Interferometric phase denoising method and device

Abstract

The embodiment of the invention provides an interferometric phase denoising method and device, and relates to the data processing technology, and the method comprises the steps: carrying out the multi-scale frequency decomposition of an original interferometric phase diagram based on a preset frequency domain filtering window, so as to obtain a hierarchical phase signal containing different frequency components; according to the difference between the principal component signals of the adjacent hierarchies, residual components of the adjacent hierarchies are calculated, and residual domain signals corresponding to all the hierarchies are obtained; de-noising processing is carried out on the residual domain signals by using a pre-trained de-noising model to obtain de-noised residual domain signals corresponding to each level; and performing phase reconstruction on the denoised residual domain signal corresponding to each level to obtain a target interferometric phase diagram. According to the invention, the authenticity and integrity of the phase structure can be effectively maintained while the noise is suppressed, the generation of artifacts is effectively reduced, the convergence stability is improved, and the reliability of phase unwrapping and subsequent deformation inversion is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to an interference phase denoising method and device. BACKGROUND

[0002] InSAR (Interferometric Synthetic Aperture Radar) is an important technology for obtaining phase information of the ground surface by using a radar imaging system, and realizing high-precision digital elevation model generation and ground surface deformation monitoring through interference processing. The basic principle is: two radar complex images of the same area obtained at different times or under different orbit conditions are subjected to interference operation to obtain an interference fringe pattern (i.e. an interferogram), and then through phase unwrapping and subsequent processing, the height information or deformation information of the ground surface can be inverted. Because it can obtain data all-weather, all-day and large-scale, InSAR has become an important means in the field of earth science and disaster monitoring.

[0003] However, the interference phase diagram of InSAR is inevitably affected by various noises in practical application. The main noise sources include: (1) temporal incoherence: due to the change of surface scattering characteristics with time caused by vegetation growth, snow change, water fluctuation, etc., the signal coherence is reduced; (2) spatial baseline effect: due to the difference in geometric conditions of two radar imaging, the spatial registration deviation of echo signal is caused, thereby introducing noise; (3) system thermal noise: caused by the instability of the radar sensor itself; (4) atmospheric delay: the phase delay caused by the tropospheric water vapor and ionospheric disturbance increases the phase error; (5) volume scattering effect: for example, the volume scattering target such as forest will cause the signal to be randomized in the propagation process. These noises not only reduce the signal-to-noise ratio of the interferogram, but also directly affect the stability of phase unwrapping and the accuracy of the final deformation result.

[0004] To solve the above problems, a variety of interference phase denoising methods are proposed in the related art, one of which is to use a spatial domain filter for processing to achieve the purpose of eliminating noise. However, although the spatial domain filter has high computational efficiency, it relies on the local uniformity assumption, which leads to phase distortion and detail loss in low-coherence or large phase gradient areas. The non-local filtering method improves the denoising performance by utilizing global similar structures, but it is still prone to introduce artifacts under strong noise conditions, and the computational complexity is large, which is difficult to meet the needs of large-scale and real-time applications. The frequency domain method (such as Goldstein filter) performs relatively stably in incoherent areas, but due to the inherent high-frequency attenuation effect, it will cause detail blur, which is not conducive to the preservation of small deformation signals. Another method is the InSAR phase denoising method based on deep learning. For example, the PhiNet network uses a residual learning framework to map the input noisy phase to a clean phase; the MONet network uses a multi-objective optimization strategy to try to balance between noise suppression and structure preservation; and the MSFF-DCNN uses multi-scale feature dynamic fusion to enhance the denoising performance. The common point of these methods is that the interference phase map is taken as the input, and the feature extraction and nonlinear mapping are performed through a deep convolutional network or an attention mechanism, and the estimated denoised phase result is output. Although their performance on the simulation data set is better than that of the traditional filter, their training process generally relies on idealized synthetic noise data, which is difficult to reflect the complex and diverse noise characteristics in the real interference map, and thus artifacts or false recovery are easily produced under low signal-to-noise ratio conditions, affecting the reliability of phase unwrapping and deformation inversion. SUMMARY

[0005] Embodiments of the present application provide an interference phase denoising method and device to alleviate or solve one or more technical problems in the prior art.

[0006] In a first aspect, embodiments of the present application provide an interference phase denoising method, comprising: performing multi-scale frequency decomposition on an original interference phase map based on a preset frequency domain filter window to obtain hierarchical phase signals containing different frequency components; the size of the frequency domain filter window increases layer by layer in order from low to high; calculating residual components of adjacent hierarchical levels according to the difference between the principal component signals of the adjacent hierarchical levels to obtain residual domain signals corresponding to each hierarchical level respectively; performing denoising processing on the residual domain signals by using a pre-trained denoising model to obtain denoised residual domain signals corresponding to each hierarchical level respectively; performing phase reconstruction on the denoised residual domain signals corresponding to each hierarchical level respectively to obtain a target interference phase map.

[0007] In a second aspect, embodiments of the present application provide an interference phase denoising device, comprising: The decomposition module is configured to perform multi-scale frequency decomposition on the original interference phase map based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components, wherein the size of the frequency domain filtering window increases layer by layer in order from low to high levels. The calculation module is configured to calculate residual components of adjacent levels according to differences between principal component signals of the adjacent levels to obtain residual domain signals corresponding to the respective levels. The processing module is configured to perform denoising processing on the residual domain signals by using a pre-trained denoising model to obtain denoised residual domain signals corresponding to the respective levels. The reconstruction module is configured to perform phase reconstruction on the denoised residual domain signals corresponding to the respective levels to obtain a target interference phase map.

[0008] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory. The processor implements any of the methods of the embodiments of the present application when executing the computer program.

[0009] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement any of the methods of the embodiments of the present application.

[0010] In a fifth aspect, a computer program product is provided, which includes a computer program. The computer program is executed by a processor to implement any of the methods of the embodiments of the present application.

[0011] According to the technical scheme of the embodiments of the present application, the original interference phase map is decomposed based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components, the size of the frequency domain filtering window increases layer by layer in order from low to high levels, residual components of adjacent levels are calculated according to differences between principal component signals of the adjacent levels to obtain residual domain signals corresponding to the respective levels, denoising processing is performed on the residual domain signals by using a pre-trained denoising model to obtain denoised residual domain signals corresponding to the respective levels, and phase reconstruction is performed on the denoised residual domain signals corresponding to the respective levels to obtain a target interference phase map. Therefore, low-frequency principal components are extracted by hierarchical frequency domain decomposition, high-frequency details are reconstructed layer by layer in the residual domain, a step-by-step optimization process from coarse to fine is constructed, and thus the authenticity and integrity of the phase structure are effectively maintained while suppressing noise. Moreover, optimization is performed in the residual domain rather than the overall phase domain, which can effectively reduce the generation of artifacts and improve convergence stability, and thus comprehensive advantages that related technologies cannot simultaneously achieve are realized in terms of noise suppression, detail preservation, computational efficiency, and adaptation to complex noise environments, thereby effectively ensuring the reliability of phase unwrapping and subsequent deformation inversion.

[0012] The above description is only a summary of the technical solutions of the present application. In order to enable the technical means of the present application to be more clearly understood, the description can be implemented according to the content of the specification, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0013] In the drawings, like reference numerals refer to same or similar functionalities throughout the several views. The drawings are not necessarily to scale. It is to be understood that the drawings only depict certain embodiments in accordance with the present application and should not be considered limiting of the scope of the present application.

[0014] Figure 1 A structural diagram of a denoising model provided by an embodiment of the present application is shown; Figure 2 A flowchart of an interference phase denoising method provided by an embodiment of the present application is shown; Figure 3 A training principle diagram of a denoising model provided by an embodiment of the present application is shown; Figure 4 An extendable reconstruction principle diagram of a denoising model provided by an embodiment of the present application is shown; Figure 5 A comparison diagram of extendable reconstruction performance of a denoising model provided by an embodiment of the present application under different scale levels is shown; Figure 6 An iterative reconstruction process principle diagram of a denoising model provided by an embodiment of the present application is shown; Figure 7 A structural diagram of an interference phase denoising device provided by an embodiment of the present application is shown; Figure 8 A block diagram of an electronic device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] In the following, only certain exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the concept or scope of the present application. Therefore, the drawings and the description are considered to be exemplary in nature, rather than limiting.

[0016] In order to facilitate understanding of the technical solutions of the embodiments of the present application, the related technologies of the embodiments of the present application are described below. The following related technologies can be combined with the technical solutions of the embodiments of the present application in any way as optional solutions, which all belong to the protection scope of the embodiments of the present application.

[0017] The present application aims to propose a MSRD-Net (Multi-Scale Residual Denoising Network) to cope with the denoising challenge of InSAR interferometric phase under different noise intensity conditions. The core idea is to adopt a frequency-space hybrid strategy: first, use hierarchical frequency domain decomposition to construct multi-scale phase signals, and then gradually reconstruct the phase in the residual domain, so as to realize step-by-step optimization from coarse to fine, which can effectively suppress noise and maintain structural integrity. As shown in Figure 1 The overall network is composed of three key modules: a multi-scale principal component construction module, a phase reconstruction module (PR-Block) based on residuals, and a denoising network designed for complex input. After differential processing, the multi-scale phase principal component obtains a multi-scale phase residual component, which is then denoised to obtain a multi-scale denoised residual component, and then fused to obtain a multi-scale reconstructed phase.

[0018] Using the method, by hierarchical frequency domain decomposition to extract low-frequency principal components, and then reconstructing high-frequency details layer by layer in the residual domain, a step-by-step optimization process from coarse to fine is constructed, so as to effectively maintain the integrity of the phase structure while suppressing noise. Moreover, by optimizing in the residual domain rather than the overall phase domain, the generation of artifacts can be effectively reduced and the convergence stability can be improved. At the same time, combined with the self-enhancing training strategy based on real noise, the denoising model can continuously evolve and enhance the adaptability and robustness to complex noise environments.

[0019] Through the technical solutions provided by the present application, balanced performance improvement can be achieved in various typical scenarios such as earthquakes, landslides, rainforests and cities, which not only outperforms traditional methods in noise suppression, but also has significant advantages in structure preservation and computational efficiency. The number of multi-scale layers can be flexibly adjusted to adapt to different noise intensity conditions; the iterative reconstruction feature ensures that reliable phase results can still be recovered step by step in extreme noise environments. Therefore, the technical contradiction between denoising effect, structure preservation and computational efficiency in the prior art is effectively solved, providing a practical and reliable technical means for large-scale commercial monitoring and high-precision earth science research.

[0020] The technical solutions of the present application and how the technical solutions of the present application solve the foregoing technical problems will be described in detail below with specific embodiments. Several specific embodiments listed can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0021] Figure 2 A flowchart of the interferometric phase denoising method provided by the embodiments of the present application is shown in Figure 2As shown, the method can include step S201, step S202, step S203 and step S204.

[0022] Step S201, performing multi-scale frequency decomposition on the original interferometric phase map based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components; the size of the frequency domain filtering window increases layer by layer in order from low to high.

[0023] In this step, for the input original interferometric phase map, multi-scale frequency decomposition is performed based on a preset frequency domain filtering window to extract low-frequency principal components and high-frequency detail components respectively, so as to realize multi-scale hierarchical division in the frequency domain without changing the spatial resolution. The multi-scale hierarchical signals obtained after decomposition include several principal component signals with different frequency ranges, and each layer can reflect the structure and noise characteristics at different scales.

[0024] Step S202, calculating the residual components of adjacent hierarchical layers according to the differences between the principal component signals of adjacent hierarchical layers to obtain the residual domain signals corresponding to each hierarchical layer respectively.

[0025] Step S203, using a pre-trained denoising model to perform denoising processing on the residual domain signals to obtain the denoised residual domain signals corresponding to each hierarchical layer respectively.

[0026] Among them, the residual domain signal mainly contains the high-frequency information and noise components corresponding to the layer. For the residual domain signal of each hierarchical layer, a pre-trained denoising model or filtering module is used for noise suppression and detail recovery to obtain the reconstructed residual domain signal. By denoising in the residual domain, direct strong filtering operation on the complete phase map can be avoided, the risk of artifacts can be reduced, and the stability and fidelity of denoising can be improved.

[0027] Step S204, performing phase reconstruction on the denoised residual domain signals corresponding to each hierarchical layer respectively to obtain a target interferometric phase map.

[0028] According to the technical scheme of the embodiment of the present application, the original interference phase map is subjected to multi-scale frequency decomposition based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components, and the size of the frequency domain filtering window is increased layer by layer in order from low to high level. The residual components of adjacent levels are calculated according to the difference between the principal component signals of adjacent levels, and the residual domain signals corresponding to each level are obtained. The residual domain signals are subjected to denoising processing by using a pre-trained denoising model, and the residual domain signals after denoising corresponding to each level are obtained. The residual domain signals after denoising corresponding to each level are subjected to phase reconstruction, and the target interference phase map is obtained. Therefore, by extracting low-frequency principal components through hierarchical frequency domain decomposition and then reconstructing high-frequency details layer by layer in the residual domain, a step-by-step optimization process from coarse to fine is constructed, so that the noise is suppressed while the authenticity and integrity of the phase structure are effectively maintained. Moreover, by optimizing in the residual domain rather than in the overall phase domain, the generation of artifacts can be effectively reduced and the convergence stability can be improved, so that the comprehensive advantages that the related art cannot achieve are realized in terms of noise suppression, detail preservation, calculation efficiency and adaptation to complex noise environment, thereby effectively ensuring the reliability of phase unwrapping and subsequent deformation inversion.

[0029] In some embodiments, when the original interference phase map is subjected to multi-scale frequency decomposition based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components, the following steps A1 and A2 can be performed: Step A1, a plurality of filtering windows corresponding to a plurality of levels are determined; wherein the filtering window corresponding to each level is an integer multiple of the filtering window corresponding to its adjacent next level.

[0030] Step A2, based on the original interference phase signal, the frequency decomposition is performed by using the filtering window corresponding to each level in order from low to high (i.e. from high frequency to low frequency) of the multi-scale level, and the phase signals of the multiple levels are obtained.

[0031] Optionally, the denoising model first constructs a multi-scale phase principal component by using a hierarchical Goldstein filter . Let the original interference phase be , the initial size of the filtering window is 32 pixels, and the filtering window is doubled in order from low to high level, i.e. the filtering window corresponding to the previous level is 2 times the filtering window corresponding to the adjacent next level. The constructed multi-scale principal component signal can be expressed as the following formula:

[0032] wherein, represents a Goldstein filter with a filtering window size of , and is the original interferometric phase. The multi-scale phase principal component signals generated by the hierarchical processing can cover different scale signals from high frequency to low frequency. The top component is a smooth low frequency structure obtained under a larger filter window, with less noise; the bottom component contains more high frequency details, but is more seriously contaminated by noise, see (a, b) in Figure 1 .

[0033] In some embodiments, the residual domain signal includes a high frequency signal and a noise component corresponding to the level. When the residual domain signal is denoised by using a pre-trained denoising model to obtain a denoised residual domain signal corresponding to each level, the high frequency signal in the residual domain signal can be extracted by using the pre-trained denoising model, thereby obtaining the denoised residual domain signal. The training method of the denoising model will be described in detail in the following embodiments.

[0034] In this embodiment, step S202 can be implemented as the following step B1, and step S203 can be implemented as the following step B2: Step B1, for a first layer in the multi-scale level phase signal, a residual high frequency component signal of the first layer is determined according to a low frequency component signal corresponding to a second layer; wherein the first layer is any level in the plurality of levels; and the second layer is a previous level of the first layer.

[0035] Step B2, the residual high frequency component signal is denoised to obtain a denoised interferometric phase signal corresponding to the first layer, i.e. a denoised residual domain signal.

[0036] In some embodiments, when step S204 is performed, it can be specifically implemented as: taking the principal component signal of the lowest level as a reference, the denoised residual domain signals are stacked layer by layer in order from low to high levels to obtain a target interferometric phase map. Specifically, taking the principal component signal of the lowest level as a reference, the reconstructed residual domain signals are stacked layer by layer in order from low to high frequencies to complete the hierarchical phase reconstruction. In the phase reconstruction of each level, the denoised residual of the current level is stacked into the principal component signal of the next level, thereby gradually recovering the high frequency details, and finally outputting the denoised target interferometric phase map, realizing high-precision phase reconstruction considering structure preservation and noise suppression.

[0037] In this embodiment, in order to restore the clean phase, the denoising model introduces a phase reconstruction block (PR-module) at each layer of the multi-scale level data. As shown in (b) of Figure 1 , the PR-module of the first layer first retains the low frequency baseline , and then extracts the residual high frequency component signal, which is represented by the following formula:

[0038] wherein, It primarily contains noise and fine structural information. Unlike traditional filtering that directly affects the entire phase... Denoising, the denoising model exists only in the residual domain. Denoising is performed to limit the optimization space to a smaller range. This not only helps to accelerate convergence and improve stability, but also reduces the risk of generating spurious fringes. Then, a dedicated denoising network is used. For residual high-frequency component signals After restoring the details, the phase reconstructed from the high-frequency information can be represented as:

[0039] This residual-based phase optimization strategy can suppress high-frequency noise while maintaining the integrity of the low-frequency phase structure.

[0040] Reference Figure 1 In parts (b) and (c), the denoising model optimizes phase estimation progressively at multiple levels by cascading multiple PR modules. For Layer configuration, denoising model starts from the coarsest low-frequency baseline Initially, through structure estimation, feature recovery, and detail enhancement, fine-grained phase information is recovered layer by layer. This coarse-to-fine strategy utilizes multi-scale frequency decomposition to suppress noise while maintaining structural integrity. The low-frequency baseline serves as an anchor point, providing a robust phase foundation and reducing artifacts generated when directly applying full-phase filtering.

[0041] Furthermore, the number of layers in multi-scale structural data can be adaptively configured according to noise intensity. Deeper layers of data can be reconstructed starting from cleaner interferometric phases, improving denoising performance under high-noise conditions. It's important to note that, unlike traditional frequency domain filtering which irreversibly loses high-frequency information, each PR-module can recover phase details from the residual domain, giving the network non-destructive, iterative frequency filtering characteristics. Because of this, the denoising model can be applied sequentially multiple times under extreme noise conditions to gradually recover phase information.

[0042] like Figure 1As shown in part (d) of FIG. 1, the denoising model is an encoder-decoder architecture designed for complex residual input. Both the input and output are 2-channel tensors, encoding the real and imaginary components of the complex residual, respectively. The encoder contains four down-sampling blocks (Down1-4) that reduce the spatial resolution by stride convolution, while increasing the channel depth from 256 to 1024. Residual blocks and self-attention mechanisms are embedded in the encoder to capture long-range dependencies and non-local features. The bottleneck part consists of three residual-attention blocks (Mid 1-3) with a fixed channel number of 1024 to aggregate global context at the lowest resolution. The decoder path is symmetric to the encoder, using transposed convolutions for up-sampling (Up 1-4) and fusing the encoder features through skip connections to recover high-frequency details using multi-scale information. The final denoised residual is obtained. The complete network architecture of the denoising model is shown in Table 1.

[0043] Table 1

[0044] In some embodiments, before step S201 is performed, the denoising model is pre-trained based on the structure shown in FIG. 2. In this embodiment, a semi-supervised self-enhancing training strategy is used, and the denoising model is trained using real noise. The training method of the denoising model can include the following steps C1, step C2 and step C3: Figure 1 Step C1, obtaining a sample noise pool and a sample label pool; the sample noise pool includes multi-scale noise samples extracted from the noise residual component of the actual interferogram, and the sample label pool includes multi-scale label samples composed of the residual component of the label phase image.

[0045] Optionally, step C1 is specifically performed as follows: First, an actual interferogram (i.e., a real phase interferogram) is obtained, and a noise residual component is extracted from the actual interferogram to construct a sample noise pool.

[0046] Second, a label phase image corresponding to the actual interferogram is obtained by simulating and combining multiple phase components of the actual interferogram, and a sample label pool is constructed based on the label phase image.

[0047] Third, multi-scale frequency domain decomposition is performed on the noise interferogram in the sample noise pool to obtain multi-scale residual data; and multi-scale frequency domain decomposition is performed on the label phase image in the sample label pool to obtain multi-scale label samples.

[0048] Figure 3 ​The diagram illustrates the training principle of the denoising model. To capture the complete characteristics of real noise and comprehensively cover its feature distribution, sample noise interferograms can be selected from the following four representative scene types: vegetated areas, affected by seasonal changes and dynamic land cover images, exhibiting moderate temporal incoherence; mountainous terrain, commonly characterized by steep slopes and geometric distortion; tropical rainforest areas, with dense canopies leading to continuous and strong volume scattering incoherence; and plains, possessing moderate coherence but exhibiting significant temporal variations due to agricultural activities or hydrological changes (see [link to relevant documentation]). Figure 3 Part (a) in the diagram: Interferogram used for noise extraction.

[0049] Clean sample label maps (i.e., clean phase labels) are generated through simulation by randomly combining four phase components, including trend components, topographic components, atmospheric turbulence components, and deformation components, such as... Figure 3 As shown in section (f), these simulated clean-phase and real-noise interferograms are used to construct their multi-scale residual components. A denoising network is applied to the residual components of the noisy interferogram to extract noise, forming a noise pool; while the residual components of the clean phase constitute the corresponding label pool (e.g., ...). Figure 3 (As shown in sections (d) and (e)). Subsequently, residual sample pairs are randomly drawn from these two pools and combined into synthetic "noisy-clean" sample pairs, which are used to train the denoising branch of MSRD-Net.

[0050] Step C2 involves inputting the training data into the neural network model to be trained, and using the neural network model to predict the clean phase map corresponding to the actual interferogram. The training data is determined based on a sample noise pool and a sample label pool. Optionally, multiple data pairs are randomly selected from the sample noise pool and the sample label pool as training data.

[0051] Optionally, after inputting the training data into the neural network model to be trained, the neural network model is used to extract the predicted noise in the actual interferogram; then the sample noise pool is updated based on the predicted noise.

[0052] Step C3: Based on the predicted clean phase map and the sample label pool, iteratively train the neural network model to obtain the trained denoising model.

[0053] In this embodiment, the entire training process employs an iterative self-enhancing mechanism, enabling the model and dataset to co-evolve during training. Specifically, in each iteration, MSRD-Net is first applied to the real interferogram to estimate the clean phase and extract the corresponding noise; the extracted noise is then updated in the noise pool and used to generate new synthetic training sample pairs for fine-tuning MSRD-Net in the next round. As the iteration progresses, the model's denoising capability continuously improves, thereby enabling more accurate noise extraction.

[0054] In some embodiments, when iteratively training the neural network model based on the predicted clean phase map and the sample label pool to obtain the trained denoising model (i.e., step C3), the following steps can be specifically performed: First, based on the predicted clean phase map and the sample label pool, the phase loss function and the magnitude loss function of the neural network model are determined. The phase loss function is used to represent the phase difference between the predicted clean phase map and the sample label map. The magnitude loss function is used to represent the difference between the magnitude corresponding to the predicted clean phase map and 1.

[0055] Secondly, the target loss function of the neural network model is determined based on the phase loss function and the magnitude loss function.

[0056] Next, the model parameters of the neural network model are adjusted according to the target loss function to obtain the denoising model.

[0057] In this embodiment, the phase loss function This can be expressed as the following formula:

[0058] Magnitude loss function This can be expressed as the following formula:

[0059] in, and These represent the first and second phases on the phase diagram, respectively. i The predicted complex phase value and the true complex phase value of each pixel. Indicates complex conjugation. Represents the phase angle of a complex value. Represents the magnitude of a complex value. Phase loss function. The goal is to recover the filtered phase angle, while the magnitude loss function... This is used to ensure that the signal maintains its normalized magnitude. The target loss function of the neural network model. This can be expressed as the sum of two losses, as shown in the following formula:

[0060] In some embodiments, to evaluate denoising performance on real interferograms lacking true labels (i.e., noisy phase interferograms), multiple complementary metrics are used to evaluate model performance. Specifically, evaluation metrics for the denoising model are determined, and the model performance of the denoising model is evaluated based on these metrics.

[0061] The evaluation indexes include at least one of the following: LRCR (Log Residue Count Reduction), MNI (Mean Noise Intensity), MMCC (Mean Main Component Coherence), EPR (Edge Preservation Ratio), and data processing time length. These evaluation indexes can comprehensively evaluate different aspects of the denoising quality of the model, including phase continuity, residual noise level, structural integrity, and calculation efficiency. Each evaluation index is described in detail below.

[0062] Residual points are phase discontinuities usually caused by noise or artifacts, which can cause errors in the phase unwrapping process because they destroy local phase continuity. Therefore, reducing residual points can generally improve phase smoothness and unwrapping performance. To quantify this improvement, a percentage index, Log Residue Count Reduction LRCR, is defined and expressed as follows:

[0063] wherein and represent the number of residual points in the original phase interferogram and the target phase interferogram after denoising, respectively. The higher the LRCR value, the more residual points are reduced relative to the original phase interferogram.

[0064] The Mean Noise Intensity MNI is used to measure the average predicted noise level after denoising, which is a general indicator of residual noise, and is defined as follows:

[0065] wherein represents the noise intensity of the i-th pixel on the phase interferogram, and N is the total number of pixels.

[0066] The Mean Main Component Coherence MMCC is used to measure the structural similarity between the main components of the target phase interferogram after denoising and the original phase interferogram. A good denoising method should not only reduce noise but also maintain the integrity of the effective signal without introducing false fringes. Therefore, the Mean Main Component Coherence MMCC is defined and expressed as follows:

[0067] wherein and represent the i-th and j-th main components of the phase interferogram, respectively. iThe principal component of a pixel and the complex phase representation of the target phase interferogram. The principal component is obtained by applying a low-pass filter to the original phase interferogram. MMCC is the mean coherence between the principal component and the denoised output, the higher the value indicates the better the structural consistency is preserved.

[0068] While suppressing noise, over-smoothing can remove meaningful structures such as fringe and building-related phase gradients. To evaluate the ability to preserve these structural features, edge preservation ratio (EPR) is introduced. This metric quantifies the proportion of phase edges in the original phase interferogram that can still be detected after denoising. Specifically, the phase edge proportion is identified by applying a Canny operator to the gradient magnitude map, and the calculation of EPR is represented as follows:

[0069] wherein, represents the i-th pixel in the input original phase interferogram, i represents the pixel set of the i-th edge in the original phase interferogram, represents the pixel set of the corresponding edge in the denoised target phase interferogram, represents the number of overlapping pixels. The higher the EPR value indicates the better the structural details are preserved, and the less over-smoothing is.

[0070] Data processing duration is used to measure the time required to denoise a single phase interferogram, reflecting the computational efficiency of the method, which is an important reference for large-scale or near real-time applications.

[0071] In some embodiments, MSRD-Net is scalable. By increasing the number of multi-scale structure layers, MSRD-Net effectively doubles the frequency domain filter window size at each stage, enabling the gradual extraction of cleaner principal components. This feature allows users to flexibly adjust the denoising strength according to noise conditions. To demonstrate this capability, models with different scale layer numbers (from one layer to five layers) are applied to a certain real original interferometric phase image, as shown in Figure 4 It can be seen that deeper scale settings gradually improve the denoising performance, with the noisy phase gradually becoming clear, while the effective phase structure is still preserved.

[0072] To quantitatively illustrate this scalability, Figure 5The scalability of phase reconstruction performance of MSRD-Net is demonstrated, including data processing time and other denoising indicators. The horizontal axis represents the number of layers. The data processing time (bar chart, left y-axis) monotonically increases with the number of layers. The denoising indicators, LRCR, MNI, MMCC, and EPR, are shown in the form of broken lines on the double right y-axis to accommodate their different numerical ranges. With the increase of multi-scale levels, LRCR steadily rises, MNI steadily decreases (indicating improved denoising effect), MMCC reaches a peak at the 4th layer and slightly decreases at the 5th layer, and EPR gradually decreases, reflecting the trade-off between noise suppression and structure preservation. This visualization result highlights the balance between computational cost and denoising performance, providing practical reference for selecting the optimal level under different noise conditions.

[0073] However, when the multi-scale level is too large, a very large frequency domain window may produce unreasonable estimates in the completely incoherent area. For example, in the ROI (lake) shown in the figure, the 4th and 5th layer results show false phase recovery on the water body. This phenomenon only appears in the extreme parameter setting used here to illustrate scalability. In practical applications, such a large window should be used with caution and is generally not recommended; incoherent areas are more suitable for processing through masking in subsequent processing.

[0074] In addition, while MSRD-Net can adaptively adjust the denoising intensity through extended reconstruction and effectively deal with moderate noise conditions, it is still challenging to obtain completely clean results at once under extreme noise. To address this limitation, another inherent property of the method - iterative property is utilized, by separating stable principal components through frequency domain filtering and recovering lost details from residual signals, MSRD-Net's function is almost equivalent to a non-destructive filter: in the worst case, the result will degrade to the principal component. This property makes it naturally suitable for iterative denoising, gradually refining the phase in repeated iterations while not compromising structural integrity.

[0075] For example, MSRD-Net is applied to a real phase interferogram of a certain earthquake. The original phase interferogram is dominated by high-intensity noise, almost completely masking the deformation signal. The iterative reconstruction process of MSRD-Net is as follows: Figure 6As shown, after three iterations, it gradually recovers the residual phase information, and finally obtains a clean and structurally reliable result. Through experiments, compared with other methods, the interferometric phase denoising method of the application will produce block artifacts for spatial domain methods (such as Boxcar, InSAR-BM3D, NL-InSAR), and will generate false interference fringes for data-driven methods (such as PhiNet, MONet), which will destroy the phase continuity and reduce the reliability of the unwrapping result. In contrast, the present method realizes smooth denoising and reliable unwrapping, and exhibits its robustness in harsh real environments.

[0076] Therefore, as a hierarchical phase reconstruction method, MSRD-Net can balance noise suppression, phase integrity preservation and subtle structure recovery in InSAR data. It combines the phase robustness of frequency domain filtering with the nonlinear modeling capability of deep learning, and at the same time, through training under real noise conditions, it ensures its good generalization ability in various sensors and earth science environments. Extensive evaluation on real scenes shows that the present method can effectively suppress noise and preserve the real phase structure without introducing artifacts, thereby realizing reliable phase unwrapping and deformation analysis. In addition to denoising accuracy, MSRD-Net has two unique advantages: its scalable multi-scale design allows flexible adjustment of denoising strength according to noise conditions, and its nearly lossless iterative nature provides a practical solution for extreme noise. Overall, MSRD-Net provides a robust and flexible framework for InSAR phase denoising, which can support large-scale business monitoring and meet the needs of high-precision earth science research, while improving the reliability of surface deformation measurement and interpretation.

[0077] Corresponding to the application scenarios and methods of the method provided by the embodiments of the application, the embodiments of the application also provide an interferometric phase denoising device.

[0078] Figure 7 A block diagram of the interferometric phase denoising device provided by the embodiments of the application is shown in FIG. 7. Figure 7 As shown in FIG. 7, the interferometric phase denoising device comprises: A decomposition module 71 is configured to perform multi-scale frequency decomposition on the original interferometric phase image based on a preset frequency domain filtering window, to obtain hierarchical phase signals containing different frequency components; the size of the frequency domain filtering window increases layer by layer in order from low to high; A calculation module 72 is configured to calculate residual components of adjacent hierarchical levels according to the difference between the principal component signals of the adjacent hierarchical levels, to obtain residual domain signals corresponding to each hierarchical level respectively; A processing module 73 is configured to perform denoising processing on the residual domain signals by using a pre-trained denoising model, to obtain denoised residual domain signals corresponding to each hierarchical level respectively. The reconstruction module 74 is configured to perform phase reconstruction on the denoised residual domain signals corresponding to the respective levels to obtain a target interference phase map. In some embodiments, when the decomposition module 71 performs multi-scale frequency decomposition on the original interference phase map based on the preset frequency domain filtering window to obtain the level phase signals containing different frequency components, the following steps are performed: determining a plurality of filtering windows corresponding to a plurality of levels; wherein the filtering window corresponding to each level is an integer multiple of the filtering window corresponding to its adjacent next level; based on the original interference phase signal, sequentially performing frequency decomposition on each level corresponding to the filtering window in the order from low to high, to obtain the phase signals of the plurality of levels.

[0079] In some embodiments, the residual domain signal includes a high frequency signal and a noise component corresponding to a level; When the processing module 73 performs denoising processing on the residual domain signal using a pre-trained denoising model to obtain denoised residual domain signals corresponding to the respective levels, the following steps are performed: extracting the high frequency signal in the residual domain signal using a pre-trained denoising model to obtain the denoised residual domain signal.

[0080] In some embodiments, when the reconstruction module 74 performs phase reconstruction on the denoised residual domain signals corresponding to the respective levels to obtain a target interference phase map, the following steps are performed: using the principal component signal of the lowest level as a reference, layer by layer superimposing the denoised residual domain signals in the order from low to high to obtain the target interference phase map.

[0081] In some embodiments, the device further comprises: The acquisition module is configured to acquire a sample noise pool and a sample label pool; the sample noise pool includes multi-scale noise samples extracted from the noise residual components of the actual interference map, and the sample label pool includes multi-scale label samples composed of residual components of the label phase map; The prediction module is configured to input the training data into the neural network model to be trained, and use the neural network model to predict a predicted clean phase map corresponding to the actual interference map; the training data is determined based on the sample noise pool and the sample label pool; The training module is configured to perform iterative training on the neural network model based on the predicted clean phase map and the sample label pool to obtain the trained denoising model.

[0082] In some embodiments, when the acquisition module acquires the sample noise pool and the sample label pool, the following steps are performed: obtaining the actual interferogram, extracting noise residual components from the actual interferogram to construct the sample noise pool; simulating and combining a plurality of phase components of the actual interferogram to obtain a label phase map corresponding to the actual interferogram, and constructing the sample label pool based on the label phase map; performing multi-scale frequency domain decomposition on the noise interferogram in the sample noise pool to obtain the multi-scale residual data, and performing multi-scale frequency domain decomposition on the label phase map in the sample label pool to obtain the multi-scale label sample.

[0083] In some embodiments, the device further comprises: The extraction module is configured to extract, by using the neural network model, a predicted noise in the actual interferogram after the training sample is input into the neural network model to be trained. The update module is configured to update the sample noise pool based on the predicted noise.

[0084] In some embodiments, the training module performs the following steps when iteratively training the neural network model based on the predicted clean phase map and the sample label pool to obtain the denoising model after training: determining a phase loss function and a modulus loss function of the neural network model based on the predicted clean phase map and the sample label pool; the phase loss function is used to represent a phase difference between the predicted clean phase map and the sample label map; and the modulus loss function is used to represent a difference between a modulus corresponding to the predicted clean phase map and 1. determining a target loss function of the neural network model based on the phase loss function and the modulus loss function; adjusting model parameters of the neural network model based on the target loss function to obtain the denoising model.

[0085] In some embodiments, the device further comprises: The determination module is configured to determine an evaluation index of the denoising model; the evaluation index comprises at least one of the following: log residual point reduction, average noise intensity, average principal component coherence, edge retention rate, and data processing time length. The evaluation module is configured to evaluate model performance of the denoising model based on the evaluation index.

[0086] The functions of each module in each device in the embodiments can be referred to the corresponding description in the above method, and have corresponding beneficial effects, which will not be repeated here.

[0087] Figure 8This is a block diagram for implementing the electronic device provided in the embodiments of this application. Figure 8 As shown, the electronic device includes a memory 801 and a processor 802. The memory 801 stores a computer program that can run on the processor 802. When the processor 802 executes the computer program, it implements the method described in the above embodiments. The number of memories 801 and processors 802 can be one or more. In a specific implementation, the electronic device may also include a communication interface 803 for communicating with external devices and performing data exchange and transmission.

[0088] In practical implementation, if the memory 801, processor 802, and communication interface 803 are implemented independently, they can be interconnected via a bus to complete communication. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0089] Optionally, in a specific implementation, if the memory 801, the processor 802, and the communication interface 803 are integrated on a single chip, then the memory 801, the processor 802, and the communication interface 803 can communicate with each other through an internal interface.

[0090] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this application.

[0091] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method provided in this application.

[0092] This application also provides a chip including a processor for calling and executing instructions stored in a memory, causing a communication device with the chip installed to perform the method provided in this application.

[0093] The embodiment of the present application further provides a chip, comprising: an input interface, an output interface, a processor and a memory, the input interface, the output interface, the processor and the memory are connected through internal connection paths, the processor is used for executing the code in the memory, and when the code is executed, the processor is used for executing the method provided by the embodiment of the present application.

[0094] It should be understood that the processor described above can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. It should be noted that the processor can be a processor supporting an advanced RISC machine (ARM) architecture.

[0095] Further, the memory can include a read-only memory and a random access memory, optionally. The memory can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memory. The non-volatile memory can include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM), or a flash memory, among others. The volatile memory can include a random access memory (RAM), which is used as an external cache. By way of example, and not limitation, many forms of RAM are available. The RAM can include a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate SDRAM (DDR SDRAM), an enhanced SDRAM (ESDRAM), a Sync link DRAM (SLDRAM), and a direct Rambus RAM (DR RAM), among others.

[0096] In the above-described embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the present disclosure are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium.

[0097] In the description of the application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. Moreover, the specific features, structures, materials or characteristics described can be combined in any suitable manner in one or more embodiments or examples. In addition, different embodiments or examples described in the specification and characteristics of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0098] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0099] Any process or method described in the flowchart or otherwise described herein can be understood as representing a module, segment, or portion of code that includes executable instructions for implementing specific logical functions or steps. And the scope of the preferred embodiments of the application includes additional implementations, in which the functions can be performed in an order other than that shown or discussed, including in a substantially simultaneous manner or in reverse order, according to the functions involved.

[0100] The logic and / or steps described in the flowchart or otherwise described herein, for example, can be considered as a list of executable instructions for implementing the logic function, which can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus, such as a computer-based system, a system including a processor, or other system that can fetch instructions from the instruction execution system, device or apparatus and execute the instructions, or in conjunction with these instruction execution systems, devices or apparatus.

[0101] It should be understood that parts of the application can be implemented in hardware, software, firmware or a combination thereof. In the above-described embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above-described embodiment method can be instructed by the relevant hardware through a program, which can be stored in a computer-readable storage medium, and the program includes one or a combination of the steps of the method embodiment when executed.

[0102] In addition, each of the function units in each embodiment of the present application can be integrated in one processing module, or each unit can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module. When the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium. The storage medium can be a read-only memory, a magnetic disk or an optical disk, etc.

[0103] The above is only exemplary embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An interferometric phase denoising method, characterized in that, include: The original interferometric phase map is decomposed into multiple scales based on a preset frequency domain filtering window to obtain a hierarchical phase signal containing different frequency components. The size of the frequency domain filtering window increases progressively in order of increasing hierarchy; Based on the differences between the principal component signals of adjacent levels, the residual components of the adjacent levels are calculated to obtain the residual domain signals corresponding to each level. The residual domain signal is denoised using a pre-trained denoising model to obtain the denoised residual domain signal corresponding to each level. Phase reconstruction is performed on the denoised residual domain signals corresponding to each level to obtain the target interferometric phase map.

2. The method according to claim 1, characterized in that, The process of performing multi-scale frequency decomposition on the original interferometric phase map based on a preset frequency domain filtering window to obtain hierarchical phase signals containing different frequency components includes: Determine multiple filtering windows corresponding to multiple levels; wherein the filtering window corresponding to each level is an integer multiple of the filtering window corresponding to its adjacent next level; Based on the original interference phase signal, frequency decomposition is performed sequentially using the filtering window corresponding to each level, in order from low to high level, to obtain the phase signals of the multiple levels.

3. The method according to claim 1, characterized in that, The residual domain signal includes high-frequency signals and noise components of the corresponding level; The step of using a pre-trained denoising model to denoise the residual domain signal to obtain the denoised residual domain signal corresponding to each level includes: Using a pre-trained denoising model, the high-frequency signal in the residual domain signal is extracted to obtain the denoised residual domain signal.

4. The method according to claim 1, characterized in that, The step of performing phase reconstruction on the denoised residual domain signals corresponding to each level to obtain the target interferometric phase map includes: Using the principal component signal at the lowest level as a reference, the denoised residual domain signal is superimposed layer by layer in ascending order of level to obtain the target interferometric phase map.

5. The method according to claim 1, characterized in that, The method further includes: Obtain a sample noise pool and a sample label pool; the sample noise pool includes multi-scale noise samples extracted from the noise residual components of the actual interferogram, and the sample label pool includes multi-scale label samples composed of the residual components of the label phase map. The trained data is input into the neural network model to be trained, and the neural network model is used to predict the predicted clean phase map corresponding to the actual interferogram; the training data is determined based on the sample noise pool and the sample label pool. Based on the predicted clean phase map and the sample label pool, the neural network model is iteratively trained to obtain the trained denoising model.

6. The method according to claim 5, characterized in that, The acquisition of the sample noise pool and sample label pool includes: Obtain the actual interferogram, and extract the noise residual component from the actual interferogram to construct the sample noise pool; Multiple phase components of the actual interferogram are simulated and combined to obtain a label phase map corresponding to the actual interferogram, and the sample label pool is constructed based on the label phase map; The noise interferogram in the sample noise pool is decomposed into multi-scale frequency domain to obtain the multi-scale residual data; the label phase map in the sample label pool is decomposed into multi-scale frequency domain to obtain the multi-scale label samples.

7. The method according to claim 5, characterized in that, After inputting the training data into the neural network model to be trained, the method further includes: Using the neural network model, the predicted noise in the actual interferogram is extracted; The sample noise pool is updated based on the predicted noise.

8. The method according to claim 5, characterized in that, The step of iteratively training the neural network model based on the predicted clean phase map and the sample label pool to obtain the trained denoising model includes: Based on the predicted clean phase map and the sample label pool, the phase loss function and the magnitude loss function of the neural network model are determined; the phase loss function is used to represent the phase difference between the predicted clean phase map and the sample label map; the magnitude loss function is used to represent the difference between the magnitude corresponding to the predicted clean phase map and 1. The target loss function of the neural network model is determined based on the phase loss function and the magnitude loss function. The model parameters of the neural network model are adjusted according to the target loss function to obtain the denoising model.

9. The method according to claim 1, characterized in that, The method further includes: Determine the evaluation metrics for the denoising model; the evaluation metrics include at least one of the following: logarithmic residual point reduction, average noise intensity, average principal component coherence, edge preservation rate, and data processing time; The performance of the denoising model is evaluated based on the evaluation metrics.

10. An interferometric phase denoising device, characterized in that, include: The decomposition module is used to perform multi-scale frequency decomposition on the original interferometric phase map based on a preset frequency domain filtering window to obtain a hierarchical phase signal containing different frequency components. The size of the frequency domain filtering window increases progressively in order of increasing hierarchy; The calculation module is used to calculate the residual components of adjacent levels based on the differences between the principal component signals of adjacent levels, so as to obtain the residual domain signals corresponding to each level. The processing module is used to denoise the residual domain signal using a pre-trained denoising model to obtain the denoised residual domain signal corresponding to each level. The reconstruction module is used to perform phase reconstruction on the denoised residual domain signals corresponding to each level to obtain the target interferometric phase map.

Images