to obtain the streetlight control point coordinates. This invention not only improves positioning accuracy but also significantly reduces the need for manual intervention, making the entire process more efficient and reliable.

Description

Technical Field

[0001] This invention relates to a method and system for automatic extraction of control points from a stereo synthetic aperture radar (SAR) image assisted by a digital orthophoto map (DOM). Background Technology

[0002] In the fields of geographic mapping and remote sensing, SAR technology has become crucial due to its high-resolution imaging capabilities of the Earth's surface. This technology can be applied to various aspects, including topographic mapping, deformation monitoring, and disaster management. However, accurately detecting and locating ground control points using SAR data has always been a challenge. Traditional methods typically rely on manual identification of permanent scattering points in SAR images, which is not only time-consuming but also prone to errors. This problem poses a challenge to the automatic extraction of SAR control points. Summary of the Invention

[0003] To overcome the shortcomings of the prior art, this invention provides a DOM-assisted method and system for automatic extraction of stereo SAR control points. With the assistance of sub-meter-level DOM optical image data, the specific radar position of a street lamp in satellite-borne SAR images from different perspectives is determined. Then, by using stereo SAR methods and combining the radar coordinates of the street lamp in SAR images obtained from different perspectives, the precise geographic coordinates of the target are recovered.

[0004] According to one aspect of the present invention, a method for automatic extraction of stereo SAR control points with DOM assistance is provided, comprising:

[0005] Read sub-meter level DOM and DEM image data with the same geographical location as the spaceborne SAR image data, and perform preprocessing;

[0006] Extract street light template slice data and obtain the planar geographic coordinates of the street light base location in the slice;

[0007] The approximate planar geographic coordinates of the street lamp base location in the DOM image are extracted by template matching. Then, the approximate geographic coordinates of the street lamp base location are extracted by combining the preprocessed DOM image data, DEM image data and the street lamp base location in the template tile.

[0008] Read SAR image data and RPC parameters, construct an RPC model, and obtain the radar coordinates of the street lamp base location in the current SAR image by registering DOM image with SAR image. Repeat the current step to obtain the radar coordinates of the same street lamp base location in two or more SAR images.

[0009] Based on the radar coordinates of the same street lamp base location in two or more SAR images, the radar coordinates of the street lamp base location are extracted in a refined manner using the point target analysis method.

[0010] An RD positioning model is constructed by reading SAR image metadata files. The radar coordinates of the same street lamp base location in two or more refined SAR images are combined with stereo SAR processing methods to obtain the precise geographic coordinates of the street lamp base location.

[0011] As a further technical solution, the preprocessing includes:

[0012] The approximate planar geographic coordinates of pixels are determined based on the existing parameters of the sub-meter level DOM image data, and High Boost Filtering sharpening is applied to the read sub-meter level DOM image data.

[0013] The planar geographic coordinates and elevation information of the pixel location are determined based on the existing parameters of the DEM image data.

[0014] As a further technical solution, by combining the preprocessed DOM image data, DEM image data, and the location of the street lamp base in the template tiles, the approximate geographical coordinates of the street lamp base location are extracted, including:

[0015] The normalized cross-correlation template matching method is used to match DOM images and street lamp template slice data to extract the approximate planar geographic coordinates of the street lamp base location in the DOM image;

[0016] By combining the location of the street lamp base in the template slice with the planar geographic coordinates and elevation parameters of the DOM and DEM images, the approximate geographic coordinates of the street lamp base location are extracted.

[0017] As a further technical solution, the street light template slice data is extracted and the planar geographic coordinates of the street light base location in the slice are obtained, including:

[0018] Extract street light template slice data, and obtain the planar geographic coordinates of the street light base location in the slice by interpreting the street light distribution in the DOM image data.

[0019] As a further technical solution, by registering DOM imagery with SAR imagery, the radar coordinates of the street lamp base location in the current SAR imagery are obtained, including:

[0020] The DOM image and the current SAR image are registered to obtain the radar coordinate transformation relationship between the DOM image and the SAR image, and then the radar coordinates of the street lamp base position in the current SAR image are obtained.

[0021] As a further technical solution, the method also includes:

[0022] Extract the geographic coordinates of the street lamp base location, and crop the DOM and SAR images based on the extracted geographic coordinates to form a cropping region.

[0023] Resample the cropped DOM and SAR images;

[0024] Registration processing is performed on DOM and SAR images that have been resampled and have the same scale;

[0025] The obtained registration relationship, the extracted geographic coordinates of the street lamp base location, and the RPC parameters of the SAR image are used to determine the radar coordinates of the street lamp in the SAR image.

[0026] As a further technical solution, the current steps are repeated to obtain the radar coordinates of the same street lamp base location in two or more SAR images, including:

[0027] Read other spaceborne SAR data and corresponding RPC parameters, repeat the RPC model construction and image configuration process, obtain the radar coordinates of the street lamp base locations in other SAR images, and use the corresponding information of the street lamp base locations in DOM images to obtain the radar coordinates of the same street lamp base locations in two or more SAR images.

[0028] According to one aspect of the present invention, a DOM-assisted automatic control point extraction system for stereo SAR is provided, comprising:

[0029] The first main module is used to read sub-meter level DOM and DEM image data with the same geographical location as the spaceborne SAR image data, and to perform preprocessing.

[0030] The second main module is used to extract street light template slice data and obtain the planar geographic coordinates of the street light base location in the slice;

[0031] The third main module is used to extract the approximate planar geographic coordinates of the street lamp base location in the DOM image using template matching. Then, it combines the preprocessed DOM image data, DEM image data, and the street lamp base location in the template slice to extract the approximate geographic coordinates of the street lamp base location.

[0032] The fourth main module is used to read SAR image data and RPC parameters, build an RPC model, and obtain the radar coordinates of the street lamp base position in the current SAR image by registering the DOM image with the SAR image. The current step is repeated to obtain the radar coordinates of the same street lamp base position in two or more SAR images.

[0033] The fifth main module is used to extract the radar coordinates of the street lamp base location using point target analysis methods based on the radar coordinates of the same street lamp base location in two or more SAR images.

[0034] The sixth main module is used to read SAR image metadata files to construct RD positioning models. It combines the radar coordinates of the same street lamp base location in two or more refined SAR images and uses stereo SAR processing methods to obtain the precise geographic coordinates of the street lamp base location.

[0035] According to one aspect of the present invention, a DOM-assisted stereo SAR control point automatic extraction device is provided, comprising a memory and a processor, wherein the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to perform the steps of the DOM-assisted stereo SAR control point automatic extraction method.

[0036] According to one aspect of the present invention, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions that cause the computer to perform the steps of the DOM-assisted stereo SAR control point automatic extraction method.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] This invention, based on spaceborne stereo SAR technology, achieves high-precision absolute positioning of permanent scatterers in SAR images. Assisted by sub-meter-level DOM optical image data, it determines the specific radar position of a light pole in spaceborne SAR images from different viewpoints. Then, using a stereo SAR method, combined with the light pole's radar coordinates from SAR images acquired from different viewpoints, it reconstructs the target's precise geographic coordinates. This invention not only improves positioning accuracy but also significantly reduces the need for manual intervention, making the entire process more efficient and reliable. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating a DOM-assisted automatic extraction method for stereo SAR control points, provided in an embodiment of the present invention.

[0041] Figure 2This is a schematic diagram of a DOM-assisted automatic control point extraction system for stereo SAR provided in an embodiment of the present invention. Detailed Implementation

[0042] Streetlight poles are excellent permanent scattering bodies, existing stably and abundantly throughout urban areas, making them ideal targets for ground control point extraction. By determining the precise radar coordinates of the poles and combining this with spaceborne stereo SAR technology, the accurate geographic coordinates of the poles can be obtained. Furthermore, because streetlight poles exhibit unique shadow characteristics in high-precision DOM imagery, making them easily detectable, automatic SAR control point detection technology using high-precision DOM imagery data as an aid becomes possible.

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0044] Please see Figure 1 This invention provides a DOM-assisted method for automatic extraction of stereo SAR control points, comprising the following steps:

[0045] Step 1: Read sub-meter level DOM optical image data with the same geographical location as the satellite SAR image data, and determine the approximate planar geographic coordinates of the pixels based on the existing parameters of the sub-meter level DOM optical image data.

[0046] Step 2: Read the Digital Elevation Model (DEM) image data that has the same geographical location as the satellite SAR image data, and determine the planar geographic coordinates and elevation information of the pixel location based on the existing parameters of the DEM image data.

[0047] Step 3: Perform High Boost Filtering sharpening on the sub-meter DOM optical image data read in Step 1 to ensure that the outline and shadow of the streetlights in the image are more prominent for subsequent extraction and recognition.

[0048] In this step, the High Boost Filtering sharpening process is as follows:

[0049] I = I0 + aI m (1)

[0050] Where I represents the filtered image, I0 represents the original image, and a represents the scalar sharpening factor. m It's an unsharpening mask, I m The difference between I0 and its Gaussian blurred version is calculated; the higher the factor a value, the higher the sharpening level.

[0051] Gaussian blur is a commonly used technique in image processing, mainly used to reduce image noise and detail. Its principle is as follows:

[0052]

[0053] Where G(x,y) represents the pixel value at coordinates (x,y) in the image after Gaussian blurring, and σ represents the standard deviation of the normal distribution. In two-dimensional space, the contour lines of the surface generated by this formula are concentric circles with a normal distribution starting from the center. The convolution matrix composed of pixels with non-zero distribution is used to transform the original image. The value of each pixel is a weighted average of the values ​​of its surrounding pixels. The original pixel has the largest Gaussian distribution value, so it has the largest weight. The weight of neighboring pixels decreases as they become farther away from the original pixel. This blurring process preserves edge effects better than other equal blur filters.

[0054] Step 4: Extract street light template slice data. By interpreting the street light distribution on the road in the DOM, obtain the planar geographic coordinates of the street light base location in the slice.

[0055] Step 5: Use the normalized cross-correlation template matching method to match the DOM image and the street lamp template tile data, and extract the approximate planar geographic coordinates of the street lamp base location in the DOM image. Then, combine the street lamp base location in the template tile with the planar geographic coordinates and elevation parameters of the DOM and DEM images to extract the approximate geographic coordinates of the street lamp base location.

[0056] In this step, the normalized cross-correlation template matching method is a similarity measurement method widely used in image processing and computer vision, especially in tasks such as image registration. It finds the optimal relative displacement by comparing the similarity between two signal or image regions. The calculation method is as follows:

[0057]

[0058] The motivation for using cross-correlation for template matching is a distance metric (squared Euclidean distance), where f is the image, and d is obtained by taking the squared difference between the feature window t located at u,v and the image pixel by pixel. 2 Expanding, we get:

[0059]

[0060] Where, ∑t 2 (xu, yv) are constants. Then if ∑f 2 If (x, y) is approximately constant, then the remaining cross-correlation terms are:

[0061] c(u,v)=∑ x,y [f(x,y)t(xu,yv)] (5)

[0062] Then c(u,v) is the similarity measure between the image and the features. The correlation coefficient is generated by normalizing the image and feature vector to unit length, resulting in a cosine-like correlation coefficient, expressed as:

[0063]

[0064] in, It is the average of the eigenvalues. It is the average value of the region f(x,y) under the feature.

[0065] Step 6: Read the spaceborne SAR image data, organize the Rational Polynomial Coefficient (RPC) parameter file and SAR image metadata file format to conform to international standards, read the RPC positioning parameter file of the spaceborne SAR image, and construct the RPC model of the spaceborne SAR image using the RPC model parameters in the RPC positioning parameter file.

[0066] In this step, the RPC model is defined as follows:

[0067]

[0068] For ease of expression, equation (7) is also written in the following form:

[0069]

[0070] Where: (X) n ,Y n Z n ) represents the regularized geographic coordinates of the ground point, (r) n ,c n F represents the regularized radar coordinates. r (X n ,Y nZ n ), F c (X n ,Y n Z n ) respectively with r n c n The meanings are the same, P1(X) n ,Y n Z n P2(X) n ,Y n Z n P3(X) n ,Y n Z n P4(X) n ,Y n Z n Let these be P1, P2, P3, and P4, where P1, P2, P3, and P4 are terms relating to X. n Y n Z n The cubic polynomial has no actual physical meaning.

[0071] P1=a0+a1Z n +a2Y n +a3X n +a4Z n Y n +a5Z n X n +a6Z n X n +a7Z n 2 +a8Y n 2 +a9X n 2 + n10 Z n Y n X n +a 11 Z n 2 Y n +a 12 Z n 2 X n +a 13 Y n 2 Z n +a 14 Y n 2 X n +a 15 Z n X n 2+a 16 Y n X n 2 +a 17 Z n 3 +a 18 Y n 3 +a 19 X n 3

[0072] P2=b0+b1Z n +b2Y n +b3X n +b4Z n Y n +b5Z n X n +b6Y n X n +b7Z n 2 +b8Y n 2 +b9X n 2 +b 10 Z n Y n X n +b 11 Z n 2 Y n +b 12 Z n 2 X n +b 13 Y n 2 Z n +b 14 Y n 2 X n +b 15 Z n X n 2 +b 16 Y n X n 2 +b 17 Z n 3 +b 18 Y n 3 +b 19 X n 3

[0073] Q3=c0+c1Zn +c2Y n +c3X n +c4Z n Y n +c5Z n X n +c6Y n X n +c7z n 2 +c8Y n 2 +c9X n 2 +c 10 Z n Y n X n +c 11 Z n 2 Y n +c 12 Z n 2 X n +c 13 Y n 2 Z n +c 14 Y n 2 X n +c 15 Z n X n 2 +c 16 Y n X n 2 +c 17 Z n 3 +c 18 Y n 3 +c 19 X n 3

[0074] Q4=d0+d1Z n +d2Y n +d3X n +d4Z n Y n +d5Z n X n +d6Y n X n +d7Z n 2 +d8Y n 2 +d9X n2 +d 10 Z n Y n X n +d 11 Z n 2 Y n +d 12 Z n 2 X n +d 13 Y n 2 Z n +d 14 Y n 2 X n +d 15 Z n X n 2 +d 16 Y n X n 2 +d 17 Z n 3 +d 18 Y n 3 +d 19 X n 3

[0075] Where a0,…,a 19 b0,…,b 19 c0,…,c 19 ,d0,…,d 19 These are model parameters that come with the RPC parameter file.

[0076] Regularized ground point geodetic coordinates (X n ,Y n Z n ), regularized radar coordinates (r) n ,c n The correspondence between the ground point's geodetic coordinates (X,Y,Z) and radar coordinates (R,C) is shown below:

[0077]

[0078] In equations (9) and (10), X is longitude, Y is latitude, Z is elevation, and X o Y o Z o R o C o X is the translation parameter for regularization; sY s Z s R s C s These are regularization scaling coefficients, all of which are model parameters included in the RPC file.

[0079] Step 7: By registering the DOM and the current SAR image, the radar coordinate transformation relationship between the DOM and the SAR image is obtained, and then the radar coordinates of the street lamp base in the current SAR image are obtained.

[0080] The steps for calculating the radar coordinates of the streetlight base in the SAR image are as follows:

[0081] Step 7.1 requires extracting the geographic coordinates of the streetlight base locations. These coordinates are typically obtained from imagery or other data sources through a specific method. Next, based on these geographic coordinates, a minimum bounding rectangle is constructed. The boundary of this rectangle should completely encompass all streetlight base locations. To ensure the area containing the streetlights is sufficiently complete and clear, this rectangle can be appropriately extended by a certain distance (e.g., a certain proportion or a fixed distance) to ensure its extent is broad enough to include surrounding areas that might affect streetlight detection. After obtaining this extended rectangular area, the next step is to use this rectangle to crop the DOM and SAR images. The cropped image area will contain the streetlight bases and their surrounding environment, providing sufficient information for subsequent image registration and analysis, thus preparing for the subsequent registration and analysis.

[0082] Step 7.2: In this step, the cropped DOM and SAR images first need to be resampled. Since the DOM and SAR images come from different sources, they may differ in resolution, coordinate system, and scale. Therefore, uniform processing is required to ensure consistency in their resolution and coordinate system. For DOM images, resampling typically relies on the image's six parameters (translation, rotation, scaling, and tilt). These parameters help convert the DOM image into a standard map coordinate system, ensuring geometric accuracy. For SAR images, resampling is performed using the SAR image's RPC parameters. The RPC model allows for the inverse calculation of the geographic coordinate system from radar coordinates in the SAR image, or vice versa, converting geographic coordinates to radar coordinates to obtain radar coordinates. During resampling, elevation values ​​need to be selected from the DEM image to determine the virtual elevation plane of the image. After resampling, the resulting DOM and SAR images will have the same scale and resolution, providing a consistent data foundation for the next step of image registration.

[0083] Step 7.3 involves registering the resampled DOM and SAR images at the same scale. Specifically, this is done using an image matching method based on the Radiation-Variation Insensitive Feature Transform (RIFT). The RIFT method finds correspondences between images by detecting and matching feature points (such as corners and edges) in the image pair. RIFT can find robust feature points across different image types and perform matching. During image matching, the RIFT algorithm calculates corresponding points (pixels representing the same feature or geographic point in both images) based on local features (such as texture, edges, and corners). These corresponding points are crucial for subsequent registration. By extracting and matching corresponding points in the image pair, the spatial relationships between these points can be used to determine the pixel registration relationship of the image pair. The registration relationship is typically expressed as a transformation matrix, which describes how radar coordinates in one image are mapped to corresponding radar coordinates in another image.

[0084] Step 7.4 uses the registration relationship obtained in the previous step, the extracted geographic coordinates of the street lamp base location, and the RPC parameters of the SAR image to determine the radar coordinates of the street lamp in the SAR image. Specifically, firstly, based on the extracted geographic coordinates of the street lamp base location, the geographic coordinates are converted into radar coordinates of the SAR image using the RPC parameters of the SAR image. This process relies on the RPC model, which can accurately convert the relationship between spatial coordinates and image coordinates. Then, using the pixel registration relationship obtained in the previous step, the radar coordinates in the SAR image converted from the geographic coordinates can be further adjusted. This is because the registration relationship between the DOM image and the SAR image may introduce certain errors or deviations, so it is necessary to use the registration relationship between corresponding points to perform accurate radar coordinate adjustment. In this way, the accurate radar coordinates of the street lamp base location in the SAR image can be obtained.

[0085] Step 8: Read other spaceborne SAR data and repeat steps 6 to 7 to obtain the radar coordinates of the street lamp base locations in other SAR images. Then, using the corresponding information of the street lamp base locations in the DOM images, obtain the radar coordinates of the same street lamp base locations in two or more SAR images.

[0086] Step 9: Using the radar coordinates of the same street lamp base location in two or more SAR images, and based on the unique reflection characteristics of the street lamp, which presents a bright isolated focal point in the SAR image, the radar coordinates of the street lamp base location are extracted in a refined manner using the point target analysis method.

[0087] In this step, the target of the point target analysis algorithm is to measure point target scatterers (such as corner reflectors (CRs), street lamp bases, etc.). After imaging processing, point target scatterers appear as Sinc impulse responses in SAR images. However, due to the bandwidth limitations of SAR signals in terms of range and azimuth, the impulse response of the two-dimensional Sinc impulse is broadened, causing signal energy to be dispersed across other pixels. Since the shape of the Sinc signal can be well defined and analyzed, the response of the point target can be accurately interpolated after certain signal processing. Finally, the peak position of the point target response in the interpolated image patch is located with sub-pixel accuracy. The calculation steps are as follows:

[0088] Step 9.1: Using the initial image coordinates obtained from geometric inverse calculation as the center, extract a 128*128 image block and find the position of the maximum peak. Then, using the maximum peak as the center, extract a 32*32 sample window and perform a two-dimensional discrete Fourier transform. After zero-padding in the frequency domain (the oversampling factor can be 32, 64, or 128), perform a two-dimensional discrete inverse Fourier transform to obtain the upsampled point target image block. Since the sampled Sinc pulse signal has a fairly smooth shape, the signal shape between the oversampled pixels around the peak can be fitted by a model function. An elliptic paraboloid is a suitable function to describe the peak.

[0089] f(x,y)=a0+a1y+a2x+a3xy+a4y 2 +a5x 2 (11)

[0090] Step 9.2: For the oversampled image block, a 3*3 sub-block is cropped from the point of maximum peak value, and the six parameters a are solved by least squares in equation (11). i (0≤i≤5), after solving for the parameters, the vertex coordinates of the parabola can be analyzed, thereby obtaining the peak position coordinates and peak power more accurately. Once the sub-pixel positions (x,y) of the scatterers in the image are estimated, they can be converted into the corresponding range-time vector τ based on the following parameters obtained from the SAR image metadata file. m and azimuth time t m :

[0091]

[0092] Where, τ img0 The fast time for the range to the first range sample is given by RSF, where RSF is the range sampling frequency, and t is the distance sampling frequency. img0Here, ASF is the azimuth start time, and ANSI is the azimuth sampling frequency. As long as no resampling is applied during SAR processing, its value is equal to the pulse repetition frequency (PRF) in the original data. The accuracy of these parameters can affect the acquired time precision during this process.

[0093] Step 10: Read the SAR image metadata file and use the RD model parameters in the image metadata file to construct the range-Doppler (RD) positioning model of the spaceborne SAR image.

[0094] In this step, the RD positioning model is defined as follows:

[0095]

[0096] Among them, X T Represented as the target location, X S Indicated as sensor position, Represented as the sensor velocity vector, where c is the speed of light, f dc τ is the Doppler center frequency. r λ is the two-way propagation delay, and λ is the wavelength.

[0097] The RD positioning model is the basis for the conversion between SAR image (time) coordinates and spatial coordinates.

[0098] However, since SAR image (time) coordinates are two-dimensional, there is still one degree of freedom when converting SAR image coordinates to spatial coordinates. To solve this problem, additional information is needed, such as a reference plane (Earth ellipsoid, digital terrain model):

[0099]

[0100] Among them, the radius R of the Earth ellipsoid p =(1-f)R e Target elevation Solving equations (13) and (14) simultaneously is also called the forward calculation of the RD model, which is to obtain the spatial rectangular coordinates of the target. Conversely, the process of converting target geographic coordinates to SAR image (time) coordinates is called RD model inverse calculation.

[0101] Step 11: Using the radar coordinates of the same street lamp base location in two or more refined SAR images, the precise geographic coordinates of the street lamp base location are obtained by using the stereo SAR processing method.

[0102] In this step, the stereo SAR processing method is essentially a general type of least squares parameter estimation problem. It linearizes the range-Doppler geometric positioning model and solves for the geographic coordinates of the observed target while considering the uncertainty of the coordinate observation values. Suitable observation targets require a stable phase center and a stable radar backscattering cross section (RCS) over a wide incident angle range. Therefore, strong scatterers and corner reflectors are generally used as observation targets in stereo imagery. The solution principle of stereo SAR processing methods is as follows:

[0103] According to equation (13), the geometric localization process of SAR images can be derived from f R f A and Doppler center frequency f dc The range-Doppler equations represent:

[0104]

[0105] In the formula X T X represents the position vector of the target to be determined. s and It is related to the orientation and time t a The relevant SAR antenna phase center position and velocity vector, τ r It represents the two-way propagation time of the distance, which can be converted into slant range by using half the speed of light, c. Note that the Doppler equation has two cases: zero Doppler and non-zero Doppler.

[0106] Azimuth to time t a It is introduced through the orbital polynomial model, and the model coefficients can be determined by fitting the orbital data in the product metadata using polynomial modeling:

[0107]

[0108] Where a i b i c i Represents the polynomial coefficients.

[0109] Regarding the observation geometry, taking the zero-Doppler case as an example, at each observation signal acquisition, the target at the zero-Doppler position lies relative to the sensor on a circle along the track with a radius equal to the slant range. At least two or more observations are required to determine the position vector X. T .

[0110] Range observation τ r and azimuth observation t a and the unknown parameter X TThe target coordinates are implicitly contained in the range-Doppler equations coupled with the orbital model. In parameter adjustment, the range-Doppler equations can be considered conditional equations; therefore, the problem is usually expressed as conditional adjustment with unknown parameters. Because the range-Doppler equations are nonlinear, the first step in solving the least-squares adjustment is to linearize the adjustment parameters and observations using Taylor series, abbreviated as:

[0111]

[0112] Assuming there are n pairs of observations, the target coordinates are to be determined. After linearizing equation (17), the following error equation can be obtained:

[0113]

[0114] After linearization, two coefficient matrices B and A are formed. These coefficient matrices are obtained by applying the distance-Doppler equation to τ. r and azimuth observation t a and the unknown parameter X T The partial derivatives are used to generate the error equation. Vector v represents the observation residuals, while w = Bl - b represents the inconsistencies in the observation conditions. These are solved using x and v. The entire error equation is solved using the least squares rule. T The solution is obtained by minimizing the variance (pv). The matrix weights each observation, typically the inverse of the variance-covariance matrix of the observations. In short, the above equation is an optimization problem that considers the uncertainty of the observations and the unknown location vector to be solved, ultimately solved by minimizing the residuals.

[0115]

[0116] Due to the nonlinearity of the observation equations, the entire adjustment must be performed iteratively. This requires determining the desired location X of the observed target. T =X T0 Perform initial coordinate estimation. When the target's geographic coordinates X... T When the solution converges in the iterative steps, the inconsistency will eventually decay, and the range residual and azimuth residual v will tend to be minimized. For a specific observation target, there are 3 unknowns, and each SAR image provides a set of range and azimuth observations, which can form two observation equations. Therefore, the geographic coordinates of the observation target can be solved if and only if n>2 SAR images are acquired.

[0117] The implementation of the various embodiments of the present invention is based on programmed processing by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of the present invention are encapsulated into various modules. Based on this reality, and building upon the above embodiments, the embodiments of the present invention provide a DOM-assisted automatic stereo SAR control point extraction system, which is used to execute a DOM-assisted automatic stereo SAR control point extraction method from the above method embodiments.

[0118] See Figure 2 The system includes: a first main module for reading sub-meter level DOM and DEM image data with the same geographical location as the satellite-borne SAR image data and performing preprocessing; a second main module for extracting street lamp template slice data and obtaining the planar geographic coordinates of the street lamp base location in the slice; a third main module for extracting the approximate planar geographic coordinates of the street lamp base location in the DOM image using template matching, and then combining the preprocessed DOM image data, DEM image data, and the street lamp base location in the template slice to extract the approximate geographic coordinates of the street lamp base location; and a fourth main module for reading SAR image data and RPC parameters, constructing an RPC model, and using D... The first main module registers OM imagery with SAR imagery to obtain the radar coordinates of the street lamp base location in the current SAR imagery. This step is repeated to obtain the radar coordinates of the same street lamp base location in two or more SAR images. The second main module uses point target analysis to refine the radar coordinates of the street lamp base location based on the radar coordinates of the same street lamp base location in two or more SAR images. The third main module reads the SAR image metadata file to construct an RD positioning model. Combined with the refined extracted radar coordinates of the same street lamp base location in two or more SAR images, the module uses stereo SAR processing to obtain the precise geographic coordinates of the street lamp base location.

[0119] This invention provides a DOM-assisted automatic control point extraction system for stereo SAR, addressing the shortcomings of traditional methods that rely on manual identification of permanent scattering points in SAR images, which is not only time-consuming but also prone to errors. Figure 2 Several modules within the system, aided by sub-meter-level DOM optical image data, determine the specific radar location of streetlights in satellite-borne SAR images from different perspectives; then, using a stereo SAR method, combined with the radar coordinates of the streetlights in SAR images acquired from different perspectives, the system recovers the precise geographic coordinates of the target.

[0120] It should be noted that the system embodiments provided by the present invention are used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The only difference is that corresponding functional modules are set. The principle is basically the same as that of the above system embodiments provided by the present invention. As long as those skilled in the art can improve the modules in the above system embodiments by referring to the specific technical solutions in other method embodiments and combining technical features to obtain corresponding technical means and technical solutions composed of these technical means, on the basis of the above system embodiments, and on the premise of ensuring the practicality of the technical solutions, they can obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments.

[0121] Based on the same inventive concept as the above embodiments, this embodiment of the invention also provides a DOM-assisted stereo SAR control point automatic extraction device, including a memory and a processor. The memory stores program instructions that are executed by the processor, and the processor calls the program instructions to execute the steps of the DOM-assisted stereo SAR control point automatic extraction method.

[0122] Based on the same inventive concept as the above embodiments, this embodiment of the invention also provides a non-transitory computer-readable storage medium storing computer instructions that cause the computer to execute the steps of the DOM-assisted stereo SAR control point automatic extraction method.

[0123] In summary, this invention discloses a DOM-assisted automatic extraction method for stereo SAR control points. The method involves: reading sub-meter level DOM image data; reading DEM image data; sharpening the DOM image; extracting streetlight templates and obtaining the streetlight base positions; template matching to extract streetlight base positions from the DOM image; reading SAR image data and RPC parameters to construct an RPC model; registering the DOM and SAR images to obtain the streetlight coordinates in the SAR image; repeating the processing of other SAR images; refining the streetlight coordinates through point target analysis; reading SAR image metadata files to construct an RD positioning model; and using stereo SAR processing to obtain the streetlight control point coordinates. This invention not only improves positioning accuracy but also significantly reduces the need for manual intervention, making the entire process more efficient and reliable.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A DOM-assisted automatic extraction method of stereo SAR control points, characterized in that, The method comprises the following steps: reading sub-meter DOM image data and DEM image data having the same geographical position as the space-borne SAR image data and pre-processing the same; extracting road lamp template slice data and obtaining the planar geographical coordinates of the road lamp base position in the slice; extracting the approximate planar geographical coordinates of the road lamp base position in the DOM image by using the template matching method, and combining the pre-processed DOM image data, DEM image data and the road lamp base position in the template slice to extract the approximate geographical coordinates of the road lamp base position; reading the SAR image data and RPC parameters, constructing the RPC model, and obtaining the radar coordinates of the road lamp base position in the current SAR image by registering the DOM image and the SAR image, and repeating the current step to obtain the radar coordinates of the same road lamp base position in two or more SAR images; based on the radar coordinates of the same road lamp base position in two or more SAR images, using the point target analysis method to finely extract the radar coordinates of the road lamp base position; reading the SAR image metadata file to construct the RD positioning model, combining the finely extracted radar coordinates of the same road lamp base position in two or more SAR images, and using the stereo SAR processing method to obtain the accurate geographical coordinates of the road lamp base position.

2. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 1, wherein, The pre-processing comprises: determining the approximate planar geographical coordinates of the pixels according to the existing parameters of the sub-meter DOM image data, and performing High Boost Filtering sharpening processing on the read sub-meter DOM image data; determining the planar geographical coordinates and elevation information of the pixel position according to the existing parameters of the DEM image data.

3. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 2, wherein, Combining the pre-processed DOM image data, DEM image data and road lamp base position in the template slice to extract the approximate geographical coordinates of the road lamp base position comprises: using the normalized cross-correlation template matching method to match the DOM image and the road lamp template slice data to extract the approximate planar geographical coordinates of the road lamp base position in the DOM image; combining the road lamp base position in the template slice and the planar geographical coordinates and elevation parameters of the DOM and DEM images to extract the approximate geographical coordinates of the road lamp base position.

4. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 1, wherein, Extracting road lamp template slice data and obtaining the planar geographical coordinates of the road lamp base position in the slice comprises: extracting the road lamp template slice data, and obtaining the planar geographical coordinates of the road lamp base position in the slice by interpreting the distribution of the road lamps in the DOM image data.

5. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 1, wherein, Obtaining the radar coordinates of the road lamp base position in the current SAR image by registering the DOM image and the SAR image comprises: registering the DOM image and the current SAR image to obtain the radar coordinate conversion relationship between the DOM image and the SAR image, and then obtaining the radar coordinates of the road lamp base position in the current SAR image.

6. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 5, wherein, The method further comprises: extracting the geographical coordinates of the road lamp base position, and cutting the DOM image and the SAR image according to the cutting area formed according to the extracted geographical coordinates; resampling the cut DOM and SAR images; registering the resampled DOM and SAR images having the same scale; The obtained registration relationship, the extracted geographic coordinates of the street lamp base position, and the RPC parameters of the SAR image are used to determine the radar coordinates of the street lamp in the SAR image.

7. The DOM aided stereoscopic SAR control point automatic extraction method according to claim 1, wherein, The current step is repeated to obtain the radar coordinates of the same street lamp base position in two or more SAR images, including: The other spaceborne SAR data and corresponding RPC parameters are read, the RPC model construction and image configuration process are repeated, the radar coordinates of the street lamp base position in other SAR images are obtained, and the corresponding information of the street lamp base position in the DOM image is used to obtain the radar coordinates of the same street lamp base position in two or more SAR images.

8. A DOM-assisted stereoscopic SAR control point automatic extraction system, characterized in that, It includes: The first main module is used to read the sub-meter DOM image data and DEM image data with the same geographic position as the spaceborne SAR image data, and to perform preprocessing; The second main module is used to extract the street lamp template slice data and obtain the planar geographic coordinates of the street lamp base position in the slice; The third main module is used to extract the approximate planar geographic coordinates of the street lamp base position in the DOM image by using the template matching method, and to extract the approximate geographic coordinates of the street lamp base position in combination with the preprocessed DOM image data, DEM image data, and street lamp base position in the template slice; The fourth main module is used to read the SAR image data and RPC parameters, construct the RPC model, and obtain the radar coordinates of the street lamp base position in the current SAR image by registering the DOM image and the SAR image, and the current step is repeated to obtain the radar coordinates of the same street lamp base position in two or more SAR images; The fifth main module is used to use the point target analysis method to finely extract the radar coordinates of the street lamp base position based on the radar coordinates of the same street lamp base position in two or more SAR images; The sixth main module is used to read the SAR image metadata file to construct the RD positioning model, combine the radar coordinates of the same street lamp base position in two or more SAR images after fine extraction, and use the stereo SAR processing method to obtain the accurate geographic coordinates of the street lamp base position.

9. A DOM-assisted stereoscopic SAR control point automatic extraction apparatus, characterized by, It includes a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the DOM-assisted stereo SAR control point automatic extraction method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, comprising: The non-transitory computer readable storage medium stores computer instructions, which make the computer execute the steps of the DOM-assisted stereo SAR control point automatic extraction method according to any one of claims 1 to 7.

Images