Double-layer network atmospheric phase error correction and height estimation method based on optimized reference network

Abstract

The invention discloses a double-layer network atmospheric phase error correction and height estimation method based on an optimized reference network. The method comprises the following steps: acquiring data of a plurality of preprocessed two-dimensional single-view complex images; sPS candidate points are extracted, an initial Delaunay reference network is constructed, correction of atmospheric phase errors is realized by using the constructed and connected phase difference signals, minimum spanning tree optimization is carried out based on the residual energy ratio of each side, and a final reference network is generated; performing height estimation on the SPS point in the final reference network relative to a certain reference SPS point by adopting weighted least square estimation; pS candidate points are extracted to be connected with adjacent SPS points in the reference network, a star network is generated, and height estimation of the PS points is completed after atmospheric phase error correction; and generating an elevation map result based on the height information of all the points acquired by the double-layer network. The atmospheric phase does not need to be separated in advance, and atmospheric phase error correction and PS point recognition can be directly achieved by performing phase difference on adjacent pixels; the three-dimensional imaging result is more accurate, and noisy points are fewer.

Description

Technical Field

[0001] This invention belongs to the field of Synthetic Aperture Radar (SAR) data processing, specifically involving a two-layer network atmospheric phase error correction and altitude estimation method based on an optimized reference network. Background Technology

[0002] Synthetic Aperture Radar Tomography (TomoSAR) imaging technology has been used for a long time. During spaceborne TomoSAR imaging, SAR images acquired by the system on different expeditions are affected by atmospheric phase errors. Zhu and Dong et al. mentioned that these errors are closely related to factors such as temperature, humidity, air pressure, and the total electron content of the ionosphere. These atmospheric phase errors can be approximated as exhibiting random variations similar to white noise, fluctuating with the observation time. This random error reduces the coherence between tomographic signals, making it difficult to effectively focus the signals in the altitude direction, thus significantly impacting the quality of TomoSAR imaging.

[0003] Zhu proposed using a traditional Delaunay triangulation as a reference network to correct atmospheric phase errors, rather than using single permanent scatterer (PS) points. This approach achieves accurate atmospheric phase error correction. However, traditional Delaunay triangulation has high redundancy, increasing the computational burden during height estimation. This invention only discusses two types of PS points: single permanent scatterer (SPS) points and double permanent scatterer (DPS) points. The Delaunay triangulation is optimized using a minimum spanning tree (MST), reducing redundancy while ensuring uniform coverage of the entire imaging scene, thus achieving atmospheric phase error correction for large scenes. In subsequent star-shaped network connections, points in the reference network are used for phase difference correction to correct the atmospheric phase, significantly reducing the influence of the atmospheric phase screen on the signal phase and making the TomoSAR imaging results more accurate. In the subsequent relative height estimation, the present invention employs weighted least squares estimation to estimate the relative height of an SPS point in the reference network relative to a selected SPS point. Summary of the Invention

[0004] Purpose of the invention: To reduce the impact of atmospheric phase error on the quality of 3D imaging, this invention proposes a two-layer atmospheric phase error correction and height estimation method based on an optimized reference network. By utilizing the advantage of the reference network uniformly covering the scene and combining it with a star-shaped network, accurate atmospheric phase error correction can be achieved.

[0005] Technical solution: The present invention provides a two-layer network atmospheric phase error correction and altitude estimation method based on an optimized reference network, comprising the following steps:

[0006] (1) Acquire multiple pre-processed two-dimensional single-view complex image data in advance;

[0007] (2) Extract candidate points of single permanent scatterer (SPS) and construct an initial Delaunay reference network. Use the constructed and connected phase difference signals to correct atmospheric phase error. Perform minimum spanning tree optimization based on the residual energy ratio of each edge to generate the final reference network.

[0008] (3) Weighted least squares estimation is used to estimate the height of the SPS points in the final reference network relative to a certain reference SPS point, which is used for atmospheric phase error correction and height estimation of the subsequent star network;

[0009] (4) Extract the permanent scatterer PS candidate points and connect them with the neighboring SPS points in the reference grid to generate a star grid. After atmospheric phase error correction, complete the height estimation of the PS points.

[0010] (5) Generate elevation map results based on the height information of all points obtained from the double-layer network.

[0011] Furthermore, the process for extracting candidate points for a single permanent scatterer (SPS) in step (2) is as follows:

[0012] Selecting SPS candidate points using the amplitude deviation index thresholding method:

[0013] ,

[0014] in, Indicates the number of observations. Indicates the first The pixel location corresponding to the next observation data The amplitude value at a certain point; by setting a fixed threshold, candidate points for SPS are selected. When the ADI of a certain point is less than the threshold, that point is selected as a candidate point for SPS.

[0015] Furthermore, the process of generating the final reference net in step (2) is as follows:

[0016] The atmospheric phase difference between endpoints is controlled using a side length thresholding method, retaining edges in the initial reference network whose lengths are less than the threshold. When the spatial distance between two SPS points in the reference network is small, their atmospheric phase delays are likely to be similar. The differential signal obtained by phase difference between the two points is:

[0017] ,

[0018] in, and They represent the first The signal observations of the starting and ending points of the corresponding edge during the next voyage. Indicates the end point Complex scattering coefficients of the second observation This represents the phase difference between the complex scattering coefficients at two points. This represents the height difference between the two endpoints. The spatial sampling rate is used to obtain the differential signal. Then, the differential signals of the remaining sides are used to perform three-dimensional imaging to obtain the relative height. ;

[0019] Calculate the residual energy ratio (RSR) of each individual permanent scatterer edge:

[0020] ,

[0021] in, It is an estimate of the phase difference signal on one side of the signal. For SPS, In For the height difference between the two endpoints, the smaller the RSR, the better the match between the observed and estimated signal values.

[0022] Delaunay is optimized using the RSR value of each edge as the edge weight to obtain the final reference network.

[0023] Furthermore, the reference SPS point mentioned in step (3) is a ground point in the candidate scene.

[0024] Furthermore, the implementation process of step (3) is as follows:

[0025] The relative height relationship between the SPS points in the final reference network and a certain reference SPS point is as follows:

[0026] ,

[0027] ,

[0028] in, It is The column vector represents the relative height of all high-quality SPS edges in the reference network. It is The column vector represents the actual height of the corresponding SPS point. It is The matrix represents the transformation matrix from the SPS point height to the relative height of the SPS edge; It is diagonal matrix , Indicates the first The RSR value corresponding to each SPS edge needs to be selected before calculation. Use a certain SPS point as a reference point and remove it. The corresponding column.

[0029] Furthermore, the matrix The elements in the array are only 0, 1, and -1. In each row, -1 represents the starting point of the corresponding SPS edge, 1 represents the ending point, and the rest are 0.

[0030] Furthermore, the implementation process of step (4) is as follows:

[0031] Set an amplitude threshold condition to filter out candidate points that are not single permanent scatterers and remove single permanent scatterer points from the reference net;

[0032] Connect the non-single permanent scatterer candidate points to the nearest single permanent scatterer point in the reference network and perform phase difference calculation. In the star network, single permanent scatterer edges and SPS-DPS edges need to be retained. Calculate the phase difference signals of the SPS edges and SPS-DPS edges, where the phase difference signal of the SPS-DPS edge is:

[0033] ,

[0034] in, Indicates the first DPS endpoint The phase difference between the complex scattering coefficients of each scattering element and the initial SPS point Indicates the first DPS endpoint The height between each scattering element and the initial SPS point is obtained by 3D imaging, which yields the relative heights of the SPS edge and the SPS-DPS edge.

[0035] Add the relative height to the actual height of the SPS point in the reference network to obtain the actual height of the SPS point and DPS point extracted from the star network.

[0036] Furthermore, the dual-layer network includes a first-layer reference network and a second-layer star network.

[0037] Beneficial effects: Compared with the prior art, the beneficial effects of the present invention are: the present invention does not require pre-separation of atmospheric phase, and atmospheric phase error correction and PS point identification can be directly achieved by phase difference between adjacent pixels; the present invention makes the three-dimensional imaging results more accurate and has less noise. Attached Figure Description

[0038] Figure 1 This is a flowchart of the present invention;

[0039] Figure 2 Spatiotemporal baseline plot for ascending orbit data;

[0040] Figure 3 For the spatiotemporal baseline map of the down-orbit data;

[0041] Figure 4 The elevation estimation results were obtained by using traditional atmospheric phase error correction methods for the orbital data;

[0042] Figure 5 The elevation estimation results of the method of this invention are used for the orbital data;

[0043] Figure 6 The elevation estimation results were obtained by using traditional atmospheric phase error correction methods for the down-orbit data;

[0044] Figure 7 The elevation estimation results of the method of this invention are used for the orbit reduction data. Detailed Implementation

[0045] The present invention will now be described in further detail with reference to the accompanying drawings.

[0046] like Figure 1 As shown, this invention proposes a two-layer network atmospheric phase error correction and altitude estimation method based on an optimized reference network. The specific implementation steps are as follows:

[0047] Step 1: Obtain the preprocessed data Two-dimensional single-view complex image data .

[0048] Step 2: Extract SPS candidate points and construct an initial Delaunay reference network. Use the constructed connected phase difference signals to correct atmospheric phase errors, and perform MST optimization based on the residual-to-signal ratio (RSR) of each edge to generate the final reference network.

[0049] (2.1) The reference network was created to find SPS points distributed throughout the scene, providing a reference for subsequent atmospheric phase error correction. Candidate SPS points were selected using the Amplitude Dispersion Index (ADI) thresholding method:

[0050] ,

[0051] in, Indicates the number of observations. Indicates the first The pixel location corresponding to the next observation data The amplitude value at a given point. The ADI threshold method filters SPS candidate points by setting a fixed threshold. When the ADI of a point is less than the threshold, that point is selected as an SPS candidate point.

[0052] (2.2) Construct a Delaunay initial reference network using the selected SPS candidate points. Use the edge length thresholding method to control the atmospheric phase difference between endpoints, retaining edges in the initial reference network with edge lengths less than the threshold. When the spatial distance between two SPS points in the reference network is small, the atmospheric phase delay between them is likely to be similar. The differential signal obtained by phase difference between the two points is:

[0053] ,

[0054] in and They represent the first The signal observations of the starting and ending points of the corresponding edge during the next voyage. Indicates the end point Complex scattering coefficients of the second observation This represents the phase difference between the complex scattering coefficients at two points. This represents the height difference between the two endpoints. The spatial sampling rate is used to obtain the differential signal. Then, the differential signals of the remaining sides are used to perform three-dimensional imaging to obtain the relative height. .

[0055] (2.3) Calculate the RSR of each SPS edge:

[0056] ,

[0057] in, It is an estimate of the phase difference signal on one side of the signal. For SPS, In For the height difference between the two endpoints, the smaller the RSR, the better the match between the observed and estimated signal values.

[0058] (2.4) Use the RSR value of each edge as the edge weight value to perform MST planning on Delaunay to obtain the final reference network.

[0059] Step 3: Use weighted least squares (WLS) estimation to estimate the height of the SPS points in the final reference network relative to a certain reference SPS point. The relative height relationship can be expressed as:

[0060] ,

[0061] ,

[0062] in, It is The column vector represents the relative height of all high-quality SPS edges in the reference network. It is The column vector represents the actual height of the corresponding SPS point. It is The matrix represents the transformation matrix from the height of an SPS point to the relative height of an SPS edge. Its elements are only 0, 1, and -1. In each row, -1 indicates the starting point of the corresponding SPS edge, 1 indicates the ending point, and the rest are 0. It is diagonal matrix , Indicates the first The RSR value corresponding to each SPS edge needs to be selected before calculation. Use a certain SPS point as a reference point and remove it. The corresponding column.

[0063] Step 4: Extract PS candidate points and connect them with neighboring SPS points in the reference network to generate a star network. After atmospheric phase error correction, complete the height estimation of the network.

[0064] Amplitude threshold conditions are set to filter out PS candidate points and remove SPS points from the reference network. The PS candidate points are connected to their nearest SPS points in the reference network and phase difference is performed. In the star network, SPS edges and SPS-DPS edges need to be retained. The phase difference signal of the SPS edge has been explained in step (2.2). The phase difference signal of the SPS-DPS edge is:

[0065] ,

[0066] in, Indicates the first DPS endpoint The phase difference between the complex scattering coefficients of each scattering element and the initial SPS point Indicates the first DPS endpoint The height between each scattering element and the initial SPS point is obtained by 3D imaging, which yields the relative heights of the SPS edge and the SPS-DPS edge.

[0067] Add the obtained relative height to the actual height of the SPS point in the reference network to obtain the actual height of the SPS point and DPS point extracted in the star network.

[0068] Step 5: Generate an elevation map based on the height information of all points obtained from the two-layer network (the first layer is the reference network, and the second layer is the star network).

[0069] The proposed method is experimentally verified using data from the Fucheng-1 satellite. The Fucheng-1 satellite data observes the Magdeburg urban area in Germany, a region with relatively flat terrain and mostly low-rise buildings. The dataset used in this invention is C-band data, containing data collected during fourteen ascent passes and fifteen descent passes, with the spatiotemporal baseline as shown below. Figure 2 , Figure 3 As shown, the slant distance of the center of the ascending orbit data scene is 587,306.682 meters, and the center of the descending orbit data scene is 596,961.248 meters.

[0070] To verify the feasibility and effectiveness of the method of the present invention, atmospheric phase error correction and altitude estimation experiments were conducted using the present invention. Figures 4 to 7 The elevation estimation results of the traditional atmospheric phase error correction method and the elevation estimation results of the method of this invention are shown respectively for ascending and descending orbit data. The results show that the height of the PS point using the method of this invention remains unchanged along the horizontal contour of the building and changes uniformly along the vertical contour. Moreover, there are basically no noise points with abrupt height changes in the whole scene compared with the traditional method, reflecting the effectiveness of atmospheric phase error correction.

[0071] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for atmospheric phase error correction and altitude estimation based on a two-layer network with an optimized reference network, characterized in that, Includes the following steps: (1) Acquire multiple pre-processed two-dimensional single-view complex image data in advance; (2) Extract candidate points of single permanent scatterer (SPS) and construct an initial Delaunay reference network. Use the constructed and connected phase difference signals to correct atmospheric phase error. Perform minimum spanning tree optimization based on the residual energy ratio of each edge to generate the final reference network. (3) Weighted least squares estimation is used to estimate the height of the SPS points in the final reference network relative to a certain reference SPS point, which is used for atmospheric phase error correction and height estimation of the subsequent star network; (4) Extract the permanent scatterer PS candidate points and connect them with the neighboring SPS points in the reference grid to generate a star grid. After atmospheric phase error correction, complete the height estimation of the PS points. (5) Generate elevation map results based on the height information of all points obtained from the double-layer network.

2. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The process of extracting candidate points for a single permanent scatterer (SPS) in step (2) is as follows: Selecting SPS candidate points using the amplitude deviation index thresholding method: ， in, Indicates the number of observations. Indicates the first The pixel location corresponding to the next observation data The amplitude value at a certain point; by setting a fixed threshold, SPS candidate points are selected. When the ADI of a certain point is less than the threshold, that point is selected as an SPS candidate point.

3. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The process of generating the final reference net in step (2) is as follows: The atmospheric phase difference between endpoints is controlled using a side length thresholding method, retaining edges in the initial reference network whose lengths are less than the threshold. When the spatial distance between two SPS points in the reference network is small, their atmospheric phase delays are likely to be similar. The differential signal obtained by phase difference between the two points is: ， in, and They represent the first The signal observations of the starting and ending points of the corresponding edge during the next voyage. Indicates the end point Complex scattering coefficients of the second observation This represents the phase difference between the complex scattering coefficients at two points. This represents the height difference between the two endpoints. The spatial sampling rate is used to obtain the differential signal. Then, the differential signals of the remaining sides are used to perform three-dimensional imaging to obtain the relative height. ; Calculate the residual energy ratio (RSR) of each individual permanent scatterer edge: ， in, It is an estimate of the phase difference signal on one side of the signal. For SPS, In For the height difference between the two endpoints, the smaller the RSR, the better the match between the observed and estimated signal values. Delaunay is optimized using the RSR value of each edge as the edge weight to obtain the final reference network.

4. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The reference SPS point mentioned in step (3) is a ground point in the candidate scene.

5. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The implementation process of step (3) is as follows: The relative height relationship between the SPS points in the final reference network and a certain reference SPS point is as follows: ， ， in, It is The column vector represents the relative height of all high-quality SPS edges in the reference network. It is The column vector represents the actual height of the corresponding SPS point. It is The matrix represents the transformation matrix from the SPS point height to the relative height of the SPS edge; It is diagonal matrix , Indicates the first The RSR value corresponding to each SPS edge needs to be selected before calculation. Use a specific SPS point as a reference point and remove it. The corresponding column.

6. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network using a two-layer network according to claim 5, characterized in that, The matrix The elements in the array are only 0, 1, and -1. In each row, -1 represents the starting point of the corresponding SPS edge, 1 represents the ending point, and the rest are 0.

7. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The implementation process of step (4) is as follows: Set an amplitude threshold condition to filter out candidate points that are not single permanent scatterers and remove single permanent scatterer points from the reference net; Connect the non-single permanent scatterer candidate points to the nearest single permanent scatterer point in the reference network and perform phase difference calculation. In the star network, single permanent scatterer edges and SPS-DPS edges need to be retained. Calculate the phase difference signals of the SPS edges and SPS-DPS edges, where the phase difference signal of the SPS-DPS edge is: ， in, Indicates the first DPS endpoint The phase difference between the complex scattering coefficients of each scattering element and the initial SPS point Indicates the first DPS endpoint The height between each scattering element and the initial SPS point is obtained by 3D imaging, which yields the relative heights of the SPS edge and the SPS-DPS edge. Add the relative height to the actual height of the SPS point in the reference network to obtain the actual height of the SPS point and DPS point extracted from the star network.

8. The method for atmospheric phase error correction and altitude estimation based on an optimized reference network in a two-layer network according to claim 1, characterized in that, The two-layer network includes a first-layer reference network and a second-layer star network.

Images