An unmanned aerial vehicle InSAR adaptive weighted filtering method, device and medium

Abstract

The present application relates to synthetic aperture radar signal processing technical field, especially to a kind of unmanned aerial vehicle InSAR adaptive weighted filtering method, equipment and medium, the method comprises: obtaining the image data of the same area unmanned aerial vehicle InSAR;Image data azimuth and range are uniformly blocked, and azimuth spectrum and range spectrum are obtained by frequency domain transformation;Spectrum overlap index and energy confidence index are constructed;Spectrum weighting function is constructed, and it is used to adaptively weight processing to the azimuth spectrum component and range spectrum component of the main image and auxiliary image;After spectrum is weighted and processed, inverse transformation is carried out, and filtered synthetic aperture radar image data are obtained.The present application method is directly aimed at the core problem that spectrum is inconsistent and energy is unstable due to motion error of airborne SAR by spectrum overlap and energy confidence double constraint mechanism, effectively distinguishes real signal from invalid spectrum component caused by motion error.

Description

Technical Field

[0001] This invention relates to the field of synthetic aperture radar signal processing technology, and in particular to an adaptive weighted filtering method, device and medium for InSAR of unmanned aerial vehicles. Background Technology

[0002] In Interferometric Synthetic Aperture Radar (InSAR), spectral matching is crucial for obtaining high-quality interferometric phase. Compared to spaceborne platforms, airborne SAR is more significantly and complexly affected by residual motion errors such as platform attitude jitter, trajectory deviation, and flight speed instability. These errors cause non-uniform translation, broadening, or distortion of the azimuth and / or range spectra of the primary and secondary images, especially the azimuth spectrum, which is difficult to completely eliminate even with precise motion compensation. This spectral inconsistency and instability makes traditional frequency domain filtering methods based on fixed bandwidth or single indicators unsuitable, often resulting in a dilemma of filtering too narrowly, losing effective signals, or filtering too widely, introducing noise. Particularly in airborne SAR, due to the large platform motion error, the effective interferometric spectral ranges of the primary and secondary images often do not completely overlap, resulting in inconsistent frequencies corresponding to spectral peaks and different spectral shapes. This leads to a significant increase in interferometric phase noise, a decrease in coherence, and a non-uniform distribution of azimuth coherence, sometimes fluctuating, and even causing local or overall interferometric failure. This phenomenon is difficult to completely eliminate even through precise registration methods such as block processing, high-precision navigation data, and adaptive motion compensation.

[0003] To address the aforementioned issues, existing technologies typically employ fixed-bandwidth azimuth or range bandpass filtering to truncate the spectra of primary and secondary images, preserving their common spectral components. However, these methods often rely on empirical parameter settings, using fixed or semi-fixed filtering bandwidths, making it difficult to accommodate spectral differences between different regions and frequency components. They are suitable for spaceborne SAR scenarios with stable baselines and minimal motion errors, based on the core assumption that the spectral mismatch patterns of primary and secondary images are relatively simple and spatially change slowly. However, for airborne SAR data with significant motion errors, spectral mismatch exhibits strong spatial variability and local bursts. Traditional fixed-bandwidth filtering or adaptive-bandwidth methods struggle to balance spectral differences between different regions and frequency components, failing to simultaneously and precisely address the dual challenges of drastic changes in spectral geometric consistency and sudden drops in local spectral energy reliability. This leads to the following problems: firstly, excessively narrow filtering bandwidths result in over-suppression of effective signals, reducing image resolution and losing ground feature edge information; secondly, excessively wide filtering bandwidths fail to effectively suppress frequency components with inconsistent spectra or low signal-to-noise ratios, still introducing strong interference noise. Existing adaptive filtering methods mainly select the spectrum based on a single metric, such as band-limiting based solely on the overlap range of the primary and secondary images' spectra, or weighting based solely on the magnitude of the spectral energy. These methods fail to simultaneously consider the geometric consistency of the primary and secondary images in the frequency domain and the reliability constraints of the spectral components against a noisy background. Under complex imaging conditions (such as the unstable azimuth coherence of airborne SAR), it remains difficult to obtain stable and reliable interferometric results.

[0004] Therefore, it is necessary to provide a new synthetic aperture radar bandpass filtering method that can adaptively weight the spectrum by comprehensively considering the consistency characteristics of the spectrum of the main and auxiliary images and the energy reliability of the spectral components in the azimuth and range frequency domains. This method can suppress low-reliability spectral components while maximizing the preservation of spectral information effective for interferometry and image detail preservation, thereby improving the phase stability and interferometry success rate in UAV-borne InSAR processing. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes an adaptive weighted filtering method, device, and medium for InSAR for unmanned aerial vehicles (UAVs).

[0006] In a first aspect, the present invention provides an adaptive weighted filtering method for InSAR of unmanned aerial vehicles, comprising: S1. Acquire single-view complex data of main and auxiliary InSAR images of the same area from UAVs. S2. The azimuth and range directions of the single-view complex data of the main image and the auxiliary image are uniformly divided into blocks. For each sub-block, frequency domain transformation is performed along the azimuth and range directions to obtain the azimuth spectrum and range spectrum corresponding to the main image and the auxiliary image. S3. Based on the frequency domain components of the main image and the auxiliary image in the azimuth and range directions, construct a spectral overlap index to characterize the spectral consistency between the main image and the auxiliary image, and an energy confidence index to characterize the reliability of the spectral components. S4. Construct a spectral weighting function based on the spectral overlap index and the energy confidence index. The spectral weighting function is used to perform adaptive weighting processing on the azimuth spectral components and range spectral components of the main image and the auxiliary image. S5. Perform an inverse transformation on the weighted spectrum to obtain filtered synthetic aperture radar image data.

[0007] Optionally, in S1, the UAV platform track deviation is better than ±5m, the antenna pointing angle is better than 0.1°, and the overlap of the azimuth and range spectra of the main and auxiliary images is greater than 50%.

[0008] Optionally, in S2, the azimuth and range spectra corresponding to the main image and the auxiliary image are obtained. The process includes: The registered main image and auxiliary image are evenly divided into blocks, and overlapping intervals are designed between each sub-block; Fast Fourier transform is performed on the same sub-block in the azimuth direction of the main image and the auxiliary image to obtain the corresponding azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image. The azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image are aligned and centered to obtain the azimuth spectrum centered at the zero Doppler frequency. Fast Fourier transform is performed on the same sub-block in the range direction of the main image and the auxiliary image to obtain the corresponding range direction spectrum of the main image and the range direction spectrum of the auxiliary image. The range spectrum of the main image and the range spectrum of the auxiliary image are aligned and centered to obtain a range spectrum centered at the zero Doppler frequency.

[0009] Optionally, in S3, the process of constructing the spectral overlap index includes: Based on the spectral amplitude values ​​of the main image and the auxiliary image at the same frequency, a spectral overlap index is calculated to characterize the degree of consistency of the spectral energy of the main image and the auxiliary image at that frequency. The spectral overlap index is used to reflect the geometric consistency of the main image and the auxiliary image in the azimuth / range frequency domain.

[0010] Optionally, in S3, the process of constructing the energy confidence index includes: Based on the sum of spectral energies of the main image and the auxiliary image at corresponding frequencies in the azimuth and range directions, an energy confidence index is constructed to characterize the reliability of the frequency components under noise background, and the energy confidence index is normalized.

[0011] Optionally, in S4, the azimuth spectral weighting function is a composite weighting function simultaneously constrained by both the spectral overlap index and the energy confidence index, and the azimuth spectral weighting function includes at least one of the following forms: A weighted function is constructed based on the product of the spectral overlap index and the energy confidence index; A weighted function is constructed based on the power function form of the aforementioned spectral overlap index and energy confidence index.

[0012] In a second aspect, a computer device includes a memory and a processor; The memory is used to store computer programs that can run on the processor; When the processor executes the computer program, it implements the steps of the UAV InSAR adaptive weighted filtering method described above.

[0013] In a third aspect, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the UAV InSAR adaptive weighted filtering method described above.

[0014] The application of the technical solution of the present invention has at least the following beneficial effects: (1) This invention provides an adaptive weighted filtering method for UAV InSAR. The joint constraint of the spectral overlap index and the energy confidence index enables adaptive weighting for different regions and frequency components, avoiding the problem of insufficient adaptability of fixed bandwidth filtering to complex spectral conditions, as well as the universality of different equipment and data frequency components that are not similar. Compared with the prior art, the dual constraint mechanism of spectral overlap and energy confidence proposed in this invention directly addresses the core problems of spectral inconsistency and energy instability caused by motion errors in airborne SAR, and can effectively distinguish between real signals and invalid spectral components caused by motion errors.

[0015] (2) The method of the present invention effectively reduces interference phase noise, improves interference coherence and phase stability, and reduces phase residual points by suppressing spectral components with low spectral consistency or low signal-to-noise ratio. Compared with simple spectral truncation methods, the present invention uses a continuous weighting method to modulate the spectrum, which preserves the effective spectral information to the maximum extent while ensuring interference quality, which is beneficial to maintaining image resolution and edge index. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of the UAV InSAR adaptive weighted filtering method in a preferred embodiment of the present invention. Figure 2 This is a schematic diagram of the spectrum of the main image and the auxiliary image in the high coherence region in a preferred embodiment of the present invention, wherein (a) represents the azimuth spectrum and (b) represents the range spectrum; Figure 3 This is a schematic diagram of the spectrum of the main image and the auxiliary image in the low coherence region in a preferred embodiment of the present invention, wherein (a) represents the azimuth spectrum and (b) represents the range spectrum; the blue solid line represents the spectrum of the main image and the red dashed line represents the spectrum of the auxiliary image; Figure 4 This is a schematic diagram of spectral overlap and energy confidence in a preferred embodiment of the present invention (the blue solid line represents the spectral overlap index). η The red dashed line represents the energy confidence index. ω ), where (a) represents the azimuth spectrum of the high coherence region, (b) represents the range spectrum of the high coherence region, (c) represents the azimuth spectrum of the low coherence region, and (d) represents the range spectrum of the low coherence region; the blue solid line represents the spectral overlap, and the red dashed line represents the energy confidence level; Figure 5 This is a schematic diagram of the spectral weighting function in a preferred embodiment of the present invention, wherein (a) represents the azimuth spectral weighting function in the high coherence region, (b) represents the range spectral weighting function in the high coherence region, (c) represents the azimuth spectral weighting function in the low coherence region, and (d) represents the range spectral weighting function in the low coherence region; Figure 6 This is a schematic diagram comparing the spectrum before and after weighting processing in a preferred embodiment of the present invention, wherein (a) represents the azimuth spectrum of the primary and secondary images in the high coherence region, (b) represents the range spectrum of the primary and secondary images in the high coherence region, (c) represents the azimuth spectrum of the primary and secondary images in the low coherence region, and (d) represents the range spectrum of the primary and secondary images in the low coherence region. Figure 7This is a schematic diagram comparing the coherence distribution of the spectrum before and after weighted filtering in a preferred embodiment of the present invention, wherein (a) represents the original coherence distribution, (b) represents the coherence distribution after azimuth filtering, (c) represents the coherence distribution after range filtering, and (d) represents the coherence distribution after simultaneous azimuth and range filtering. Figure 8 This is a schematic diagram of the statistical histogram of interference coherence probability density under different processing methods in a preferred embodiment of the present invention; Figure 9 This is a schematic diagram comparing the image intensity (local) of the main image before and after spectral weighted filtering in a preferred embodiment of the present invention. (a) represents the image intensity diagram of the main image without filtering, (b) represents the image intensity diagram of the main image with only azimuth filtering, (c) represents the image intensity diagram of the main image with only range filtering, and (d) represents the image intensity diagram of the main image with both azimuth and range filtering. Figure 10 This is a comparative schematic diagram of the image interference phase in a preferred embodiment of the present invention. The interference phase is obtained by multiplying the main and auxiliary images by their conjugates. In this diagram, (a) represents the unfiltered phase, (b) represents the interference phase after filtering only in the azimuth direction, (c) represents the interference phase after filtering only in the range direction, and (d) represents the interference phase after filtering both in the azimuth and range directions. Figure 11 This is a coherence statistical histogram under different directional spectral window function conditions in a preferred embodiment of the present invention; Figure 12 This is a coherence statistical histogram under different distance-direction spectral window function conditions in a preferred embodiment of the present invention. Detailed Implementation

[0018] 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: like Figure 1 As shown, this embodiment provides an adaptive weighted filtering method for UAV InSAR, including (steps S1 to S5): S1. Acquire single-view complex data of the main and auxiliary InSAR images of the same area on a UAV; the UAV platform track deviation is better than ±5m, the antenna pointing angle is better than 0.1°, and the azimuth spectrum and range spectrum overlap of the main and auxiliary images is greater than 50%.

[0020] S2. Divide the single-view complex data of the main image and auxiliary image into uniform blocks along the azimuth and range directions. Perform frequency domain transformation on each sub-block along the azimuth and range directions to obtain the following... Figure 2 and Figure 3 The azimuth and range spectra of the main and auxiliary images shown are as follows: Figure 2 The spectrum of the high coherence region, Figure 3 For the spectrum of the low coherence region, in Figure 2 and Figure 3 In the image, the solid blue line represents the spectrum of the primary image, and the dashed red line represents the spectrum of the secondary image.

[0021] To improve the spatial adaptability of frequency domain analysis, this embodiment employs a uniform block processing strategy for the registered main image and auxiliary image. The process includes: S2.1 Divide the registered main image and auxiliary image into uniform blocks, and design overlapping intervals between each sub-block to reduce the boundary effect that may be introduced at the spatial stitching point during block processing.

[0022] S2.2 Perform Fast Fourier Transform on the same sub-block of the main image and the auxiliary image in the azimuth direction to obtain the corresponding azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image; perform spectrum centering processing on the obtained azimuth spectrum so that the zero Doppler frequency is located in the center of the spectrum, which facilitates the consistency analysis of different frequency components in the subsequent process.

[0023] S2.3 Perform spectrum range alignment and centering processing on the azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image to obtain the azimuth spectrum centered on the zero Doppler frequency; perform spectrum centering processing on the obtained azimuth spectrum so that the zero Doppler frequency is located at the center of the spectrum, which facilitates the consistency analysis of different frequency components in the subsequent process.

[0024] S2.4 Perform Fast Fourier Transform on the same sub-block of the main image and the auxiliary image in the range direction to obtain the corresponding range spectrum of the main image and the range spectrum of the auxiliary image; perform spectrum centering processing on the obtained range spectrum so that the zero Doppler frequency is located in the center of the spectrum, which facilitates the consistency analysis of different frequency components in the subsequent process.

[0025] S2.5 Perform spectrum range alignment and centering processing on the range spectrum of the main image and the range spectrum of the auxiliary image to obtain a range spectrum centered at the zero Doppler frequency.

[0026] S3. Based on the frequency domain components of the main image and the auxiliary image in the azimuth and range directions, construct a spectral overlap index to characterize the spectral consistency between the main image and the auxiliary image, and an energy confidence index to characterize the reliability of the spectral components.

[0027] This embodiment constructs a spectral overlap index and an energy confidence index based on the energy distribution and consistency characteristics of the main image and the auxiliary image in the frequency domain, which are used to construct the subsequent adaptive spectral weighting function. The spectral overlap and energy confidence in this embodiment are as follows: Figure 4 As shown, the blue solid line represents the spectral overlap index. η The red dashed line represents the energy confidence index. ω , where (a) represents the azimuth spectrum of the high coherence region, (b) represents the range spectrum of the high coherence region, (c) represents the azimuth spectrum of the low coherence region, and (d) represents the range spectrum of the low coherence region.

[0028] In this embodiment, the process of constructing the spectral overlap index includes: Based on the spectral amplitude values ​​of the main image and the auxiliary image at the same frequency, a spectral overlap index is calculated to characterize the degree of consistency of the spectral energy of the main image and the auxiliary image at that frequency. The spectral overlap index is used to reflect the geometric consistency of the main image and the auxiliary image in the azimuth / range frequency domain.

[0029] Specifically, the spectral overlap index Based on the spectral amplitude information of the primary and secondary images at the same frequency, a spectral overlap index is constructed to characterize the degree of spectral consistency between the two at that frequency. The spectral overlap index reflects the geometric consistency characteristics of the primary and secondary images in the azimuth and range frequency domains: Let the primary and secondary images be within a certain sub-block at a certain frequency... The complex spectrum at each position is respectively Its corresponding amplitude spectrum is The spectral overlap index is defined as follows: ; in, When the primary and secondary images are in frequency When the energy distribution is the same or similar, Approaching 1; when there is a significant mismatch or shift in the spectrum of the primary and secondary images. Significantly reduced. The spectral overlap index is used to quantitatively describe the geometric consistency and effective overlap of primary and secondary images in the frequency domain.

[0030] Furthermore, the process of constructing the energy confidence index includes: Based on the sum of spectral energies of the main image and the auxiliary image at corresponding frequencies in the azimuth and range directions, an energy confidence index is constructed to characterize the reliability of the frequency components under noise background, and the energy confidence index is normalized.

[0031] To avoid introducing low signal-to-noise ratio frequency components based solely on spectral overlap, this embodiment further introduces an energy confidence index. The energy confidence index is calculated based on the sum of the spectral energy of the main image and the auxiliary image at corresponding frequencies, and is normalized to make different frequency components comparable, which is used to measure the reliability of spectral components in a noisy background.

[0032] Assuming at frequency The spectral energies of the primary and secondary images are respectively The energy confidence index is defined as follows: ； in, This represents the maximum sum of spectral energies at all frequencies within the current sub-block; normalization is used to make the energy confidence levels of different frequency components comparable; when the energy of a certain frequency component is low (e.g., ... Figure 3 This embodiment constructs a spectral overlap index and an energy confidence index based on the energy distribution and consistency characteristics of the main image and the auxiliary image in the frequency domain, which are used to construct the subsequent adaptive spectral weighting function. When the spectral energy at a certain frequency is low, even if its spectral overlap is high, its energy confidence is still low and will be suppressed in the subsequent weighting process.

[0033] S4. Construct a spectral weighting function based on the spectral overlap index and the energy confidence index. The spectral weighting function is used to adaptively weight the azimuth spectral components and range spectral components of the main image and the auxiliary image.

[0034] In this embodiment, the spectral weighting function is a composite weighting function that is simultaneously constrained by both the spectral overlap index and the energy confidence index. The spectral weighting function includes at least one of the following forms: A weighted function is constructed based on the product of the spectral overlap index and the energy confidence index; A weighted function is constructed based on the power function form of the aforementioned spectral overlap index and energy confidence index.

[0035] This embodiment constructs a spectral weighting function in the frequency domain, constrained by both a spectral overlap index and an energy confidence index, for adaptive weighting of the azimuth or range spectrum. The spectral weighting function in this embodiment is as follows: Figure 5 As shown, (a) represents the azimuth spectral weighting function for the high coherence region, (b) represents the range spectral weighting function for the high coherence region, (c) represents the azimuth spectral weighting function for the low coherence region, and (d) represents the range spectral weighting function for the low coherence region.

[0036] The spectrum weighting function is defined as: ; in, , representing the spectrum overlap control index; , representing the energy confidence control index; Expanding further, we can obtain: ; In regions of high coherence, the spectral overlap between the primary and secondary images is high, and the overall energy confidence index is relatively large. The constructed spectral weighting function approaches 1 within the effective frequency band, only slightly suppressing frequency components at the band edges or with unstable energy, thereby preserving the spectral information effective for interferometry to the maximum extent and avoiding resolution loss. In regions of low coherence, due to factors such as large baselines, motion errors, or low signal-to-noise ratios, the spectral consistency between the primary and secondary images decreases significantly, and the energy confidence of some frequency components is low.

[0037] It should be noted that the spectral weighting function has the following characteristics: within a frequency range where the primary and secondary images have highly consistent spectra and sufficient energy, In frequency ranges with spectral mismatch or insufficient energy, The continuous decrease in frequency weighting achieves soft suppression; the spectral weighting function changes continuously in the frequency domain without introducing hard truncation boundaries, avoiding the sidelobe effect caused by traditional rectangular bandpass filtering. In this case, the adaptive spectral weighting function will exert stronger suppression on frequency components with low spectral overlap or low energy confidence, thereby effectively weakening the negative impact of unstable spectral components on the interference phase and improving overall phase stability.

[0038] S5. Perform inverse Fourier transform on the weighted azimuth spectrum or range spectrum to obtain complex data of the main image and auxiliary image after adaptive weighted filtering, and use them in subsequent interferometric processing.

[0039] The following is a comparison between the method in this embodiment and the traditional window function filtering method:

[0040] Figure 6The diagram shows a comparison of the spectra before and after weighting processing. (a) represents the azimuth spectrum of the primary and secondary images in the high-coherence region; (b) represents the range spectrum of the primary and secondary images in the high-coherence region; (c) represents the azimuth spectrum of the primary and secondary images in the low-coherence region; and (d) represents the range spectrum of the primary and secondary images in the low-coherence region. In the spectral weighting function proposed in this embodiment, the weight distribution of the image at different frequency components adapts adaptively. Traditional window functions have fixed weight distributions, and the filtering bandwidth and shape mainly rely on manual experience parameter settings, which cannot adaptively adjust with changes in the consistency of the primary and secondary image spectra. However, the spectral weighting function constructed by the method in this embodiment can continuously change in the frequency domain according to the spectral overlap and energy confidence, and is entirely data-driven. For different data and different scenarios, it adaptively adjusts the weight distribution of different frequency components.

[0041] like Figure 7 As shown, in order to quantitatively evaluate the impact of different filtering methods on the interference quality, this embodiment calculates the interference coherence of the filtered interference results and statistically analyzes its distribution characteristics.

[0042] Interference coherence is calculated using the following standard definition: ; in, The first main image 1 pixel, For the first image 1 pixel, Perform complex conjugation on the image. The number of pixels in the window (e.g., 5) 5 windows, N =25), This indicates taking the modulus.

[0043] like Figure 7 As shown, by simultaneously applying the azimuth and range spectral filtering methods proposed in this embodiment to the primary and secondary images in series, the interferometric coherence of unfiltered, azimuth-filtered alone, range-filtered, and azimuth-filtered + range-filtered images is compared and plotted. It can be seen that the original image exhibits significant loss of coherence in the azimuth direction due to motion errors, registration errors, etc. The filtering proposed in this invention can greatly improve the coherence.

[0044] Furthermore, the coherence calculation window is 7. 7. By statistically analyzing the coherence corresponding to different azimuth spectral filtering methods, the coherence histogram distribution is obtained (e.g., Figure 11As shown). The average coherence before filtering is 0.441. Three window functions were used: rect window, Hanning window, and Kaiser window, with a conventional fixed bandwidth of 0.2714 (95% of the normalized energy after Fourier transform is concentrated in the range of ±0.1357 from the center of the spectrum). β =2.5) The average coherence of the windowed images is 0.459, 0.589, and 0.519, respectively. However, when using the adaptive weighted filtering method for azimuth spectrum in synthetic aperture radar interferometry processing proposed in this embodiment of the invention ( , The average coherence of the resulting images was improved to 0.605.

[0045] Furthermore, by simultaneously applying the azimuth and range spectral filtering methods proposed in this embodiment to the primary and secondary images in a cascaded manner, the results are compared: no filtering, azimuth filtering alone, range filtering, and azimuth filtering + range filtering. The primary and secondary images are then multiplied by their conjugates, and the interferometric coherence is calculated. Statistical analysis is then performed to obtain and statistically analyze the distribution characteristics. This yields the corresponding coherence probability distribution histogram, as shown below. Figure 8 As shown, azimuth filtering can greatly improve coherence. It also shows that the method proposed in this embodiment can be implemented sequentially and improve interference coherence without interfering with each other. The coherence improvement effect is independent.

[0046] The azimuth and range spectral filtering methods proposed in this embodiment are simultaneously applied to the primary and secondary images in a cascaded manner. The intensity maps of the primary image before and after filtering (partial) are compared: no filtering, azimuth filtering alone, range filtering, and azimuth filtering + range filtering. Figure 9 As shown, (a) represents the image intensity diagram of the main image without filtering, (b) represents the image intensity diagram of the main image with only azimuth filtering, (c) represents the image intensity diagram of the main image with only range filtering, and (d) represents the image intensity diagram of the main image with both azimuth and range filtering. It can be seen that the brightness of the main image is uniform before and after filtering, and there is no obvious defocus or fragmentation, which proves that the filtering does not change the overall radiation characteristics of the image, nor does it cause edge softening.

[0047] Similarly, the azimuth and range spectral filtering methods proposed in this embodiment are simultaneously applied to the primary and secondary images in a cascaded manner. The results are compared as follows: filtering alone, azimuth filtering alone, range filtering alone, and azimuth filtering plus range filtering. The primary and secondary images are then multiplied by conjugate, and interferometric phase diagrams before and after filtering are plotted as follows: Figure 10As shown, (a) represents the interferometric phase diagram of the main image without filtering, (b) represents the interferometric phase diagram of the main image with only azimuth filtering, (c) represents the interferometric phase diagram of the main image with only range filtering, and (d) represents the interferometric phase diagram of the main image with both azimuth and range filtering. It can be seen that before filtering, the interferometric phase noise is obvious, with obvious azimuth decoherence and low signal-to-noise ratio. After filtering, the interferometric phase is significantly clear, the azimuth decoherence is reduced, the noise is greatly weakened, and the residual points are greatly reduced, which can greatly improve the phase unwrapping success rate.

[0048] Furthermore, by performing statistical analysis on the interferometric coherence after processing with different range-direction spectral filtering methods, the corresponding coherence probability distribution histograms are obtained, such as... Figure 12 As shown. The average coherence before filtering is 0.441. Three window functions were used: rect window, Hanning window, and Kaiser window. The standard fixed bandwidth of 0.5327 (95% of the normalized energy after Fourier transform is concentrated within ±0.2663 of the spectrum center) were used. β The average coherence of the windowed images (=2.5) is 0.447, 0.502, and 0.472. The range-oriented spectral adaptive weighted filtering method proposed in this embodiment of the invention is used. , After that, the average coherence of the image was further improved to 0.473.

[0049] Statistical analysis of residual error points: In order to further analyze the ability of different filtering methods to suppress residual interference errors, this embodiment statistically analyzes the residual errors of the interference phase before and after filtering.

[0050] The residual phase is defined as: ; in, Indicates the filtered interference phase. This represents the reference phase estimated using a smoothing model or local fitting method.

[0051] Furthermore, the number of pixels whose residual phase amplitude exceeds a preset threshold is counted as a residual point index: ; in, The residual phase threshold, This represents the cell index.

[0052] By statistically analyzing the number of interferometric phase residual points corresponding to different azimuth spectral filtering methods, the results show that the number of residual points in the image without filtering is 2,769,656. A conventional azimuth spectral filtering method with a fixed bandwidth of 0.2714 (95% of the normalized energy after Fourier transform is concentrated within the range of ±0.1357 from the spectral center) was adopted, and rectangular windows, Hanning windows, and Kaiser windows were used respectively. β After processing with a window of 2.5, the number of residual points in the corresponding images are 2,518,539, 1,401,032, and 1,952,901, respectively. Using the method of this embodiment (… , The number of residual points in the image was further reduced to 1,321,617, which significantly reduced the distribution density of interferometric phase discontinuities.

[0053] Furthermore, statistical analysis of the number of interferometric phase residual points after processing with different range-direction spectral filtering methods revealed that the number of residual points in the image without filtering was 2,769,656. Using a conventional range-direction spectral filtering method with a fixed bandwidth of 0.5327 (95% of the normalized energy after Fourier transform is concentrated within ±0.2663 of the spectral center), and processing with rectangular windows, Hanning windows, and Kaiser windows respectively, the number of residual points in the corresponding images were 2,605,049, 1,502,400, and 2,051,922, respectively. The number of residual points using the method in this embodiment was reduced to 2,101,482, indicating that the method in this embodiment also has a good phase consistency improvement effect in the range direction. This embodiment significantly reduces the number of residual error points while ensuring the preservation of effective spectral information, demonstrating its stronger suppression capability for unstable spectral components.

[0054] To evaluate the ability of different filtering methods to preserve the edge structure of the interference phase space while suppressing noise, this embodiment further introduces the Edge Preservation Index. Quantitative analysis was performed on the Edge Preservation Index (EPI). The EPI is defined as the ratio of the correlation between the phase gradients before and after filtering, and the expression is as follows: ; in, Indicates the interference phase before filtering. Indicates the interference phase after filtering. The EPI (Electronic Phase Filtering) represents the spatial gradient operator. It is defined as the Pearson correlation coefficient between the phase gradient fields before and after filtering. A value closer to 1 indicates a stronger ability to preserve the original spatial edge structure after filtering. Specifically, an EPI > 0.90–0.95 indicates excellent edge preservation with almost no significant blurring; 0.80–0.90 is good, with slight smoothing; and below 0.80 indicates some degree of edge loss or blurring. This index is particularly suitable for evaluating the ability of SAR interferometric phase filtering to balance improving coherence with preserving fine ground feature edges (such as buildings, roads, and fault lines).

[0055] like Figure 11 and Figure 12 As shown, a quantitative and qualitative comparative analysis was conducted between the method of this embodiment and the traditional window function filtering method.

[0056] Among them, by statistically analyzing the edge preservation index after different azimuth spectral filtering methods, the edge preservation index before filtering was 1. Using three conventional window functions (rect window, Hanning window, and Kaiser window) with a fixed bandwidth of 0.2714 (95% of the normalized energy after Fourier transform is concentrated within ±0.1357 of the spectrum center), the image edge preservation indices were 0.962, 0.857, and 0.934, respectively. The azimuth spectral index using the method in this embodiment (…) , The edge preservation index (EPI) of the image after weighted filtering is 0.870.

[0057] The edge preservation index (EPI) after different range-direction spectral filtering methods was statistically analyzed. Before filtering, the EPI was 1. Using three conventional window functions (rect window, Hanning window, and Kaiser window) with a fixed bandwidth of 0.5327 (95% of the normalized energy after Fourier transform is concentrated within ±0.2663 of the spectral center), the EPIs were 0.973, 0.8337, and 0.9402, respectively. Using the range-direction spectral filtering method of this embodiment... , The edge preservation index (EPI) of the image after weighted filtering is 0.9710.

[0058] Table 1. Comparison of average coherence, number of residual error points, and edge preservation index under different filtering methods.

[0059] As shown in Table 1, the average coherence without filtering is 0.441, the number of residual error points is as high as 2,769,656, and the EPI is 1 (reference value). Using conventional fixed bandwidths of 0.2714 (95% of the normalized energy after Fourier transform is concentrated within ±0.1357 of the spectrum center), rectangular windows, Hanning windows, and Kaiser windows are employed.β After adjusting for 2.5, although the azimuth average coherence increased to 0.459, 0.589, and 0.519 respectively, and the number of residual error points decreased, the edge preservation index (EPI) generally decreased (e.g., only 0.857 for the Hanning window). This indicates that traditional fixed window function filtering methods, due to their fixed spectral weight distribution, are prone to introducing excessive smoothing effects at frequency band edges, leading to phase edge blurring and loss of structural information. In contrast, the method in this embodiment can retain more effective spectral components in high-reliability frequency bands and implement adaptive suppression in low-reliability frequency bands based on spectral energy overlap and energy confidence characteristics. This significantly improves interferometric coherence (azimuth up to 0.605, better than all traditional window functions) while effectively enhancing the image's edge preservation capability (EPI of 0.870) and phase stability, further reducing the number of residual error points to 1,321,617, demonstrating superior overall performance.

[0060] This embodiment provides an adaptive weighted filtering method for UAV InSAR based on the spectral characteristics of residual motion errors. Through comparisons from multiple perspectives, including coherence statistics, the number of residual error points, and the edge preservation index, it can be seen that the method of this embodiment has at least the following advantages: it simultaneously considers the spectral consistency and spectral energy reliability of the primary and secondary images, avoiding the limitations of single-criterion filtering; it maximizes the preservation of effective spectral information in high-coherence regions and effectively suppresses unstable spectral components in low-coherence regions; it significantly reduces the number of phase residual error points while improving interferometric coherence; it has good phase spatial edge preservation capabilities, avoiding excessive smoothing; it is applicable to frequency domain processing in the azimuth and range directions, has strong versatility, low parameter dependence, and is easy to implement in engineering.

[0061] In addition, this embodiment also provides a computer device, including a memory and a processor; The memory is used to store computer programs that can run on the processor; When the processor executes the computer program, it implements the steps of the UAV InSAR adaptive weighted filtering method described above.

[0062] It should be noted that computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0063] The flowcharts in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in the flowchart may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowchart, and combinations of blocks in the flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0064] In addition, this embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the UAV InSAR adaptive weighted filtering method described above.

[0065] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described UAV InSAR adaptive weighted filtering method. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the UAV InSAR adaptive weighted filtering method provided in the above embodiments, and will not be repeated here.

[0066] The above description is only a preferred embodiment of the present invention and does not limit the scope of the present invention. All equivalent structural transformations made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the protection scope of the present invention.

Claims

1. An adaptive weighted filtering method for InSAR of unmanned aerial vehicles (UAVs), characterized in that, include: S1. Acquire single-view complex data of main and auxiliary InSAR images of the same area from UAVs. S2. The azimuth and range directions of the single-view complex data of the main image and the auxiliary image are uniformly divided into blocks. For each sub-block, frequency domain transformation is performed along the azimuth and range directions to obtain the azimuth spectrum and range spectrum corresponding to the main image and the auxiliary image. S3. Based on the frequency domain components of the main image and the auxiliary image in the azimuth and range directions, construct a spectral overlap index to characterize the spectral consistency between the main image and the auxiliary image, and an energy confidence index to characterize the reliability of the spectral components. S4. Construct a spectral weighting function based on the spectral overlap index and the energy confidence index. The spectral weighting function is used to perform adaptive weighting processing on the azimuth spectral components and range spectral components of the main image and the auxiliary image. S5. Perform an inverse transformation on the weighted spectrum to obtain filtered synthetic aperture radar image data.

2. The UAV InSAR adaptive weighted filtering method according to claim 1, characterized in that, In S1, the UAV platform track deviation is better than ±5m, the antenna pointing angle is better than 0.1°, and the overlap of the azimuth and range spectra of the main and auxiliary images is greater than 50%.

3. The UAV InSAR adaptive weighted filtering method according to claim 2, characterized in that, In S2, the azimuth and range spectra corresponding to the main and auxiliary images are obtained. The process includes: The registered main image and auxiliary image are evenly divided into blocks, and overlapping intervals are designed between each sub-block; Fast Fourier transform is performed on the same sub-block in the azimuth direction of the main image and the auxiliary image to obtain the corresponding azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image. The azimuth spectrum of the main image and the azimuth spectrum of the auxiliary image are aligned and centered to obtain the azimuth spectrum centered at the zero Doppler frequency. Fast Fourier transform is performed on the same sub-block in the range direction of the main image and the auxiliary image to obtain the corresponding range direction spectrum of the main image and the range direction spectrum of the auxiliary image. The range spectrum of the main image and the range spectrum of the auxiliary image are aligned and centered to obtain a range spectrum centered at the zero Doppler frequency.

4. The UAV InSAR adaptive weighted filtering method according to claim 3, characterized in that, In S3, the process of constructing the spectral overlap index includes: Based on the spectral amplitude values ​​of the main image and the auxiliary image at the same frequency, a spectral overlap index is calculated to characterize the degree of consistency of the spectral energy of the main image and the auxiliary image at that frequency. The spectral overlap index is used to reflect the geometric consistency of the main image and the auxiliary image in the azimuth / range frequency domain.

5. The UAV InSAR adaptive weighted filtering method according to claim 4, characterized in that, In S3, the process of constructing the energy confidence index includes: Based on the sum of spectral energies of the main image and the auxiliary image at corresponding frequencies in the azimuth and range directions, an energy confidence index is constructed to characterize the reliability of the frequency components under noise background, and the energy confidence index is normalized.

6. The UAV InSAR adaptive weighted filtering method according to claim 5, characterized in that, In S4, the azimuth spectral weighting function is a composite weighting function constrained by both the spectral overlap index and the energy confidence index. The azimuth spectral weighting function includes at least one of the following forms: A weighted function is constructed based on the product of the spectral overlap index and the energy confidence index; A weighted function is constructed based on the power function form of the aforementioned spectral overlap index and energy confidence index.

7. A computer device, characterized in that, Including memory and processor; The memory is used to store computer programs that can run on the processor; When the processor executes the computer program, it implements the steps of the UAV InSAR adaptive weighted filtering method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the UAV InSAR adaptive weighted filtering method as described in any one of claims 1 to 6.

Images