SAR radiation cross calibration method based on urban wide strong scattering point reference

Abstract

The invention discloses an SAR radiation cross calibration method based on an urban wide strong scattering reference point target, and the method comprises the steps: analyzing a meta-file of to-be-calibrated SAR data, extracting corresponding system parameter information and imaging region information, selecting multi-scene reference SAR data from calibrated historical SAR data, and extracting an absolute calibration constant; cutting and registering an area which coincides with the imaging area of the reference SAR data and the to-be-calibrated SAR data and is located in an urban area to make an SSDH data set; carrying out one-round screening and two-round screening on the SSDH data set based on a standard strong scattering point model, and extracting a radiation-stable high-value reference point target appearing in a full time domain; rCS data of the reference point target in the to-be-calibrated SAR data and RCS data of the reference point target in the reference SAR data are calculated; and comparing the RCS data of the same reference point target in the to-be-calibrated SAR data with the RCS data of the calibrated reference SAR data, and performing cross calibration to obtain an absolute radiometric calibration constant of the to-be-calibrated SAR data. According to the invention, low-cost and normalized SAR radiometric calibration processing is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to a SAR calibration technology, in particular to a SAR radiation cross calibration method based on a wide strong scattering point reference in an urban area. BACKGROUND

[0002] Synthetic Aperture Radar (SAR) achieves high-resolution imaging on the ground by actively transmitting and receiving electromagnetic waves combined with signal processing technology, and has the ability to observe the ground at all times and in all weather, and has been widely used in geological disaster monitoring and evaluation, ocean, agriculture and other fields. With the development of SAR technology, SAR data acquisition including multi-band, multi-imaging mode, multi-wave position has been greatly enriched. The development of spaceborne SAR has emerged a star cluster mode to improve the continuous observation capability. Good SAR radiation data quality and consistency of observation data are the basis for various quantitative applications, which rely on SAR data calibration processing to achieve.

[0003] Traditional SAR systems use internal calibration combined with pre-deployed backscatter cross section (RCS) accurately set by the external calibration of the calibration group to ensure high-precision radiation data quality of SAR. In addition, natural distribution targets such as Amazon rainforest with uniform and consistent backscatter coefficient priori known are also used for radiation calibration. By establishing a direct connection between the DN value of the SAR image and the RCS of the reference object, absolute radiation calibration is completed. Depend on accurate radar repeater or corner reflector (CR) and other artificial calibration equipment as a calibrator, such as the SAR calibration center of the German Aerospace Center (DLR) undertakes the calibration work of multi-satellite SAR system, the accuracy is better than 0.2dB (1σ). Through the global deployment of CR and active radar calibrators, ALOS-2 is radiometrically calibrated and evaluated, with a stability and radiometric accuracy of less than 0.8dB (1σ). The rainforest distribution target is used for antenna pattern calibration verification and long-term radiation level monitoring, such as ERS-2, Sentinel-1, ALOS-2, etc. In general, the rainforest distribution target has a radiation stability of 0.2-0.3dB. However, the traditional radiation external calibration method has certain limitations, including the need for a large amount of manpower and material resources in deployment and maintenance, the frequency range of calibration activities is limited, and the consistency of data is caused by different reference calibration devices in different calibration activities. With the development of SAR constellation star cluster, the workload of SAR radiation calibration and radiation quality monitoring has increased explosively, and the traditional radiation calibration method is more and more inadequate.

[0004] In view of the bottleneck problem of data quality assurance caused by the large development and application of SAR, there are two existing processing methods, one is to construct multiple calibration fields and use remote adjustable orientation angle in the manufacture of calibrator to improve the calibration ability of the calibrator, but this strategy cannot fundamentally solve the bottleneck problem. The second is to use non-cooperative targets in the scene as reference targets for SAR radiation calibration, or to use cross calibration of already calibrated SAR data. First, the permanent scatterer (PS) maintains high coherence under long time and spatial baseline, and can be used for SAR radiation calibration due to its high radiation stability. Based on ERS-2 data, the system stability is less than 0.2 dB by using PS point auxiliary calibration. In addition, based on the stable radiation characteristics or statistical parameters of special explicit target objects, such as power transmission towers, signal towers, offshore oil wells, urban corner targets, etc. Yang et al. found that the radiation stability of the median statistical parameters in the city, and used it for radiation calibration. These technologies have the problems of low accuracy, non-universality of reference targets, complex method, low controllability compared with the calibrator method, but to some extent, they get rid of the limitation of the calibrator, and have the possibility of hardware-free, real-time, normal, and automatic calibration. Under the background of existing technologies and clear SAR radiation calibration applications, the method proposed in the present application can be used for normal SAR data quality monitoring and multi-satellite SAR data consistency calibration in the life cycle of SAR system. SUMMARY

[0005] The purpose of the present application is to provide a SAR radiation cross calibration method based on widely strong scattering point reference in urban area.

[0006] The technical solution for achieving the purpose of the present application is: a SAR radiation cross calibration method based on widely strong scattering point reference in urban area, comprising the following steps:

[0007] Step 1, read the SAR data to be calibrated and the SAR historical data that has been calibrated, parse the metafile of the SAR data to be calibrated to extract the corresponding system parameter information and imaging area information, select multiple reference SAR data from the historical SAR data that has been calibrated, and extract the absolute calibration constant;

[0008] Step 2, clip the area that is overlapped in the imaging area of the reference SAR data and the SAR data to be calibrated and located in the urban area, and perform registration to obtain an SSDH data set;

[0009] Step 3, perform one round of screening and two rounds of screening on the SSDH data set based on the standard strong scattering point model, and identify and extract the high-value reference point targets with stable radiation appearing in the full time domain;

[0010] Step 4, calculate the RCS data of the reference point target in the SAR data to be calibrated and the RCS data of the reference point target in the reference SAR data;

[0011] Step 5, compare the RCS data of the same reference point target in the SAR data to be calibrated and the RCS data of the reference SAR data, cross-calibrate to obtain the absolute radiometric calibration constant of the SAR data to be calibrated.

[0012] Further, step 1, read the SAR data to be calibrated and the SAR historical data that has been calibrated, parse the metafile of the SAR data to be calibrated to extract the corresponding system parameter information and imaging area information, select multi-scene reference SAR data from the historical SAR data that has been calibrated, and extract the absolute calibration constant, wherein:

[0013] The system parameter information includes the look angle azimuth, the elevation track, the side view, the radar frequency, the working mode, and the resolution;

[0014] The imaging area information includes the longitude and latitude information of the imaging area and the urban area information of the imaging scene.

[0015] Further, the multi-scene reference SAR data is selected from the historical SAR data that has been calibrated, and the selection conditions include:

[0016] 1) The reference SAR data and the SAR data to be calibrated need to have the same radar frequency, the same wave position, the same resolution, the same elevation track, and the same side view,

[0017] 2) The reference SAR data and the SAR data to be calibrated need to image the same urban area, that is, the imaging ranges overlap;

[0018] 3) The target and the SAR radar look angle deviation of the reference SAR data and the SAR data to be calibrated does not exceed ±5°, and the target and the SAR radar azimuth deviation does not exceed ±5°.

[0019] Further, step 2, clip the area where the imaging areas of the reference SAR data and the SAR data to be calibrated overlap and are located in the urban area, and perform registration to obtain the SSDH data set, and the specific method is:

[0020] First, clip the SAR data to be calibrated and the multi-scene reference SAR data imaged in the same urban area to obtain a data set imaged in the same area, which includes the area clipping data of the SAR data to be calibrated at the latest time and the area clipping data of the reference SAR data at multiple historical times;

[0021] Next, the regional cropped data of the SAR data to be calibrated and the regional cropped data of the reference SAR data at multiple historical moments are registered to create the SSDH dataset. That is, the SAR data to be calibrated is selected as the main image and the reference SAR data is used as the auxiliary image. Fourier transform is performed on the main image and the auxiliary image to extract the phase information of the cross power spectrum. The pixel offset of the auxiliary image in the horizontal and vertical dimensions is calculated. The pixel offset in the two dimensions is compensated by translation of the auxiliary image to complete the registration process. The SAR dataset after registration is the SSDH dataset.

[0022] Furthermore, in step 3, the SSDH dataset is screened in two rounds based on the standard strong scattering point model to identify and extract high-value reference point targets with stable radiation that appear throughout the entire time domain. The specific method is as follows:

[0023] First, the SAR data for each scene in the SSDH dataset, including SAR data to be calibrated and multi-scene reference SAR data, undergoes a first round of filtering. The filtering criteria for the first round are as follows:

[0024]

[0025] In the formula, the target amplitude of the candidate point, Amp_Ti, is required to be the maximum value within a local square area with a side length of 40 pixels, which is greater than the scene mean value by a factor of TH_Point. Furthermore, the maximum value of the four background regions of 15*15 pixels centered on the point (top left, bottom left, top right, and bottom right) is less than the amplitude value by a factor of TH_Background. Area represents the local square area of ​​20 pixels before and after the target point. Mean_Scene is the mean value of the scene amplitude. Background represents the four background regions of 15*15 pixels centered on the point (top left, bottom left, top right, and bottom right).

[0026] Next, iterate through all candidate point targets obtained in the first round of screening, extract a 36*36 local slice centered on the point target, upsample it by 8 times, and then divide the local slice into three equal parts along the horizontal and vertical axes to obtain an average of 9 regions. Then, conduct the second round of screening, where the screening conditions are as follows:

[0027]

[0028] In the formula, after upsampling, the amplitude of the candidate point target is greater than TH_MM times the maximum amplitude of the remaining 8 annular blocks excluding the central block. For the three-dimensional structure of the candidate point scattering, cross sections are cut at 0.3 / 0.4 / 0.5 / 0.6 / 0.7 / 0.8 / 0.9 / 0.95 times the amplitude of the candidate point target. The cross sections at these heights are all single connected domains, that is, they pass the second round of screening, otherwise they are eliminated. NUM(Slice(·)) is the number of closed connected domains in the computation graph.

[0029] Then, for the candidate point targets obtained after two rounds of screening in each scene, the same point identification and marking are performed. If the position offset of the candidate point target in the scene at different times is less than a certain threshold, it is considered to be a point target caused by the scattering and reflection of the same entity.

[0030] Finally, points with the same name that exist in all scenes in the time domain are selected as the final high-value reference point targets.

[0031] Further, in step 4, the RCS data of the extracted reference point target in the SAR data to be calibrated and the RCS data in the reference SAR data are calculated. The specific method is as follows:

[0032] First, the RCS data of the reference point target in the SAR data to be calibrated is calculated using the integral pixel method;

[0033] Next, the RCS values ​​of each reference point target in the reference SAR data are calculated in different time domain images, and the standard deviation STD_i of all RCS values ​​of a reference point in the time domain is calculated. If the standard deviation is greater than the set threshold STD_TH, such a point target is considered not a reliable reference point target and is removed.

[0034] Finally, the mean value of the RCS of the valid reference point target in the reference SAR data at all times is calculated as the radiation benchmark of the reference point target.

[0035] Further, in step 5, the RCS data of the same reference point target in the SAR data to be calibrated is compared with the RCS data of the SAR data already calibrated, and the absolute radiometric calibration constant of the SAR data to be calibrated is obtained through cross-calibration. The specific method is as follows:

[0036] The calibration constant K for the SAR data to be calibrated is:

[0037] K = K 0 dB +(10log 10 P 1 -10log 10 P 0 (7)

[0038] K 0 dB P represents the absolute calibration constant of the calibrated SAR data. 0 For the RCS data of the reference point target in the calibrated SAR data, P 1 RCS data for reference point targets in uncalibrated SAR data.

[0039] A SAR radiometric cross-calibration system based on a widely used strong scattering reference point target in an urban area includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the SAR radiometric cross-calibration method based on a widely used strong scattering reference point target in an urban area, thereby achieving SAR radiometric cross-calibration based on a widely used strong scattering reference point target in an urban area.

[0040] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the SAR radiometric cross-calibration method based on a widely used strong scattering reference point target in an urban area, thereby achieving SAR radiometric cross-calibration based on a widely used strong scattering reference point target in an urban area.

[0041] 1. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the SAR radiometric cross-calibration method based on a widely strong scattering reference point target in an urban area is implemented, thereby achieving SAR radiometric cross-calibration based on a widely strong scattering reference point target in an urban area.

[0042] Compared with existing technologies, the significant advantages of this invention are: (1) It utilizes the cross-calibration processing approach, and the entire calibration process does not require an external calibration field. With sufficient data support, the absolute radiometric calibration of the SAR system can be completed based on data processing, achieving low-cost and routine SAR radiometric calibration processing. (2) It proposes a complete SAR absolute radiometric calibration method based on a wide range of targets in the scenario. Facing the bottleneck of a large number of radiometric calibration needs for future SAR constellations, it alleviates the shortage of calibration resources in the SAR constellation radiometric calibration field, realizes rapid and low-cost absolute radiometric calibration of SAR stars, and can be used for long-term monitoring of SAR radiometric quality, monitoring of radiometric consistency between SAR systems, etc., to ensure SAR radiometric quality to support the effectiveness of various quantitative applications of SAR. (3) It proposes a novel and widely applicable urban strong scattering point radiation stability reference target, realizing the full exploration of the radiation mechanism characteristics of strong tertiary and strong secondary scatterers formed by building corners, etc., which is mutually verified with the permanent scatterer method and inspires related subsequent applications. Attached Figure Description

[0043] Figure 1 This is a flowchart of a SAR radiation cross-calibration method based on a wide range of strong scattering points in urban areas.

[0044] Figure 2 A three-dimensional plot of the upsampling amplitude of a local region of an ideal strong scattering point target model.

[0045] Figure 3 The image shows a 3D plot of the upsampling amplitude and an amplitude contour plot for strong scattering at non-ideal points.

[0046] Figure 4 The images show the optical and SAR images corresponding to the strong scattering point structures (the top image shows the secondary reflection formed by streetlights on a flat parking lot, and the bottom image shows the tertiary reflection formed by the corner points of a building wall).

[0047] Figure 5 This is a schematic diagram illustrating the RCS region extraction of point targets in SAR images using the integral method.

[0048] Figure 6 Optical images of the GF-3 historical data imaging area and high-coverage airport area were extracted.

[0049] Figure 7 This is the result of extracting candidate points for the first scene of strong scattering in the time series.

[0050] Figure 8 This is the result of extracting candidate points for strong scattering in the fifth time-series scene.

[0051] Figure 9 To screen and obtain the radiation time-domain performance of multiple effective radiation-stable reference point targets.

[0052] Figure 10 The processing results for scene 10 (scene to be calibrated) are calculated for radiation cross-calibration. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] Natural ground object SAR signals often exhibit limited intensity and broad distribution, consisting of multiple scattering components including volume scattering. In contrast, man-made objects in urban areas often possess advantages such as strong scattering (point targets), high signal-to-noise ratio, clear scattering mechanisms, and less susceptibility to external environmental influences like weather and precipitation. These advantages make urban targets suitable as reference objects for SAR data calibration. The core of this invention is to identify a class of ground objects with consistently stable radar radiation that are widely present in urban areas, serving as a reference for SAR system absolute radiation gain calibration. Furthermore, by leveraging the cross-calibration concept commonly found in other sensors such as hyperspectral sensors, the invention addresses the issue of multi-satellite radiation consistency, enabling gain adjustment and correction of the radiation calibration constant for multi-satellite SAR data products, as well as long-term, routine radiation gain monitoring.

[0055] Based on this idea, this invention provides a SAR radiometric cross-calibration method based on widely distributed, strongly scattering reference point targets in urban areas. Specifically, this invention proposes a novel method that uses widely distributed, radiometrically stable, strongly scattering point targets in urban areas as radiometric calibration references. By comparing the difference in the reference RCS of the point targets between the absolutely radiometrically calibrated SAR image and the SAR image to be calibrated using the cross-calibration approach, the absolute calibration constant of the SAR image to be calibrated is calculated. Figure 1 As shown, it includes:

[0056] Step A:

[0057] Step A marks the beginning of the processing flow of this invention. In this step, the SAR data to be calibrated and the historically calibrated SAR data used as a reference are read in.

[0058] Specifically, the L1A level single-look complex data (hereinafter referred to as SAR data to be calibrated) of the SAR data source to be calibrated is first read. This is the L1A level SAR data obtained by imaging the radar echo, which has not undergone radiometric calibration and geometric correction.

[0059] Next, by parsing the metadata of the SAR data to be calibrated, the key system parameter information (such as the azimuth angle, ascent and descent, side look, radar frequency, working mode, and resolution) and imaging area information (latitude and longitude information of the imaging area and urban area information of the imaging scene) of the SAR data to be calibrated are extracted.

[0060] Then, using the system parameter information and imaging area information of the SAR data to be calibrated as described above, numerous historical SAR data from the already calibrated SAR data source are retrieved to obtain multiple scenes of already calibrated SAR data that can be used as references (hereinafter referred to as reference SAR data). To ensure the effectiveness and accuracy of the cross-radiometric calibration of this invention, the selected reference SAR data must have the same radar frequency, same wavefront, same resolution, same ascent / descent trajectory, and same side-look as the SAR data to be calibrated, and must image the same urban area, i.e., their imaging ranges overlap to some extent. Furthermore, the downward viewing angle deviation between the target and the SAR radar in the SAR data to be calibrated and the reference SAR data should not exceed ±5°, and the azimuth angle deviation between the target and the SAR radar should not exceed ±5°. Preferably, the resolution of the suitable SAR data in this invention should be within 10 meters.

[0061] Finally, the retrieved reference SAR data that meets the above restrictions are read, and their absolute calibration constant K is extracted as a reference for the absolute calibration constant of the final SAR data to be calibrated obtained by the radiation cross-calibration of this invention. When a sufficient number of reference SAR data (greater than or equal to 3 scenes) are available, the data resource conditions can be considered to meet the radiation cross-calibration requirements of this invention, and the subsequent processing can proceed.

[0062] It should be noted that, in principle, the more reference SAR data the better, and the imaging time interval between the reference SAR data and the data to be calibrated should not be too large. This is because, although the reference target proposed in this invention is radiometrically stable in the time domain, data with excessively long time intervals may not guarantee the validity of the reference target. The reference SAR data proposed in this invention are not limited to using the same SAR data source, but it is required that the number of reference SAR data sources from the same SAR data source be greater than or equal to three.

[0063] Step B:

[0064] like Figure 1 As shown in the corresponding section, this step mainly involves cropping the areas where the imaging regions of the reference SAR data and the SAR data to be calibrated overlap and are located in urban areas, and then registering them to create the SSDH dataset.

[0065] First, the dataset of images of the same urban area is obtained by cropping the uncalibrated SAR data and multiple reference SAR data. This dataset includes the regional cropped data of the uncalibrated SAR data at the latest time and the regional cropped data of the reference SAR data at multiple historical times.

[0066] Next, the above dataset is converted into an SSDH dataset through registration processing. Specifically, the SAR data to be calibrated is selected as the main image, and the reference SAR data is used as the auxiliary image and registered with it. Fourier transform is performed on the main image and the auxiliary image to extract the phase information of the cross power spectrum. The pixel offset of the auxiliary image relative to the main image in the horizontal and vertical dimensions is calculated, and the pixel offset in the two dimensions is compensated by translation of the auxiliary image.

[0067] Finally, the SAR dataset obtained after simple registration becomes the SSDH dataset, which is then processed in subsequent steps.

[0068] It should be noted that, for high-precision radiometric cross-calibration, this invention employs a registration process of integer pixel-level translation of the original data without altering the actual scattering intensity of each pixel in the SAR data, thus reconstructing the registered auxiliary image without interpolation resampling. Furthermore, since the reference target in this invention is the strongest scattering point in a local region, any residual pixel-level registration errors in this step are permissible.

[0069] Step C:

[0070] like Figure 1 As shown, this step is the core technology of this invention. In a specific implementation, this step identifies and extracts high-value reference point targets with stable radiation that appear throughout the entire time domain through a first-round screening and a second-round screening based on a standard strong scattering point model.

[0071] First, for the SSDH dataset produced in the previous step, this step performs the first round of pixel-by-pixel filtering for the SAR data of each scene (including SAR data to be calibrated and multi-scene reference SAR data).

[0072] The following screening criteria are used to extract a batch of candidate target points.

[0073]

[0074] In this formula, the target amplitude at a given point must first be the maximum value within a local square area with a side length of 40 pixels, and greater than the scene mean multiplied by TH_Point. Furthermore, the maximum value of the four background regions (top left, bottom left, top right, and bottom right) within a 15x15 pixel area centered on the point must be less than the amplitude multiplied by TH_Background. TH_Spoint is typically set to 5-10, and TH_Background is set to around 0.25. In the formula, Amp_Ti represents the amplitude of the candidate target point, Area represents the local square area within 20 pixels before and after the target point, Mean_Scene represents the average scene amplitude, and Background represents the four background regions (top left, bottom left, top right, and bottom right) within a 15x15 pixel area centered on the point.

[0075] Next, in the second round of filtering, each candidate point target obtained in the first round of filtering is traversed, a local region centered on the point target is extracted, upsampled, and then the following second round of filtering is performed:

[0076]

[0077] Specifically, the extracted local region is a 36*36 local slice, upsampled by a factor of 8. After upsampling, the slice is divided into three equal parts along its horizontal and vertical axes, resulting in an average of 9 regions. The amplitude of the candidate point target is greater than TH_MM times the maximum amplitude of the remaining 8 annular regions excluding the central block. Here, the TH_MM parameter is set to 0.3. Next, single scattering center screening is performed. That is, for the three-dimensional structure of (x,y,Amp(amplitude)) scattered by the candidate point, cross-sections are extracted at 0.3 / 0.4 / 0.5 / 0.6 / 0.7 / 0.8 / 0.9 / 0.95 times the amplitude of the candidate point target. The cross-sections at these heights are all single connected domains, which can pass the second round of screening; otherwise, they are eliminated. In the formula, NUM(Slice(·)) is the number of closed connected domains in the computation graph, and Amp is the amplitude in the three-dimensional structure of (x,y,Amp(amplitude)) of the local region where the upsampled point target is located.

[0078] Then, for the candidate point targets obtained after two rounds of screening in each scene, corresponding point identification and labeling (the same candidate point target at each time step) are performed. That is, if the positional offset of the candidate point target in the scene at different time steps is less than a certain threshold, it is considered to be a target caused by scattering and reflection from the same entity. Here, the offset threshold is set to an absolute pixel distance of less than 15 pixels. Since the candidate point targets obtained in the above two rounds of screening are the maximum values ​​of a local 40*40 pixel interval, point targets that meet this offset threshold are the same target, i.e., corresponding points.

[0079] Finally, points with the same name that exist in all scenes in the time domain are selected as the final high-value reference points. Due to their highly stable radiation characteristics, these points serve as calibration references for establishing radiation intensity correlations between the SAR data to be calibrated and the reference SAR data in the radiation cross-calibration of this invention.

[0080] It should be noted that the one-round and two-round screening method used in this invention is based on actual operational efficiency considerations. The first round of screening has a small computational load but traverses all pixels, while the data in the second round of screening has undergone upsampling processing, resulting in a larger computational load, but it processes the small set of candidate points after the first round of screening. Both rounds of screening are based on the same standard strong scattering point model, such as... Figure 2 As shown, it is a three-dimensional radiation intensity diagram of a standard strong scattering point target. It can be seen that it satisfies characteristics such as a local maximum peak for strong scattering, a peak value much larger than the background scattering intensity, a peak value much larger than the signal sidelobe energy, and an isolated single-peak structure. The implementation of the two-round screening conditions in this invention is based on these characteristics. Figure 3The example shown represents a model of point targets that might not conform to the standard point target in reality. It's evident that this model clearly does not satisfy the condition that the number of connected components in the last clause is one. Furthermore, it should be noted that due to image quality issues, such as defocusing in individual images, the number of candidate point targets extracted from that scene will be significantly less than the average number from other scenes in SSDH. This can be clearly identified from the time-domain curve of the number of extracted candidate point targets. In practical applications, these individual anomalous reference image data are directly discarded.

[0081] In summary, the reference point targets extracted in this invention correspond to specific reflection and scattering structures of actual ground features, such as the secondary strong scattering structure formed by streetlight poles and the ground in a flat parking lot, and the strong scattering structure of a trihedral reflector-like structure formed by three wall corners. Their signals appear as standard strong scattering point targets in SAR images. Because these reflection and scattering structures also possess stable radiation characteristics, this invention employs a screening method based on a standard strong scattering point model to obtain such reference point targets that exist throughout the entire time domain.

[0082] Step D:

[0083] This step calculates the RCS of the extracted reference point target in the SAR data to be calibrated and the RCS in the reference SAR data, providing data for cross-calibration in subsequent steps.

[0084] First, calculate the RCS of the reference point target in the SAR data to be calibrated, such as... Figure 5 As shown, the integral pixel method is used to accurately estimate the target's backscattering power. The point target response energy is the energy of the point target's integral region minus the influence of background clutter energy. Figure 5 The blue area represents the background region, and the orange area represents the energy integration region of the point target and side lobes. The formula for calculating the RCS of a point target is as follows:

[0085]

[0086] Where, δ a ,δ r Let θ be the two-dimensional pixel interval of the corresponding image, θ be the local incident angle, A be the integration region of the target point, B be the background region, Na be the number of pixels in the integration region, Nb be the number of pixels in the background region, and DN be the amplitude value of the corresponding pixel.

[0087] Next, the RCS values ​​of each reference point target in the reference SAR data are calculated in different time domain images (the calculation method is exactly the same as that for calculating the RCS of the reference point target in the reference SAR data). The standard deviation STD_i of all RCS values ​​of a reference point in the time domain is calculated. If the standard deviation is greater than the set threshold STD_TH (optionally set to 1dB), then such a point target is considered not a reliable candidate point target and is removed.

[0088] Finally, the mean RCS values ​​of all valid reference points in the reference SAR data are calculated as the radiometric reference for that reference point. The RCS of the same reference point in the SAR data to be calibrated and the radiometric reference in the reference SAR data are used as input information for the final step of cross-calibration.

[0089] It should be noted that this step utilizes historical data from multiple scenes of the reference SAR data to verify the time-domain stability of the reference point target's radiation. In particular, because the reference of this invention is a non-cooperative urban strong scattering structure, the further elimination of reference points in this step helps to remove the influence of unexpected changes (such as damage to the typical secondary reflection structure caused by parking near streetlights) on the cross-calibration results.

[0090] Step E:

[0091] This step is the final step in the embodiment of the present invention. By comparing the RCS of the SAR data to be calibrated with that of the calibrated reference SAR data (i.e., the radiometric reference of the reference SAR data) at the same reference point target, cross-calibration is performed to obtain the absolute radiometric calibration constant of the SAR data to be calibrated. The principle of this radiometric cross-calibration is clearly explained in the technical solution.

[0092] Specifically, the calibration constant K of the SAR data to be calibrated calculated in this step is:

[0093] K = K 1 dB =K 0 dB +(10log 10 P 1 -10log 10 P 0 (7)

[0094] K 0 dB P is the absolute calibration constant of the calibrated data. 0 P represents the RCS of point targets in a calibrated image. 1 The RCS of the reference point target in the uncalibrated image.

[0095] It should be noted that the radiometric accuracy of this cross-calibration is shown in the following formula, i.e., K 1 dB The error is:

[0096] ΔK 1 dB =ΔK 0 dB +Δ(10log 10 P1 -10log 10 P 0 )+Δ10log 10 σ c (8)

[0097] Its ΔK 1 dB The middle value represents the absolute radiometric calibration accuracy of SAR after cross-calibration, ΔK. 0 dB To ensure the absolute radiometric accuracy of the reference star, i.e., the calibrated star, Δ10log 10 σ c The deviation between the two observed target's true RCS values ​​is determined by the target's radiation stability under the cross-calibration constraint. Δ(10log) 10 P 1 -10log 10 P 0 This method was introduced during the measurement and processing of target power in two SAR images. In order to eliminate the systematic deviation between the two measurements, the two measurement methods adopted the exact same RCS calculation method, that is, the same upsampling factor, the same slice size, the same integration area range, and the same background range integration measurement method.

[0098] Example

[0099] To verify the effectiveness of the present invention, the following experiment was conducted.

[0100] like Figure 6 As shown, the GF-3 historical data includes multiple revisit observations of the San Diego area in the United States, providing a wealth of relevant historical data. Selecting data with the same ascent and descent orbits, imaging modes, and wavefronts reveals that the imaging range is roughly the same, exhibiting similar azimuth angles. This rich historical data on key urban areas, ports, airports, etc., forms the data foundation for the successful implementation of this invention. Figure 6 This is a region where time-series observations of an airport area overlap significantly, providing sufficient reference targets. If there are insufficient reference targets, the processed area can be appropriately enlarged.

[0101] Figure 7 , 8The results of two rounds of strong scattering candidate point target extraction (yellow diamonds) for the first and fifth time-series scenes are presented, along with the radiation reference point target set (red hexagons) obtained through step D and verified for temporal RCS stability. It can be seen that the strong scattering candidate point target extraction results differ across scenes, including the number of targets, which is affected by imaging quality. Furthermore, through screening based on the full temporal existence of corresponding points and temporal radiation stability verification, the number of remaining red reference points decreases rapidly. However, theoretically, cross-calibration can be performed as long as even one such reference point exists. Multiple reference points can be mutually verified. Figure 9 The time-domain performance of RCS for several stable radiation reference points is presented, and the radiation time-domain stability of most reference targets is better than 0.5 dB(1σ).

[0102] Figure 10 The results of the radiation cross-calibration are presented. For the uncalibrated GF-3B satellite data (scene 10) from the same source, a stable radiation reference (green line) can be achieved from a radiation reference point corresponding to the first nine calibrated GF-3A scenes. This reference point represents the true RCS of the scatterer under the observation conditions. By calculating the RCS of the point target in the uncalibrated image, the difference between the two can be compared, yielding a deviation of 10.9 dB. This deviation is obtained by correcting the absolute calibration constant of the uncalibrated image to the calibration constant of GF3A by 10.9 dB.

[0103] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. It should be noted that implementations not illustrated or described in the drawings or the main text of the specification are forms known to those skilled in the art and are not described in detail. Furthermore, the definitions of the methods above are not limited to the specific methods mentioned in the embodiments; those skilled in the art can easily modify or substitute them, for example:

[0104] (1) In addition to the integral method, other classic methods such as the peak method can be used to calculate the RCS of a point target.

[0105] (2) Other methods can be used to perform image registration processing, such as image correlation-based methods and various other sub-pixel level fine registration processing.

[0106] Based on the above description, those skilled in the art should have a clear understanding of the SAR radiation cross-calibration method based on a reference of a wide range of strong scattering points in urban areas according to the present invention.

[0107] It should also be noted that, unless otherwise stated, the numerical parameters in this specification and the appended claims are approximate values ​​and can be changed according to the desired characteristics obtained from the content of this invention. Specifically, all figures used in the specification and claims to indicate the content of composition, reaction conditions, etc., should be understood to be modified by the term "about" in all cases. Generally, this means that there may be variations of ±10% in some embodiments, ±5% in some embodiments, ±1% in some embodiments, and ±0.5% in some embodiments.

[0108] Furthermore, unless specifically described or required to occur in a specific order, the order of the above steps is not limited to those listed above and can be varied or rearranged according to the desired design. Moreover, the above embodiments can be used in combination with each other or with other embodiments based on design and reliability considerations; that is, technical features from different embodiments can be freely combined to form more embodiments.

[0109] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0110] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the related devices according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0111] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A SAR radiometric cross-calibration method based on a widely distributed strong scattering reference point target in an urban area, characterized in that, Includes the following steps: Step 1: Read the SAR data to be calibrated and the historical SAR data that has already been calibrated; parse the metadata of the SAR data to be calibrated to extract the corresponding system parameter information and imaging area information; select multiple reference SAR data from the historical SAR data that has already been calibrated and extract the absolute calibration constants. Step 2: Cropping the areas in the urban area where the imaging regions of the reference SAR data and the SAR data to be calibrated overlap, and then registering them to create the SSDH dataset; Step 3: Based on the standard strong scattering point model, the SSDH dataset is screened in one round and two rounds to identify and extract high-value reference point targets with stable radiation that appear in the entire time domain. Step 4: Calculate the RCS data of the extracted reference point target in the SAR data to be calibrated and the RCS data in the reference SAR data. Step 5: Compare the RCS data of the same reference point target in the SAR data to be calibrated with the RCS data of the calibrated reference SAR data, and cross-calibrate to obtain the absolute radiometric calibration constant of the SAR data to be calibrated.

2. The SAR radiation cross-calibration method based on a widely dispersed strong scattering reference point target in an urban area according to claim 1, characterized in that, Step 1: Read the SAR data to be calibrated and the historical SAR data that has already been calibrated. Parse the metadata of the SAR data to be calibrated to extract the corresponding system parameter information and imaging area information. Select multiple reference SAR data from the historical SAR data that has already been calibrated and extract the absolute calibration constants, where: System parameter information includes the azimuth angle of the area's lower view, the riser / faller rail, the side view, the radar frequency, the operating mode, and the resolution; The imaging area information includes the latitude and longitude of the imaging area and the urban area information of the imaging scene.

3. The SAR radiation cross-calibration method based on a widely used strong scattering reference point target in an urban area according to claim 2, characterized in that, Multiple reference SAR data sets were selected from calibrated historical SAR data, with selection criteria including: 1) The reference SAR data and the SAR data to be calibrated need to have the same radar frequency, the same wavefront, the same resolution, the same ascent and descent orbits, and the same side-looking parameters. 2) The reference SAR data and the SAR data to be calibrated need to image the same urban area, that is, the imaging ranges overlap to some extent; 3) The downward viewing angle deviation between the target and the SAR radar in the reference SAR data and the SAR data to be calibrated shall not exceed ±5°, and the azimuth angle deviation between the target and the SAR radar shall not exceed ±5°.

4. The SAR radiation cross-calibration method based on a widely dispersed strong scattering reference point target in an urban area according to claim 1, characterized in that, Step 2: Cropping is performed on the overlapping imaging areas of the reference SAR data and the SAR data to be calibrated, especially those located in urban areas. Registration is then performed to create the SSDH dataset. The specific method is as follows: First, the uncalibrated SAR data and multiple reference SAR data images of the same urban area are cropped to obtain a dataset of images of the same area, which includes the area cropped data of the uncalibrated SAR data at the latest time and the area cropped data of the reference SAR data at multiple historical times. Next, the regional cropped data of the SAR data to be calibrated and the regional cropped data of the reference SAR data at multiple historical moments are registered to create the SSDH dataset. That is, the SAR data to be calibrated is selected as the main image and the reference SAR data is used as the auxiliary image. Fourier transform is performed on the main image and the auxiliary image to extract the phase information of the cross power spectrum. The pixel offset of the auxiliary image in the horizontal and vertical dimensions is calculated. The pixel offset in the two dimensions is compensated by translation of the auxiliary image to complete the registration process. The SAR dataset after registration is the SSDH dataset.

5. The SAR radiation cross-calibration method based on a widely used strong scattering reference point target in an urban area according to claim 1, characterized in that, Step 3: Based on the standard strong scattering point model, the SSDH dataset undergoes a first and second round of screening to identify and extract high-value reference point targets with stable radiation across the entire time domain. The specific method is as follows: First, the SAR data for each scene in the SSDH dataset, including SAR data to be calibrated and multi-scene reference SAR data, undergoes a first round of filtering. The filtering criteria for the first round are as follows: In the formula, the target amplitude of the candidate point, Amp_Ti, is required to be the maximum value within a local square area with a side length of 40 pixels, which is greater than the scene mean value by a factor of TH_Point. Furthermore, the maximum value of the four background regions of 15*15 pixels centered on the point (top left, bottom left, top right, and bottom right) is less than the amplitude value by a factor of TH_Background. Area represents the local square area of ​​20 pixels before and after the target point. Mean_Scene is the mean value of the scene amplitude. Background represents the four background regions of 15*15 pixels centered on the point (top left, bottom left, top right, and bottom right). Next, iterate through all candidate point targets obtained in the first round of screening, extract a 36*36 local slice centered on the point target, upsample it by 8 times, and then divide the local slice into three equal parts along the horizontal and vertical axes to obtain an average of 9 regions. Then, conduct a second round of screening, with the following criteria: In the formula, after upsampling, the amplitude of the candidate point target is greater than TH_MM times the maximum amplitude of the remaining 8 annular blocks excluding the central block. For the three-dimensional structure of the candidate point scattering, cross sections are cut at 0.3 / 0.4 / 0.5 / 0.6 / 0.7 / 0.8 / 0.9 / 0.95 times the amplitude of the candidate point target. The cross sections at these heights are all single connected domains, that is, they pass the second round of screening, otherwise they are eliminated. NUM(Slice(·)) is the number of closed connected domains in the computation graph. Then, for the candidate point targets obtained after two rounds of screening in each scene, the same point identification and marking are performed. If the position offset of the candidate point target in the scene at different times is less than a certain threshold, it is considered to be a point target caused by the scattering and reflection of the same entity. Finally, points with the same name that exist in all scenes in the time domain are selected as the final high-value reference point targets.

6. The SAR radiation cross-calibration method based on a widely used strong scattering reference point target in an urban area according to claim 1, characterized in that, Step 4: Calculate the RCS data of the extracted reference point target in the SAR data to be calibrated and the RCS data in the reference SAR data. The specific method is as follows: First, the RCS data of the reference point target in the SAR data to be calibrated is calculated using the integral pixel method; Next, the RCS values ​​of each reference point target in the reference SAR data are calculated in different time domain images, and the standard deviation STD_i of all RCS values ​​of a reference point in the time domain is calculated. If the standard deviation is greater than the set threshold STD_TH, such a point target is considered not a reliable reference point target and is removed. Finally, the mean value of the RCS of the valid reference point target in the reference SAR data at all times is calculated as the radiation benchmark of the reference point target.

7. The SAR radiation cross-calibration method based on a widely dispersed strong scattering reference point target in an urban area according to claim 1, characterized in that, Step 5: Compare the RCS data of the same reference point target in the SAR data to be calibrated with the RCS data of the calibrated reference SAR data, and perform cross-calibration to obtain the absolute radiometric calibration constant of the SAR data to be calibrated. The specific method is as follows: The calibration constant K for the SAR data to be calibrated is: K=K 0 dB +(10log 10 P 1 -10log 10 P 0 )(7) K 0 dB P represents the absolute calibration constant of the calibrated SAR data. 0 For the RCS data of the reference point target in the calibrated SAR data, P 1 RCS data for reference point targets in uncalibrated SAR data.

8. A SAR radiometric cross-calibration system based on a widely dispersed strong scattering reference point target in an urban area, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the SAR radiometric cross-calibration method based on a widely dispersed strong scattering reference point target in an urban area as described in any one of claims 1-7, thereby realizing SAR radiometric cross-calibration based on a widely dispersed strong scattering reference point target in an urban area.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the SAR radiometric cross-calibration method based on a widely strong scattering reference point target in an urban area as described in any one of claims 1-7, thereby realizing SAR radiometric cross-calibration based on a widely strong scattering reference point target in an urban area.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the SAR radiometric cross-calibration method based on a widely strong scattering reference point target in an urban area as described in any one of claims 1-7, thereby realizing SAR radiometric cross-calibration based on a widely strong scattering reference point target in an urban area.

Images