Differential InSAR system, method, application and readable storage medium

Abstract

The present invention proposes a differential InSAR system, method, application and readable storage medium. The system includes laser emission, radar detection and adaptive optical compensation modules. The laser emission module supplies energy to the system and splits light; the radar detection module uses two symmetrical branches to collaboratively detect targets, collect and process echoes to achieve the purpose of differential detection; the adaptive optical compensation module uses deep learning and wavefront detection mechanisms to correct the laser wavefront distortion caused by atmospheric turbulence in real time to ensure accurate and stable long-distance remote sensing. Its innovation lies in combining deep learning to improve the performance of adaptive optics and suppress phase noise. It has far-reaching significance in many fields such as terrain mapping and geological disaster monitoring. It can accurately measure terrain elevation and small surface displacements, such as keenly capturing early millimeter-level changes in landslide warnings to seize opportunities for disaster prevention and control.

Description

Technical Field

[0001] The present invention relates to the field of radar detection technology, and in particular to a differential InSAR system, method, application and readable storage medium. Background Art

[0002] In the field of modern radar detection, synthetic aperture radar (SAR) technology continues to evolve and plays a key role in obtaining surface information. Among them, differential InSAR (D-InSAR) technology is particularly prominent. It can accurately monitor small surface deformations by interfering SAR images acquired at different times, and is widely used in many fields such as terrain mapping, geological disaster warning, and urban infrastructure monitoring.

[0003] However, most current differential InSAR systems face severe challenges in long-distance remote sensing. For existing technologies represented by CN1329743C and CN101551455B, when the radar is in a long-distance working state, atmospheric turbulence becomes a key factor restricting performance improvement. Turbulence causes complex distortion of the laser radar wavefront. Due to the lack of an effective correction mechanism, the existing system cannot accurately compensate for the wavefront distortion, which in turn causes significant phase noise. This phase noise seriously interferes with the quality of the interference pattern, causing phase information distortion. In high-precision terrain mapping, the details of small terrain undulations are lost; in geological disaster monitoring scenarios, such as early small displacement monitoring of landslides, noise interference cannot capture changes in a timely and accurate manner, greatly weakening the system reliability and measurement accuracy, making it difficult to meet the growing demand for high-precision, stable and reliable long-distance surface monitoring, limiting the in-depth application and expansion of technology in key areas such as deformation monitoring of large-scale water conservancy facilities and prevention and control of geological disasters in remote mountainous areas. Summary of the invention

[0004] The embodiments of the present invention provide a differential InSAR system, method, application and readable storage medium, which are aimed at the problems such as phase noise caused by turbulence when using long-distance D-InSAR in the current technology.

[0005] The core technology of this invention is mainly to improve the correction bandwidth of the adaptive optical system through deep learning technology, and to solve the phase noise problem caused by turbulence interference of the laser radar wavefront information through the adaptive optical system.

[0006] In a first aspect, the present invention provides a differential InSAR system, comprising:

[0007] A laser emission module, used for generating and distributing a laser beam;

[0008] Radar detection module, used to detect targets and collect and process signals;

[0009] Adaptive optical compensation module, used to correct phase distortion caused by atmospheric turbulence;

[0010] Among them, the radar detection module contains two detection branches with symmetrical structures and working in coordination. The two detection branches detect the target simultaneously and collect and process the echo signals to realize the differential detection function and improve the system reliability and accuracy; the adaptive optical compensation module is based on deep learning technology and wavefront detection processing mechanism to obtain and compensate for the laser wavefront distortion caused by atmospheric turbulence in real time to ensure the differential InSAR system to achieve stable and accurate detection in long-distance remote sensing.

[0011] Further, the adaptive optical compensation module includes an adaptive optical system laser, a digital micromirror device, and a wavefront sensor;

[0012] The adaptive optical system laser emits a measuring laser when the tunable laser of the laser emission module is working. After being affected by atmospheric turbulence, the scattered wavefront is received and compressed by the digital micromirror device and received by the wavefront sensor within one exposure time.

[0013] Reconstruct continuous images from a single exposure of the wavefront sensor with the help of a trained deep learning network model to improve the frame rate and closed-loop bandwidth;

[0014] The ideal wavefront deviation is calculated based on wavefront reconstruction, the conjugate wavefront execution amount is determined, and the first deformable mirror and the second deformable mirror of the radar detection module are driven to compensate the radar receiving laser wavefront to achieve the purpose of anti-turbulence.

[0015] Furthermore, when the radar detection module works in a single-flight mode, the calibration light of the adaptive optical compensation module is located at the radar transmitting end to detect the atmospheric turbulence phase distortion in real time; the wavefront sensor and the deformable mirror are placed in the radar receiving end, and the receiving antenna receives the target reflected echo. The wavefront detector first measures the distorted calibration light, and the compensation coefficient obtained by the algorithm is sent to the deformable mirror to remove the distortion, and the corrected echo is received by the radar receiving end.

[0016] Furthermore, the laser emission module includes a tunable laser, which splits the laser into two paths through a beam splitter to enter the radar detection module.

[0017] Further, the radar detection module includes a master radar and a slave radar;

[0018] The two lasers emitted by the laser transmitting module are collimated by a collimator, polarized by a polarizer, filtered by a pinhole, and polarized by a half-wave plate in their respective branches of the main radar and the slave radar. They are then separated into local oscillator laser and outgoing laser by a polarization beam splitter. The local oscillator laser enters the receiving optical path through a reflector and a quarter-wave plate and its polarization state is transformed. After the outgoing laser irradiates the target, the reflected light is corrected for the wavefront phase by a deformable mirror and then mixed with the local oscillator laser. Part of the light is observed by a charge-coupled device to obtain the interference pattern and target alignment, and the remaining part is filtered by a pinhole and then detected by an avalanche photodiode for time series for synthetic aperture processing.

[0019] Further, the slave radar includes a first collimator, a first polarizer, a first pinhole, a first half wave plate, a first polarization beam splitter, a first quarter wave plate, a second quarter wave plate, a first reflector, a second polarization beam splitter, a first charge coupled device, a second pinhole, a first avalanche photodiode, and a first deformable mirror;

[0020] The main radar includes a second collimator, a second polarizer, a third pinhole, a second half wave plate, a third polarization beam splitter, a third quarter wave plate, a fourth quarter wave plate, a second reflector, a fourth polarization beam splitter, a second charge coupled device, a fourth pinhole, a second avalanche photodiode, and a second deformable mirror.

[0021] Furthermore, the deep learning technology based on the adaptive optical compensation module adopts the EfficientSCI deep learning network model.

[0022] In a second aspect, the present invention provides a control method for a differential InSAR system, comprising:

[0023] S00, the laser emission module generates and distributes the laser beam into the radar detection module;

[0024] S10, the two detection branches of the radar detection module detect the target simultaneously and collect and process the echo signals to realize the differential detection function;

[0025] S20, the adaptive optical compensation module uses deep learning technology and wavefront detection processing mechanism to obtain and compensate for laser wavefront distortion caused by atmospheric turbulence in real time.

[0026] In a fourth aspect, the present invention provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute the process, and the process includes the control method according to the above.

[0027] The main contributions and innovations of the present invention are as follows:

[0028] 1. Effectively suppress turbulent phase noise: In long-distance D-InSAR usage scenarios, the existing technology cannot correct wavefront distortion, resulting in significant phase noise in the radar system, which seriously interferes with measurement accuracy. The present invention relies on the exquisite collaboration of deep learning technology and adaptive optical systems to accurately correct the phase distortion caused by turbulence in the observation area in real time. By real-time monitoring of the impact of atmospheric turbulence on the laser wavefront, advanced algorithms are used to drive the deformable mirror to implement displacement compensation, effectively suppressing phase noise. In terrain surveying and mapping, this technology greatly improves the accuracy of terrain elevation measurement. In the field of geological disaster monitoring, it can keenly capture tiny surface displacement changes, such as millimeter-level displacement in the early stage of landslides, to gain critical time for disaster warning, and greatly enhance the system's detection reliability and accuracy in complex atmospheric environments.

[0029] 2. Significantly improve the performance of adaptive optical systems: The present invention innovatively introduces deep learning technology to optimize adaptive optical systems. The EfficientSCI model is used to process wavefront sensor data to reconstruct continuous images from a single exposure, which greatly increases the frame rate of the wavefront sensor and expands the closed-loop bandwidth of the adaptive system. Compared with traditional adaptive optical systems, the present invention can respond more quickly to changes in atmospheric turbulence and compensate for rapidly changing turbulence interference in a timely manner. In urban infrastructure monitoring, in the face of complex airflow interference between high-rise building clusters, the system can stably and accurately monitor tiny deformations of bridges and buildings, providing solid technical support for urban safety operation and maintenance, and effectively broadening the scope of application and application scenarios of the system.

[0030] 3. Enhance the overall stability and adaptability of the system: Its unique architecture and technology integration give the system strong stability and adaptability. It always maintains high-precision detection performance under a variety of complex climatic and geographical conditions. Whether it is the strong winds and low temperatures in high altitude areas that lead to increased atmospheric turbulence, or the humidity and salt erosion challenges in coastal areas, the system can be adaptively adjusted and operate stably and reliably. For long-term monitoring tasks, such as deformation monitoring of large-scale water conservancy facilities, it continues to accurately collect and analyze data for many years, providing a coherent and accurate basis for facility safety assessment and maintenance planning, effectively reducing the interference and impact of environmental factors on system performance, extending the service life of the system, and improving the cost-effectiveness of the entire life cycle. It plays a key role in many fields such as global geographic information collection and infrastructure security.

[0031] The details of one or more embodiments of the invention are set forth in the following drawings and description so that other features, objects, and advantages of the invention are more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0033] Figure 1 is a radar transceiver optical path structure diagram of a differential InSAR system according to an embodiment of the present invention;

[0034] Figure 2 is a coordinate relationship diagram between a radar airborne platform and a target according to an embodiment of the present invention;

[0035] Figure 3 is a diagram of radar transmission and receiving aperture distribution according to an embodiment of the present invention;

[0036] Figure 4 is a spatial relationship diagram between the interference part of the radar transceiver and the target according to an embodiment of the present invention;

[0037] Figure 5 is a spatial coordinate relationship diagram of a radar and a target according to an embodiment of the present invention;

[0038] Figure 6 is a diagram showing the relationship between radar transmitting and receiving aperture layout and target coordinates according to an embodiment of the present invention;

[0039] Figure 7 4 is a system block diagram of a differential InSAR system according to an embodiment of the present invention.

[0040] In the figure, 1, tunable laser; 2, slave radar part; 3, master radar part; 4, target; 5, beam splitter; 6-1, adaptive optical system laser; 6-2, wavefront sensor; 6-3, digital micromirror device;

[0041] 2-1, first collimator; 2-2, first polarizer; 2-3, first pinhole; 2-4, first half-wave plate; 2-5, first polarization beam splitter; 2-6, first quarter-wave plate; 2-7, second quarter-wave plate; 2-8, first reflector; 2-9, second polarization beam splitter; 2-10, first charge coupled device; 2-11, second pinhole; 2-12, first avalanche photodiode; 2-13, first deformable mirror;

[0042] 3-1, second collimator; 3-2, second polarizer; 3-3, third pinhole; 3-4, second half-wave plate; 3-5, third polarization beam splitter; 3-6, third quarter-wave plate; 3-7, fourth quarter-wave plate; 3-8, second reflector; 3-9, fourth polarization beam splitter; 3-10, second charge-coupled device; 3-11, fourth pinhole; 3-12, second avalanche photodiode; 3-13, second deformable mirror. DETAILED DESCRIPTION

[0043] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0044] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0045] Embodiment 1

[0046] The present invention aims to propose a differential InSAR system, specifically, referring to Figure 1-7 , the system comprising:

[0047] Tunable laser 1, 1-to-2 optical fiber beam splitter 5, Slave radar part 2, Master radar part 3, target 4, adaptive optical system laser 6-1, wavefront sensor 6-2, digital micromirror device 6-3;

[0048] The slave radar part 2 (slave radar) further includes a first collimator 2-1, a first polarizer 2-2, a first aperture 2-3, a first half wave plate 2-4, a first polarization beam splitter 2-5, a first quarter wave plate 2-6, a second quarter wave plate 2-7, a first reflector 2-8, a second polarization beam splitter 2-9, a first charge coupled device 2-10, a second aperture 2-11, a first avalanche photodiode 2-12, and a first deformable mirror 2-13;

[0049] The Master radar part 3 (main radar) further includes a second collimator 3-1, a second polarizer 3-2, a third aperture 3-3, a second half wave plate 3-4, a third polarization beam splitter 3-5, a third quarter wave plate 3-6, a fourth quarter wave plate 3-7, a second reflector 3-8, a fourth polarization beam splitter 3-9, a second charge coupled device 3-10, a fourth aperture 3-11, a second avalanche photodiode 3-12, and a second deformable mirror 3-13.

[0050] In this embodiment, the tunable laser 1 splits the laser into two paths through the beam splitter 5, one path enters the slave radar part 2, and the other path enters the master radar part 3. The two paths have exactly the same components and detect targets simultaneously.

[0051] The laser emitted from the optical fiber is collimated by the first collimator 2-1 and the second collimator 3-1, and then polarized by the first polarizer 2-2 and the second polarizer 3-2. The first small hole 2-3 and the third small hole 3-3 perform low-pass filtering. The first half wave plate 2-4 and the second half wave plate 3-4 rotate the polarization state of the polarized laser beam to a suitable angle to control the power ratio of the local oscillator laser and the echo signal of the Master and Slave radar parts. The first polarization beam splitter 2-5 and the third polarization beam splitter 3-5 divide the light beam into a local oscillator laser and an emitted laser. The local oscillator laser is reflected by the first quarter wave plate 2-7 and the second quarter wave plate 3-7 and the first reflector 2-8 and the second reflector 3-8 into the receiving optical path, and the polarization The polarization state is changed from vertical polarization to horizontal polarization. After the outgoing laser passes through the first quarter wave plate 2-6 and the third quarter wave plate 3-6, it irradiates the surface of the target 4. The backscattered laser passes through the first deformable mirror 2-13 and the second deformable mirror 3-13 to correct the distorted wavefront phase. After passing through the second polarization beam splitter 2-9 and the fourth polarization beam splitter 3-9, it is mixed with the local oscillator laser. A part of it is observed by the first charge coupled device 2-10 and the second charge coupled device 3-10 to obtain the interference pattern and check the target alignment. A part of it is low-pass filtered by the second aperture 2-11 and the fourth aperture 3-11, and then detected by the first avalanche photodiode 2-12 and the first avalanche photodiode 3-12 as a time series for synthetic aperture processing.

[0052] The adaptive optical system laser 6-1, the wavefront sensor 6-2, and the digital micromirror device 6-3 constitute an adaptive optical module (adaptive optical system), such as Figure 7 As shown in the figure, the calibration light of the adaptive optical system (AO) is located in the radar transmitting part and is used to detect the phase distortion caused by atmospheric turbulence in real time. The receiving antenna receives the echo information reflected from the target. The distorted calibration light is first detected by the wavefront detector, and then the compensation coefficient is calculated by the wavefront reconstruction algorithm and sent to the deformable mirror to remove the phase distortion caused by turbulence. Finally, the corrected echo information is received by the radar receiver. The measurement process of the adaptive optical system is as follows:

[0053] While the tunable laser 1 is working, the adaptive optical system laser 6-1 is also emitting a measuring laser. After being affected by atmospheric turbulence, the scattered wavefront is first received by the digital micromirror device 6-3, compressed and encoded N times, and then received by the wavefront sensor 6-2 within one exposure time. Through the trained EfficientSCI deep learning network model, N temporally continuous images are reconstructed from the one-time exposure result of the wavefront sensor 6-2, thereby improving the frame rate of the wavefront sensor 6-2, which is used to improve the overall closed-loop bandwidth of the adaptive system, thereby obtaining the influence of the atmosphere on the laser wavefront at the current time, and calculating the deviation from the ideal wavefront through wavefront reconstruction, and calculating the required execution amount of the conjugate wavefront, and then sending it to the first deformable mirror 2-13 and the second deformable mirror 3-13 to perform corresponding displacement compensation, thereby compensating the laser wavefront of the radar receiving part and achieving the purpose of anti-turbulence.

[0054] Among them, the avalanche photodiode is APD (Avalanche Photo Diode), which is a light detector with internal gain. It works under reverse bias voltage. When photons are incident on its absorption region, electron-hole pairs are generated. The high electric field region accelerates the electrons to collide with lattice atoms, triggering an avalanche multiplication effect, generating more carriers, realizing photocurrent amplification, and improving the detection sensitivity of weak light signals. It can convert weak light signals into significant electrical signals. In the present invention, the light signal after pinhole filtering is detected by the first avalanche photodiode 2-12 and the second avalanche photodiode 3-12 as a time series. This time series data is of great significance to the subsequent synthetic aperture processing.

[0055] The digital micromirror device (DMD) is a light modulation element based on a microelectromechanical system (MEMS). Its core is an array of numerous tiny mirrors with independently controllable tilt angles. By controlling the mirrors to switch between the "on" and "off" states or deflect at different angles, the direction, intensity and phase of the reflected light can be precisely modulated to achieve digital control of the spatial distribution and intensity of the optical signal. In the present invention, DMD plays a key role. Atmospheric turbulence causes the laser wavefront to be distorted. The DMD receives the scattered wavefront and implements N-times compression encoding, so that the wavefront sensor 6-2 can capture multiple wavefront information in a single exposure.

[0056] Charge-Coupled Device (CCD) is a semiconductor device used to convert optical images into digital signals. Its core is a pixel matrix composed of a series of closely arranged photosensitive elements. When light shines on the photosensitive element, photons excite the generation of electron-hole pairs, and the electrons are collected and stored in the potential well corresponding to the pixel, forming a charge packet proportional to the light intensity. Driven by a clock pulse, these charge packets are transferred in a specific order to achieve the output from a two-dimensional pixel array to a one-dimensional serial electrical signal. This electrical signal is converted into digital image data through subsequent processing processes such as amplification, sampling, quantization and encoding. In the present invention, two CCDs undertake key tasks. In the radar detection process, after the outgoing laser irradiates the target, the reflected light is processed and mixed with the local oscillator laser, and part of the mixed light is received by the CCD.

[0057] Preferably, Figure 3 Where B is the center distance between the master and slave transmitting antennas, d is the center distance between the master and slave receiving antennas, and s is the center distance between the transmitting antenna and the receiving antenna.

[0058] In this embodiment, the radar transmission process is as follows: Figure 4-6 As shown, Figure 4 of The angle between the line connecting the main transmitting antenna and the target and the vertical direction, The angle between the line connecting the transmitting antenna and the target and the vertical direction, is the distance between the main transmitting antenna and the target, is the distance from the transmitting antenna to the target, H is the height of the radar from the ground, h is the height of the target, The coordinate of the target height on the z-axis. Figure 5 The coordinate rotation process is as follows: with the transmitting antenna as the center, the direction of the line connecting the transmitting antenna center and the target as the new z-axis, and the direction perpendicular to the new z-axis pointing to the ground as the new x-axis, the new coordinate system x'oz' is translated and rotated relative to the original xoz coordinate system. The translation vector is [B / 2,0,H], and the rotation angle is . Figure 6 of is the radar system sampling interval, represents the nth sampling sequence, The slant range from the main receiving antenna to the target.

[0059] Assume that two coherent beams are emitted and interfere with each other at the target point. It is always in the plane of the transmitting antenna. Assuming that MT (master transmitter) and ST (slave transmitter) both transmit Gaussian beams, and ignoring constant terms and unimportant amplitude terms, we can get:

[0060]

[0061] x, y, z are the spatial coordinates of the light beam. The first exp term is the phase delay caused by the propagation of the light beam, and the second exp term is the quadratic phase caused by the Gaussian beam.

[0062] Assume the speed of the aircraft is , the step time interval is , the step number of the azimuth is n, then At time , the coordinates of the transmitting antenna center are The diffraction distance of the Gaussian beam is actually the distance from the transmitting antenna to the target, which is related to the side deviation angle θ. When the target is far enough, .

[0063] The coordinates of the transmitting antenna center are , the center coordinates of MT are , the center coordinates of ST are .

[0064] Assuming that the elevation information of the target is h, the coordinates of the upper end of the target point can be expressed as:

[0065] , the light field of Gaussian light emitted by ST reaching the target point is:

[0066]

[0067] The light field of the Gaussian light emitted by MT reaching the target point is:

[0068]

[0069] The interference field formed at the target point is:

[0070]

[0071] The light intensity formed by the target point, as a point target (point light source), is backscattered back to the receiving antenna, and MR receives the light intensity of the interference fringes. , the light field from the scattering point target to the receiving surface can be expressed as:

[0072]

[0073] is the coordinate of the plane where the scattering point is located, and x, y are the coordinates of the receiving surface.

[0074] During the reception process, the slant distance from the target point to the MR can be accurately measured. , is the diffraction distance of the transmitting radar to the target, so the amplitude (modulus of the complex amplitude) received by the receiving antenna MR is actually the amplitude of the target after a certain distance. The attenuation value of the quartz crystal is converted into the constant term C, and we get:

[0075]

[0076] is the reflectivity of the target, is the radar wavelength, is the slant distance from the target to the MR. The optical path difference can be obtained by inversely solving the above formula: , so that it can be accurately Obtain , that is, through the information of the transmitter we can obtain the elevation information of the target.

[0077] The receiving apertures of the radar platform are deployed along the y direction, with an interval of d between the receiving apertures. Two or more apertures continuously scan the same area, and by comparing the measurement data at different times, tiny target area change information can be extracted. The processing flow includes the following steps:

[0078] S1. Data preprocessing:

[0079] Denoising and filtering: Denoise the laser echo signal (such as wavelet transform, median filtering, Gaussian filtering), filter out environmental noise and interference signals, and improve signal quality.

[0080] Among them, the wavelet transform denoising steps are:

[0081] 1. Select the wavelet basis and decomposition level: Select a mother wavelet function that is suitable for the signal characteristics (such as Daubechies wavelet, Symlets wavelet, etc.). Determine the number of decomposition levels, usually based on the frequency characteristics of the signal and the detail requirements.

[0082] 2. Wavelet decomposition: Perform multi-layer wavelet decomposition on the original laser echo signal or image to decompose it into sub-bands of different frequencies (low frequency and high frequency).

[0083] 3. Threshold processing: Apply thresholds to the wavelet coefficients of the high-frequency subband to remove noise. Common threshold methods include:

[0084] Hard Thresholding: Set the coefficients whose absolute values ​​are below the threshold to zero, and keep the coefficients above the threshold unchanged.

[0085] Soft Thresholding: Set the coefficients whose absolute values ​​are below the threshold to zero, and subtract the threshold value from the coefficients above the threshold.

[0086] The choice of threshold can be based on statistical properties (such as the standard deviation of the noise) or empirical methods.

[0087] 4. Wavelet reconstruction: Use the thresholded coefficients to reconstruct the denoised signal or image through inverse wavelet transform.

[0088] 5. Post-processing (optional): Perform further smoothing or enhancement as needed to optimize the denoising effect.

[0089] Among them, the median filter denoising step

[0090] 1. Select the filter window size: Determine the size of the sliding window (such as 3×3, 5×5, etc.). The window size is selected according to the noise type and signal characteristics.

[0091] 2. Sliding window traversal: Slide a window of selected size pixel by pixel (or point by point) over the signal or image.

[0092] 3. Calculate the median: Sort all pixel values ​​in the window and take the middle value as the new value of the center pixel of the current window.

[0093] 4. Replace pixel value: Assign the calculated median value to the pixel at the center of the window to complete a filtering operation.

[0094] 5. Repeat traversal: Repeat the above process for the entire signal or image until all pixels are processed.

[0095] Among them, the Gaussian filter noise reduction steps are:

[0096] 1. Choose the size and standard deviation (σ) of the Gaussian kernel:

[0097] Determine the window size (such as 3×3, 5×5, etc.) and standard deviation σ of the Gaussian filter. The σ value determines the smoothness of the filter.

[0098] 2. Construct Gaussian kernel:

[0099] According to the selected window size and σ value, the weight of the Gaussian function at each window position is calculated to generate a Gaussian kernel matrix.

[0100] Make sure the sum of the Gaussian kernel is 1 to keep the image brightness constant.

[0101] 3. Image Convolution:

[0102] The Gaussian kernel is convolved with the original signal or image, that is, the Gaussian kernel is slid to cover every pixel of the signal and the weighted average is calculated.

[0103] For each pixel position, the sum of the products of the surrounding pixels and the Gaussian kernel weights is calculated as the new pixel value.

[0104] 4. Boundary processing:

[0105] To process pixels at the edge of a signal or image, common methods include zero padding, mirror padding, or edge replication to ensure that the filter is applied at the edge.

[0106] 5. Output denoising results:

[0107] After completing the convolution operation, a smoothed and denoised signal or image is obtained.

[0108] 6. Time synchronization: Ensure the time synchronization of laser echo data and navigation data (IMU, GPS) to prepare for subsequent trajectory correction and data alignment.

[0109] For example, timestamp alignment uses the timestamp recorded during data acquisition to match and align the laser echo data and navigation data according to the time tag. The operation steps are as follows:

[0110] 1. Extract timestamp: Extract respective timestamp information from laser echo data and navigation data.

[0111] 2. Matching time tags: Match the data points in the two data streams based on the timestamps.

[0112] 3. Interpolation processing: If the time steps of two data streams are inconsistent, use interpolation methods (such as linear interpolation and spline interpolation) to adjust the time step of one data stream to be consistent with the other.

[0113] 4. Data alignment: Align the adjusted data in chronological order to ensure that the time corresponding to each data point is consistent.

[0114] S2. Track correction and alignment:

[0115] 5. Trajectory reconstruction: Reconstruct the precise motion trajectory of the sensor based on IMU and GPS data.

[0116] Among them, for example, the Kalman filter in Kalman filter trajectory reconstruction is a recursive algorithm that is widely used in sensor data fusion and trajectory estimation. It is suitable for linear systems, but there are also extended versions (such as extended Kalman filter, EKF) for nonlinear systems. The operation steps are as follows:

[0117] 1. Define the state vector and observation vector: The state vector usually includes parameters such as position, velocity, acceleration, etc. For example, x=[x,y,z,vx,vy,vz]Tx=[x,y,z,vx,vy,vz]T. The observation vector includes the position data provided by GPS and the acceleration and angular velocity data provided by IMU.

[0118] 2. Establish a state transfer model:

[0119] Describes how a system moves from one state to the next, usually based on kinematic equations. For example, position updates can be based on velocity, and velocity updates based on acceleration.

[0120] 3. Establish observation model:

[0121] Describes how to generate observation data from the state vector. GPS position data is directly mapped to the position portion of the state vector, while IMU data is used to predict acceleration and angular velocity.

[0122] 4. Initialize the filter:

[0123] Set up the initial state estimate and covariance matrix.

[0124] 5. Recursive prediction and update:

[0125] Prediction step: predict the next state and covariance based on the state transition model.

[0126] Update step: Incorporate new observations to update the state estimate and covariance.

[0127] 6. Output track:

[0128] After many iterations, the filter outputs a smooth and accurate trajectory estimate.

[0129] 7. Data alignment: spatially align the scan data at different times or locations to eliminate errors caused by platform movement.

[0130] For example, the Iterative Closest Point (ICP) algorithm is a widely used algorithm for point cloud data alignment. It uses iterative optimization to find the best rigid transformation (rotation and translation) between two point clouds so that they overlap as much as possible. The steps are as follows:

[0131] 1. Initial alignment: Perform a preliminary alignment of the two point clouds, usually based on a rough sensor pose estimate.

[0132] 2. Closest point matching: For each point in the source point cloud, find the nearest point in the target point cloud as the corresponding point.

[0133] 3. Transformation estimation: Calculate the optimal rigid transformation (rotation and translation) that transforms the source point cloud to the target point cloud, usually achieved by the least squares method.

[0134] 4. Apply transformation: Apply the calculated transformation to the source point cloud.

[0135] 5. Convergence judgment: Check whether the transformation converges (such as the transformation amount is less than the threshold). If not, repeat steps 2-4.

[0136] S3, synthetic aperture processing:

[0137] 6. Coherent accumulation: superimpose the data from multiple scans and use synthetic aperture technology to improve spatial resolution.

[0138] For example, coherent accumulation with matched filtering is a signal processing technology that detects and enhances signals by matching with known signal templates. In coherent accumulation, matching filtering can further enhance the coherence of the target signal and suppress noise. The operation steps are as follows:

[0139] 1. Template selection: Select or generate a template signal for the matched filter based on the characteristics of the expected target (such as frequency, phase, etc.).

[0140] 2. Filter application: Apply a matched filter to each data segment to be accumulated and extract the part that matches the template signal.

[0141] 3. Coherent accumulation: The matched filtered data is coherently accumulated, usually using phase alignment and weighted superposition methods.

[0142] 4. Signal enhancement: Through matched filtering and coherent accumulation, the coherence of the target signal is significantly enhanced, and random noise and interference are suppressed.

[0143] 5. Phase correction: For coherent measurements, phase correction is performed to eliminate phase offset caused by platform motion and environmental factors.

[0144] S4. Difference calculation:

[0145] Differential matching: Compare high-resolution point clouds or image data at different times or locations point by point to calculate differential information.

[0146] Change detection: Based on differential data, detect small changes in the target area, such as displacement, deformation, terrain changes, etc.

[0147] For example, time series analysis can detect changes by analyzing differential data of multiple time phases to identify trends and anomalies that change over time. The steps are as follows:

[0148] 1. Multi-temporal data collection: Acquire point cloud or image data at multiple time points to generate multi-temporal differential interferograms.

[0149] 2. Time series modeling: Establish a time series model for each monitoring point, such as the autoregressive model (AR), moving average model (MA), autoregressive moving average model (ARMA), etc.

[0150] 3. Trend and anomaly detection: Analyze trends, periodicity, and anomalies in time series, and identify moments and areas of abnormal changes.

[0151] 4. Marking of change areas: Based on the deviation between model prediction and actual observation, mark the areas with significant changes.

[0152] 5. Post-processing: Combine spatial information to suppress noise and optimize connectivity of changed areas.

[0153] S5. Error correction and optimization:

[0154] Geometric error correction: Correct geometric errors caused by system errors, motion platform vibration, etc.

[0155] Accuracy optimization: Improve the accuracy and reliability of differential measurement and reduce error propagation through algorithm (least squares method) optimization. Least Squares Method is a classic parameter estimation method that optimizes parameter estimation by minimizing the sum of squared errors between observed values ​​and model predicted values. This is particularly important in trajectory correction, baseline estimation, and phase correction. The operation steps are:

[0156] 1. Establish observation equation: link actual observation data with theoretical model to form linear or nonlinear observation equation.

[0157] 2. Define the error function: The error function is usually the sum of the squares of the observation error and the model prediction error.

[0158] 3. Solve optimization problems: Use optimization techniques such as Gauss-Newton method and Levenberg-Marquardt algorithm to iteratively adjust model parameters to minimize the error function.

[0159] 4. Evaluate results: Analyze the optimized parameter estimates and their confidence intervals to ensure the rationality and accuracy of the results.

[0160] S6. Data fusion and imaging:

[0161] 3D point cloud generation: Fusion of differential data with synthetic aperture imaging data to generate a high-precision 3D point cloud model.

[0162] For example, point cloud generation based on elevation model (DEM-Based Point Cloud Generation). By combining digital elevation model (DEM) and DInSAR data, the terrain information provided by DEM is fused with interferometric phase data to generate high-precision three-dimensional point cloud. This method is suitable for data scenes with existing elevation references. The operation steps are as follows:

[0163] 1. DEM acquisition and preprocessing: obtain a high-resolution digital elevation model of the target area and perform preprocessing (such as smoothing and filtering).

[0164] 2. Interference pattern generation and height unwrapping: Generate interference pattern and use DEM as a reference to perform height unwrapping to restore the actual height of the object.

[0165] 3. Point cloud construction: Combine the unwrapped height information with the geographic coordinates to generate a three-dimensional point cloud.

[0166] 4. Data fusion: Fuse the generated 3D point cloud with other data sources (such as optical images and LiDAR data) to improve the details and accuracy of the point cloud.

[0167] Change monitoring image: Generate a monitoring image showing the changed area, which can intuitively show the slight changes in the target area. For example, threshold segmentation is the most basic change detection method, which distinguishes the changed area from the non-changed area by setting one or more thresholds. Based on the segmentation results, a binary change monitoring image is generated, in which the changed area is usually highlighted with a specific color or mark. The operation steps are as follows:

[0168] 1. Differential data generation: A differential interference pattern or a displacement pattern is generated through a differential matching step (S4).

[0169] 2. Set thresholds: Set one or more thresholds based on the statistical characteristics of the data (such as mean and standard deviation) to distinguish between changing and non-changing areas.

[0170] 3. Apply threshold: The differential data is compared with the threshold, and the pixels exceeding the threshold are marked as changed areas.

[0171] 4. Generate a binary image: Convert the labeling result into a binary image. The changed area is usually represented by white or red, and the non-changed area is represented by black or other colors.

[0172] 5. Post-processing: Perform morphological operations (such as dilation and erosion) to remove isolated noise points, fill small holes, and improve the coherence and accuracy of the changed areas.

[0173] In this way, D-InSAR differential synthetic aperture lidar achieves high-resolution, high-precision three-dimensional imaging and change monitoring by combining synthetic aperture technology, interferometry technology and differential measurement methods. Its implementation process involves precise integration of system design, accurate acquisition of multi-phase or multi-view data, efficient data processing and application of differential algorithms.

[0174] Embodiment 2

[0175] Based on the same concept, the present invention also proposes a control method for a differential InSAR system, comprising:

[0176] S00, the laser emission module generates and distributes the laser beam into the radar detection module;

[0177] S10, the two detection branches of the radar detection module detect the target simultaneously and collect and process the echo signals to realize the differential detection function;

[0178] S20, the adaptive optical compensation module uses deep learning technology and wavefront detection processing mechanism to obtain and compensate for laser wavefront distortion caused by atmospheric turbulence in real time.

[0179] In this embodiment, the deep learning technology adopts the EfficientSCI model, and the training process of the model is as follows:

[0180] 1) Data collection and preprocessing

[0181] Data collection: Samples are collected from a large number of wavefront sensor measured data under different atmospheric turbulence conditions, covering various turbulence scenarios and corresponding wavefront distortion information, to build a rich and diverse data set that truthfully records the wavefront changes at different times, climates, and geographical environments, providing comprehensive material for the model to learn the characteristics of turbulence.

[0182] Preprocessing: noise reduction and normalization of raw data. Use filtering algorithms to remove artifacts and fluctuations caused by measurement noise and environmental interference, reduce data feature values ​​to a specific range, improve data quality and stability, help the model focus on learning the essential characteristics of wavefront distortion, avoid misleading learning direction due to differences in data magnitude and distribution, and ensure efficient and accurate model training.

[0183] 2) Model architecture construction

[0184] The multi-layer convolutional neural network architecture is designed as the main body, and residual connections and attention mechanism optimization are integrated. The convolution layer automatically extracts local features of the image, and multi-layer stacking expands the receptive field to capture the complex spatial structure of wavefront distortion; residual connections alleviate gradient disappearance and help deep networks to train efficiently; the attention mechanism focuses on key areas and features of the image, enhances the learning weights of important features, such as highlighting the wavefront details in the turbulent strong interference area, and improves the model feature extraction accuracy and representation capabilities.

[0185] 3) Training algorithm selection and parameter setting

[0186] Training algorithm: Use stochastic gradient descent variants (such as Adam) to optimize model parameters, flexibly adjust hyperparameters such as learning rate and momentum according to data scale, computing resources, and convergence speed requirements, balance training efficiency and model accuracy, enable the model to quickly approach the optimal solution, avoid falling into local extremes, and efficiently learn the mapping relationship between wavefront distortion and correction strategy.

[0187] Parameter setting: Reasonably divide the ratio of training set, validation set, and test set (commonly 7:2:1), and use the validation set to monitor training and prevent overfitting. Determine the batch size and number of training rounds based on the characteristics of the data set and hardware performance to ensure the generalization ability and stability of the model. After multiple rounds of iterative optimization, improve the performance of processing wavefront distortion in unknown turbulent scenes.

[0188] 4) Model evaluation and optimization

[0189] Evaluation indicators: Select indicators such as mean square error (MSE) and structural similarity index (SSIM) to quantify the similarity between the model-reconstructed image and the real image. MSE measures the mean square of the pixel value error to reflect the overall deviation. SSIM evaluates the perceptual quality from multiple dimensions such as brightness, contrast, and structure. The comprehensive indicators accurately evaluate the model performance and provide guidance for optimization.

[0190] Optimization strategy: fine-tune the network architecture, parameters or data enhancement technology based on the evaluation results. For example, amplify the data to enrich the sample diversity, improve the model complexity balance fitting ability, and improve the accuracy and reliability of the model in processing wavefront distortion under complex turbulent conditions through multiple rounds of iterative optimization to meet the high-precision adaptive optical compensation requirements of the radar system.

[0191] Embodiment 3

[0192] This embodiment provides an application of the first embodiment to an aircraft, such as a civil aircraft. Figure 2 As shown in the figure, the flight direction coordinate system of the aircraft where the radar is located. The radar system works in a single pass mode (the platform where the radar is located flies over the target once). The radar layout is as follows: Figure 3 As shown, the calibration light of the adaptive optical system is located at the radar transmitting end (the main transmitter and the slave transmitter), and the wavefront sensor 6-2 and the deformable mirror are located in the radar receiving end (the main receiver and the slave transmitter).

[0193] Among them, the radar transmitting end here includes a tunable laser 1, a one-to-two fiber beam splitter 5, a first collimator 2-1 and a second collimator 3-1, a first polarizer 2-2 and a second polarizer 3-2, a first half wave plate 2-4 and a second half wave plate 3-4, a first polarization beam splitter 2-5 and a third polarization beam splitter 3-5, a first quarter wave plate 2-6 and a third quarter wave plate 3-6, a first reflector 2-8 and a second reflector 3-8.

[0194] The radar receiving end includes a first deformable mirror 2-13 and a second deformable mirror 3-13, a second polarization beam splitter 2-9 and a fourth polarization beam splitter 3-9, a first charge coupled device 2-10 and a second charge coupled device 3-10, a second aperture 2-11 and a fourth aperture 3-11, a first avalanche photodiode 2-12 and a second avalanche photodiode 3-12.

[0195] Using the point target model, the target point coordinates are ,when At this moment, the target p is located in the normal plane of the transmitting antenna (xz plane). At time , the aircraft steps forward a distance. The transmitting antenna is located in the normal plane (xz plane), and the receiving antenna is arranged along the azimuth direction.

[0196] Embodiment 4

[0197] This embodiment further provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute the process. The process includes the control method according to the second embodiment.

[0198] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0199] In general, various embodiments may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flow charts, or using some other graphical representation, it should be understood that, as non-limiting examples, the boxes, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0200] Embodiments of the present invention can be implemented by computer software, which is executable by a data processor of a mobile device, such as in a processor entity, or implemented by hardware, or implemented by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros can be stored in any device readable data storage medium, and they include program instructions for performing specific tasks. Computer program products can include one or more computer executable components configured to perform embodiments when the program is running. One or more computer executable components can be at least one software code or a part thereof. In addition, at this point, it should be noted that any box of the logic flow can represent a program step, or interconnected logic circuits, boxes and functions, or a combination of program steps and logic circuits, boxes and functions. Software can be stored in physical media such as memory chips or storage blocks implemented in processors, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and data variants thereof, CDs. Physical media are non-transient media.

[0201] Those skilled in the art should understand that the technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A differential InSAR system, characterized in that: include: A laser emission module, used for generating and distributing a laser beam; Radar detection module, used to detect targets and collect and process signals; Adaptive optical compensation module, used to correct phase distortion caused by atmospheric turbulence; The radar detection module comprises two detection branches that are symmetrical in structure and work in coordination, including a main radar (3) and a slave radar (2), wherein the two detection branches respectively detect the target simultaneously and collect and process echo signals, so as to realize a differential detection function and improve system reliability and accuracy; The two laser beams emitted by the laser emission module are collimated by a collimator, polarized by a polarizer, filtered by a pinhole, and polarized by a half-wave plate in the respective branches of the main radar (3) and the slave radar (2), and are separated into local oscillator laser and output laser by a polarization beam splitter. The local oscillator laser enters the receiving optical path through a reflector and a quarter-wave plate and its polarization state is transformed. After the output laser irradiates the target, the reflected light is corrected for wavefront phase by a deformable mirror and then mixed with the local oscillator laser. Part of the light is observed by a charge coupled device to obtain interference patterns and target alignment conditions, and the remaining part is filtered by a pinhole and then detected by an avalanche photodiode to detect time series for synthetic aperture processing; The adaptive optical compensation module acquires and compensates for the laser wavefront distortion caused by atmospheric turbulence in real time based on deep learning technology and wavefront detection processing mechanism, so as to ensure stable and accurate detection of the differential InSAR system during long-distance remote sensing.

2. A differential InSAR system according to claim 1, characterized in that: The adaptive optical compensation module comprises an adaptive optical system laser (6-1), a digital micromirror device (6-3) and a wavefront sensor (6-2); The adaptive optical system laser (6-1) emits a measuring laser when the tunable laser (1) of the laser emission module is working, and after the laser is affected by atmospheric turbulence, the scattered wavefront is sequentially received and compressed by the digital micromirror device (6-3) and received by the wavefront sensor (6-2) within one exposure time; Reconstructing continuous images from a single exposure result of the wavefront sensor (6-2) with the help of a trained deep learning network model to improve the frame rate and closed-loop bandwidth; The ideal wavefront deviation is calculated based on wavefront reconstruction, the conjugate wavefront execution amount is determined, and the first deformable mirror (2-13) and the second deformable mirror (3-13) of the radar detection module are driven to compensate for the radar receiving laser wavefront to achieve the purpose of anti-turbulence.

3. A differential InSAR system as claimed in claim 2, characterized in that: When the radar detection module operates in a single-pass mode, the calibration light of the adaptive optical compensation module is located at the radar transmitting end to detect atmospheric turbulence phase distortion in real time; the wavefront sensor (6-2) and the deformable mirror are placed in the radar receiving end, and the receiving antenna receives the target reflected echo. The wavefront detector (6-2) first measures the distorted calibration light, and the compensation coefficient obtained by the algorithm is sent to the deformable mirror to remove the distortion, and the corrected echo is received by the radar receiving end.

4. A differential InSAR system according to claim 1, characterized in that: The laser emission module includes a tunable laser, which divides the laser into two paths through a beam splitter to enter the radar detection module.

5. A differential InSAR system according to claim 1, characterized in that: The slave radar (2) comprises a first collimator (2-1), a first polarizer (2-2), a first pinhole (2-3), a first half-wave plate (2-4), a first polarization beam splitter (2-5), a first quarter-wave plate (2-6), a second quarter-wave plate (2-7), a first reflector (2-8), a second polarization beam splitter (2-9), a first charge coupled device (2-10), a second pinhole (2-11), a first avalanche photodiode (2-12), and a first deformable mirror (2-13); The main radar (3) comprises a second collimator (3-1), a second polarizer (3-2), a third small hole (3-3), a second half wave plate (3-4), a third polarization beam splitter (3-5), a third quarter wave plate (3-6), a fourth quarter wave plate (3-7), a second reflector (3-8), a fourth polarization beam splitter (3-9), a second charge coupled device (3-10), a fourth small hole (3-11), a second avalanche photodiode (3-12), and a second deformable mirror (3-13).

6. A differential InSAR system according to any one of claims 1 to 5, characterized in that: The deep learning technology based on the adaptive optical compensation module adopts the EfficientSCI deep learning network model.

7. A control method for a differential InSAR system according to any one of claims 1 to 6, characterized in that: include: S00, the laser emission module generates and distributes the laser beam into the radar detection module; S10, the two detection branches of the radar detection module detect the target simultaneously and collect and process the echo signals to realize the differential detection function; S20, the adaptive optical compensation module uses deep learning technology and wavefront detection processing mechanism to obtain and compensate for laser wavefront distortion caused by atmospheric turbulence in real time.

8. A differential InSAR system as claimed in any one of claims 1 to 6 applied to an aircraft.

9. A readable storage medium, characterized in that: The readable storage medium stores therein a computer program including a program code for controlling a process to execute the process, the process including the control method according to claim 7.

Images