A super-resolution detection method for array radar

Abstract

The present invention discloses an array radar super-resolution detection method. First, based on the preliminary detection results of the traditional real aperture array radar, a fine angle grid is divided on the distance unit where the target is located, and the echo signals of each channel at each sampling moment are adaptively modulated. After the mutual accumulation of observation results at different moments, the original length L is converted into 1 The real aperture array is expanded to a length of L 2 The invention further utilizes the prior information of the detection model to update and correct the observation vector and the model, thereby improving the robustness of target information reconstruction. Compared with the prior art, the invention realizes angular super-resolution detection on the basis of the original radar array by constructing an equivalent aperture that is much larger than the actual aperture size.

Description

Technical Field

[0001] The present invention relates to the field of radar technology, and in particular to an array radar super-resolution detection method. Background Art

[0002] Radar detection technology is widely used in many military and civilian fields. With the increasing diversification of detection scenarios and the continuous growth of application needs, radar detection technology also needs to be improved accordingly.

[0003] Range resolution and angle resolution are two important indicators for measuring radar detection performance, which determine the radar's ability to distinguish targets in space. Among them, range resolution is the minimum distance at which the radar can distinguish two targets in the distance dimension, which is mainly determined by the radar's working bandwidth. Angle resolution is the minimum angle at which the radar can distinguish two targets in the angle dimension, which is mainly determined by the radar's aperture size at a given working frequency. With the rapid development of military technologies such as electronic information and weapons and equipment, the attack mode faced by modern battlefields has expanded from single offensive and defensive operations to multi-to-multi complex scenarios and confrontation and game of targets. At present, common complex targets on the battlefield mainly include clustered cooperative targets such as drone clusters and cruise missile clusters. They are small in size and highly maneuverable. After forming a cluster, they have the ability to cooperate in combat, and their power is doubled. At the same time, the continuous development of technologies such as artificial intelligence, satellite communications, and 5G has further enhanced the anti-destruction ability of cluster operations, enabling them to implement large-scale distributed multi-point and multi-directional assaults, which poses a great challenge to traditional defense systems. In order to seize the initiative in battlefield defense and confrontation, it is necessary to have the ability to accurately detect and track large-scale cluster targets of the enemy. At present, advanced radar defense systems for complex target detection in actual scenarios at home and abroad have the ability to detect the target as a whole, but they still face huge challenges in distinguishing its internal features and details. Due to the small velocity difference of scattering points inside complex targets, it is usually difficult to distinguish multiple scattering information in the Doppler dimension. Resolving the internal information of complex targets in the spatial dimension requires extremely high angular resolution. The angular resolution of traditional radar systems is limited by the actual aperture size of the radar array, which is far from meeting the high-resolution detection requirements for complex targets and cannot effectively reconstruct the interior of complex targets. Therefore, in order to achieve the resolution of the interior of complex targets, a large number of transceiver array elements are still required to form a sufficiently large aperture, which poses severe challenges in terms of system complexity, volume, power consumption, etc. On the other hand, for limited space combat platforms such as airborne, shipborne and missile-borne, it is unrealistic to rely on expanding the array size to improve the angular resolution detection capability. In this context, it is very urgent to study new radar detection methods for high-resolution detection of complex targets, which has extremely important application needs. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide an array radar super-resolution detection method, which realizes angular super-resolution detection on the basis of the original radar array by constructing an equivalent aperture that is far larger than the actual aperture size.

[0005] The present invention specifically adopts the following technical solutions to solve the above technical problems:

[0006] An array radar super-resolution detection method comprises the following steps:

[0007] Step 1: Based on the array radar with an actual aperture size of L1, the traditional array beamforming method is used to obtain the overall distance and angle information of the target; according to the expected angular resolution ρ a2 With the corresponding aperture size L2, each distance resolution unit where the target distribution area is located is divided into K fine angle grids of the same size, where L2>L1 and K is an integer greater than 1; the relationship between the echo received by the actual aperture of the array and the detection area can be expressed as:

[0008]

[0009] Among them, y=[y(1,t),y(2,t),…,y(Q,t)] T is the echo signal vector received by the actual aperture of the array, y(q,t) is the echo signal received by the qth element of the real aperture array at time t, q = 1, 2, ..., Q, Q is the number of elements of the array radar, St q (r k ,t) is the detection signal of the qth array element of the real aperture array to the kth grid, β=[β1,β2,…,β K ] T is the target scattering coefficient vector, β k represents the kth grid center r k Scattering coefficient at , k = 1, 2, …, K;

[0010] Sample the echo of each channel, t=[t1,t2,…,t T ], thus obtaining the preliminary echo signal vector At the same time, the observation vector Then we get the fine detection model y = S·β;

[0011] Step 2: According to the finely divided angle grid and the spatial position relationship between the set apertures L2 and L1, a modulation interval Φ is set, and T groups of mutually independent white noise are randomly generated within the modulation interval Φ according to a uniformly distributed probability criterion to achieve phase modulation. The modulation coefficient matrix composed of the generated multiple groups of modulation coefficients is used to process the echo signal vector y and the observation vector S of the Q channels of the real aperture array in parallel to obtain a new fine detection model y′=S′·β; wherein,

[0012]

[0013] Step 3: Solve the new fine detection model y′=S′·β to obtain the target scattering coefficient vector β, thereby completing the super-resolution reconstruction of the angle information in the target area.

[0014] Furthermore, the array radar super-resolution detection method further includes:

[0015] Step 4. First, update the target scattering coefficient vector β obtained in step 3 according to the following method: according to the actual aperture of the array, use the traditional array radar detection method to detect the fine angle grid, find the pixel units in the defined grid whose energy is less than the preset energy threshold value, set the grid where these pixels are located as the non-target area, and assign its scattering coefficient to 0; then, substitute the updated target scattering coefficient vector β' as prior information into the new fine detection model, so as to achieve the optimization update of the new fine detection model, and then obtain a low-noise super-resolution reconstruction result.

[0016] Further preferably, the preset energy threshold value is taken as the average energy value of the fine angle grid detection results according to the traditional array radar detection method.

[0017] Preferably, the solving is performed using a direct space matched filtering method.

[0018] Preferably, the modulation interval Φ is set based on an improved Gerchberg-Saxton algorithm, specifically comprising the following steps:

[0019] a) Initialize a set of ideal scattering points with the number K; initialize a small-size real aperture array with the aperture size L1 and the number of channels Q; initialize a two-dimensional random phase matrix Initialize a large-size real aperture array with an aperture size of L2 and a number of channels of Q, and generate the target information reconstructed by the matching method under the conditions of the array. Initialize convergence error Δ, maximum number of cycles i max ;

[0020] b) Using a two-dimensional random phase matrix The echo and observation matrix of the small-size real aperture array are phase modulated to obtain a new detection model

[0021] c) Reconstruct the target information through matching method based on the new detection model Preserve Phase Replace the magnitude with the magnitude of the reference information |β ref |, synthesize new target information

[0022] d) for β i Perform inverse solution transformation to obtain the corresponding echo information y i , retaining its phase angle(y i ), calculate the additional phase matrix of this cycle, and update the operation function for the next iteration;

[0023] e) Calculate the root mean square error between the target reconstruction information and the reference value in the current cycle. If the preset iteration stop condition is met, the algorithm stops, outputs the phase distribution that meets the condition and sets the interval to which it belongs as the modulation interval Φ; otherwise, enter step b) and start the next iteration.

[0024] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0025] The present invention performs multiple groups of random modulations that obey uniform distribution on multi-channel echoes, which is equivalent to multiple groups of large-size observation apertures at different positions. The aperture size is further expanded on the basis of the actual aperture of the original array, the formed detection beam is sharpened, and the redundancy of target information in radar echoes is avoided by utilizing the incoherence between different groups of modulations. The target information is quickly reconstructed through the solution method of matched filtering to achieve angle super-resolution detection.

[0026] The present invention further utilizes the angle detection results of the real aperture array with low resolution and high number of pixels to obtain prior information of the detection scene and the target, optimizes and corrects the observation vector and the modulation coefficient, and improves the robustness of target information reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of the principle of the array radar super-resolution detection method of the present invention;

[0028] Figure 2 The detection results of two corner reflectors by the traditional real aperture array detection method;

[0029] Figure 3 This is the detection result of two corner reflectors by the super-resolution detection method of the present invention. DETAILED DESCRIPTION

[0030] In view of the shortcomings of the existing technology, the solution of the present invention is to perform fine angular grid division on the distance unit where the target is located based on the preliminary detection results of the traditional real aperture array radar, and adaptively modulate the echo signals of each channel at each sampling moment. After mutual accumulation of observation results at different moments, the original real aperture array with a length of L1 is expanded to a large aperture array with a length of L2, thereby realizing angular super-resolution detection.

[0031] The present invention specifically adopts the following technical solutions to solve the above technical problems:

[0032] An array radar super-resolution detection method comprises the following steps:

[0033] Step 1: Based on the array radar with an actual aperture size of L1, the traditional array beamforming method is used to obtain the overall distance and angle information of the target; according to the expected angular resolution ρ a2 With the corresponding aperture size L2, each distance resolution unit where the target distribution area is located is divided into K fine angle grids of the same size, where L2>L1 and K is an integer greater than 1; the relationship between the echo received by the actual aperture of the array and the detection area can be expressed as:

[0034]

[0035] Among them, y=[y(1,t),y(2,t),…,y(Q,t)] T is the echo signal vector received by the actual aperture of the array, y(q,t) is the echo signal received by the qth element of the real aperture array at time t, q = 1, 2, ..., Q, Q is the number of elements of the array radar, St q (r k ,t) is the detection signal of the qth array element of the real aperture array to the kth grid, β=[β1,β2,…,β K ] T is the target scattering coefficient vector, β k represents the kth grid center r k Scattering coefficient at , k = 1, 2, …, K;

[0036] Sample the echo of each channel, t=[t1,t2,…,t T ], thus obtaining the preliminary echo signal vector At the same time, the observation vector Then we get the fine detection model y = S·β;

[0037] Step 2: According to the finely divided angle grid and the spatial position relationship between the set apertures L2 and L1, a modulation interval Φ is set, and T groups of mutually independent white noise are randomly generated within the modulation interval Φ according to a uniformly distributed probability criterion to achieve phase modulation. The modulation coefficient matrix composed of the generated multiple groups of modulation coefficients is used to process the echo signal vector y and the observation vector S of the Q channels of the real aperture array in parallel to obtain a new fine detection model y′=S′·β; wherein,

[0038]

[0039] Step 3: Solve the new fine detection model y′=S′·β to obtain the target scattering coefficient vector β, thereby completing the super-resolution reconstruction of the angle information in the target area.

[0040] The present invention can further utilize the prior information and the expanded aperture for joint reconstruction to suppress the influence of noise on the reconstruction quality. Specifically, the array radar super-resolution detection method further includes:

[0041] Step 4. First, update the target scattering coefficient vector β obtained in step 3 according to the following method: according to the actual aperture of the array, use the traditional array radar detection method to detect the fine angle grid, find the pixel units in the defined grid whose energy is less than the preset energy threshold value, set the grid where these pixels are located as the non-target area, and assign its scattering coefficient to 0; then, substitute the updated target scattering coefficient vector β' as prior information into the new fine detection model, so as to achieve the optimization update of the new fine detection model, and then obtain a low-noise super-resolution reconstruction result.

[0042] The modulation interval Φ can be set based on experience or based on the Gerchberg-Saxton algorithm. The Gerchberg-Saxton algorithm is one of the iterative algorithms for Fourier phase retrieval. Its basic idea is to perform repeated iterative Fourier transforms between the object plane and the spectrum plane, and set two amplitude constraint restrictions on the object plane and the spectrum plane, so as to maximize the recovery of the phase distribution on the object plane. However, the traditional Gerchberg-Saxton algorithm only uses a single Fourier transform and an inverse Fourier transform as the operation function in the iterative process, and it is fixed in different iterations. The effect is not ideal for the complex model relationship between the observation matrix, additional phase, target information, and scattered echoes in the present invention. Therefore, the present invention is based on the traditional Gerchberg-Saxton algorithm phase retrieval iterative process. The forward and inverse process of reconstructing the target information using the matching method is used as the operation function, replacing the original Fourier transform and inverse Fourier transform, and the additional phase matrix calculated in each iteration is used to update the operation function for the next iteration.

[0043] The improved Gerchberg-Saxton algorithm proposed in the present invention specifically includes the following steps:

[0044] a) Initialize a set of ideal scattering points with the number K; initialize a small-size real aperture array with the aperture size L1 and the number of channels Q; initialize a two-dimensional random phase matrix Initialize a large-size real aperture array with an aperture size of L2 and a number of channels of Q, and generate the target information reconstructed by the matching method under the conditions of the array. Initialize convergence error Δ, maximum number of cycles i max ;

[0045] b) Using a two-dimensional random phase matrix The echo and observation matrix of the small-size real aperture array are phase modulated to obtain a new detection model

[0046] c) Reconstruct the target information through matching method based on the new detection model Preserve Phase Replace the magnitude with the magnitude of the reference information |β ref |, synthesize new target information

[0047] d) for β i Perform inverse solution transformation to obtain the corresponding echo information y i , retaining its phase angle(y i ), calculate the additional phase matrix of this cycle, and update the operation function for the next iteration;

[0048] e) Calculate the root mean square error between the target reconstruction information and the reference value in the current cycle. If the preset iteration stop condition is met, the algorithm stops, outputs the phase distribution that meets the condition and sets the interval to which it belongs as the modulation interval Φ; otherwise, enter step b) and start the next iteration.

[0049] In order to facilitate public understanding, the technical solution of the present invention is further described in detail below with reference to the accompanying drawings:

[0050] S1. First, build a fine detection model based on the actual aperture of the traditional array radar:

[0051] Based on the array radar with an actual aperture size of L1, the traditional beamforming method is first used to reconstruct a low-resolution image of the detection area. This obtains the overall distance and angle information of the complex target and delineates the approximate distribution area of ​​the target in the imaging scene (the expanded observation aperture will perform angle super-resolution detection for this area). Furthermore, the expected angular resolution ρ is set according to the requirements.a2 , and based on this, the size of the expanded observation aperture is solved

[0052] L2=λ / ρ a2

[0053] where λ = c / f c is the wavelength of the detection signal. Each distance resolution unit in the area where the complex target is located is re-divided into K fine angle grids of the same size, and the size of each resolution unit along the angle direction is equal to the set angle resolution ρ a2 , and assume that the corresponding target scattering coefficient vector is

[0054] β=[β1,β2,…,β K ] T

[0055] where β k represents the kth grid center r k The scattering coefficient at the location; then the relationship between the echo received by the actual aperture of the array and the detection area can be expressed as

[0056]

[0057] Among themSt q (r k ,t) is the detection signal of the qth array element of the real aperture array to the kth grid. Sample the echo of each channel, t=[t1,t2,…,t T ], thus obtaining the preliminary echo signal vector At the same time, the observation vector Then the detection model y=S·β is obtained.

[0058] The parameters can be determined according to the specific imaging system. In this embodiment, the array aperture size L1 is selected to be 5.7 mm, the number of array elements Q is 4, the detection signal is a linear frequency modulation signal with a bandwidth of 4 GHz (77-81 GHz), and the echo is de-skewed at the receiving end. The expected angular resolution ρ a2 =0.018rad.

[0059] S2. Based on the improved Gerchberg-Saxton algorithm, the modulation interval is solved according to the aperture position relationship. Specifically:

[0060] Initialize a set of K ideal scattering points located at the center of the defined fine angle grid, and their scattering coefficients are β1=β2=,…,=β K =1;

[0061] Initialize a small-size real aperture array with an aperture size of L1 and a channel number of Q. After detecting the set ideal scattering point, the de-slanted target reflection echo received is Observation vector

[0062] Initialize a 2D random phase matrix

[0063] Initialize a large-size real aperture array with an aperture size of L2 and a channel number of Q. After detecting the set ideal scattering point, the de-slanted target reflection echo received is Observation vector Based on the detection model in S1, the target information is reconstructed by matching method:

[0064] Initialize convergence error Δ, maximum number of cycles i max ;

[0065] The iterative process of the improved Gerchberg-Saxton algorithm is divided into the following six steps:

[0066] a) Using Phase Matrix The echo and observation matrix of the small aperture array are phase modulated to obtain a new detection model:

[0067]

[0068] b) Reconstruct the target information through matching method based on the newly created detection model:

[0069]

[0070] c) Amplitude constraint, retaining the phase of the target reconstruction information Replace its magnitude with the magnitude of the reference information |β ref |, get the synthesized target information

[0071]

[0072] In the formula, |.| is the value operator, angle(.) is the phase operator, β i is the synthesis result in the i-th iteration process.

[0073] d) for β i Perform inverse solution transformation to obtain the corresponding echo information y i

[0074]

[0075] e) The phase angle(y) of the target reconstruction information is preservedi ), calculate the additional phase matrix of this cycle

[0076]

[0077] Based on this order And update the operation functions in steps b) and d) for the next iteration respectively

[0078] f) Calculate the root mean square error between the target reconstruction information and the reference value in the i-th cycle. If RMSE≤Δ or i>i max , that is, when the algorithm meets the error setting conditions or exceeds the set maximum number of cycles, the algorithm stops and outputs the phase matrix that meets the conditions Setting the Phase Matrix The corresponding interval Φ is the modulation interval. If the set condition is not met, the process goes to step a) and starts the next iteration.

[0079] S3, adaptive modulation of echo and super-resolution reconstruction of target information;

[0080] In the process of super-resolution reconstruction of target information, the significance of the modulation interval Φ is that when the aperture size is expanded from L1 to L2, it needs to be loaded into the effective range of random modulation of the multi-channel signal. In the process of array radar super-resolution detection, the modulation of a single sample can only provide a set of sparse aperture observation results, which cannot meet the conditions for better angular resolution of the defined detection grid. Therefore, the present invention adopts a method of multiple parallel modulation of L1 multi-channel echoes to form multiple groups of observation apertures at random positions in L2, which is equivalent to the process of the radar transmitting and receiving signals multiple times in different groups of apertures. Specifically, for the super-resolution reconstruction scenario, the modulation interval Φ is first adaptively solved based on the improved Gerchberg-Saxton algorithm of S2 according to the finely divided angle grid and the set aperture. In order to ensure that the contributions of the observation aperture positions between and within different groups are the same and avoid redundancy, the present invention randomly generates T groups of independent white noises based on a uniformly distributed probability criterion to achieve phase modulation.

[0081]

[0082]

[0083] In this way, the Q channel data of the real aperture array (the distance movement and Doppler broadening compensation between channels have been completed) are processed to obtain a new echo signal vector

[0084]

[0085] Similarly, new observation vectors corresponding to different samples can be obtained The equivalent observation process is that Q antennas receive radar echo signals at random observation positions within an aperture of length L2.

[0086] To reconstruct the target information, the echo data can be processed using imaging methods such as back projection, power spectrum estimation, and regularization. However, different imaging methods have different effects on different application scenarios. Therefore, after a large number of experimental verifications, the present invention prefers the direct matched filtering solution method, which is a linear process, simple to implement and can obtain stable results. It can improve the efficiency of super-resolution reconstruction of angle information in the target area while ensuring the stability of the system. The specific solution process can be described as:

[0087]

[0088] As the number of observation samples increases, the accumulated information can be equivalent to the detection results of a uniform linear array of length L2. By coherently accumulating multiple sets of large-aperture target inversion results, the complete scattering information of the fine grid can be mapped to obtain the super-resolution radar detection results, that is,

[0089]

[0090] S4, optimize the detection model to suppress the influence of noise;

[0091] First, based on the small aperture L1, the traditional beamforming method is used to reconstruct the information of the fine angle grid redefined in S1 and produce a low-resolution and high-pixel reconstruction result. The threshold is set according to the average energy of this result. The grid cells whose scattering coefficient is greater than this threshold are judged as the target area, and its value β is maintained. k The scattering coefficient is less than the threshold σ T The grid cell is judged as the non-target area ∧, and its corresponding scattering coefficient is determined as:

[0092] β k =0,k∈Λ

[0093] That is β ∧ =0.

[0094] The updated target scattering coefficient vector β' is used as prior information to substitute into the fine detection model construction process in S1, and the real aperture observation vector S is updated. Similarly, β' is substituted into the initialization process in S2 to construct a new real aperture array detection model, and the observation vectors S1, S2 and echoes y1, y2 are updated to optimize the adaptive modulation coefficient matrix, realize the update of the super-resolution detection model, and then produce high-quality radar detection results with low noise interference.

[0095] In order to verify the technical effect of the present invention, the method of the present invention is used to detect and reconstruct the image of two corner reflectors with a size of 1cm×1cm×1cm, an angular interval of 1°, and a distance of 1.2m from the radar, and the detection results are compared with those of the traditional real aperture array detection method. Figure 2 and Figure 3 The two-dimensional imaging results of the traditional real aperture array beamforming method and the array radar super-resolution detection method proposed in this invention are given respectively. Figure 2 As shown in the figure, the detection method of the traditional array is limited by the actual aperture size, the angular resolution is low, and the target information is redundantly overlapped and cannot be distinguished. Figure 3 As shown in the figure, after adopting the array radar super-resolution detection method proposed in the present invention, the angular resolution capability is greatly improved, the two targets are well separated and the reconstructed position information is accurate, the focusing performance is good, and it is not affected by interference in the actual measurement environment. Compared with the traditional real aperture array detection method, the present invention achieves nearly 30 times the super-resolution capability in the angular direction.

[0096] It can be seen that compared with traditional real aperture array radar detection, the present invention performs multiple groups of random modulations that obey uniform distribution on multi-channel echoes, which is equivalent to multiple groups of observation apertures at different positions, sharpens the formed detection beam, and utilizes the incoherence between different groups of modulations to avoid the redundancy of target information in radar echoes; and further combines the set threshold to obtain prior information of the detection model, and quickly reconstructs the target information through the matched filtering solution method to achieve noise-resistant angle super-resolution detection.

Claims

1. A super-resolution detection method for array radar, characterized in that: The following steps are involved: Step 1: Based on the array radar with an actual aperture size of L1, the traditional array beamforming method is used to obtain the overall distance and angle information of the target; according to the expected angle resolution and the corresponding aperture size L2, each distance resolution unit where the target distribution area is located is divided into K fine angle grids of the same size, where L2>L1 and K is an integer greater than 1; the relationship between the echo received by the actual aperture of the array and the detection area can be expressed as: Among them, y=[y(1,t),y(2,t),…,y(Q,t)] T is the echo signal vector received by the actual aperture of the array, y(q,t) is the echo signal received by the qth element of the real aperture array at time t, q = 1, 2, ..., Q, Q is the number of elements of the array radar, St q (r k ,t) is the detection signal of the qth array element of the real aperture array to the kth grid, β=[β1,β2,…,β K ] T is the target scattering coefficient vector, β k represents the kth grid center r k Scattering coefficient at , k = 1, 2, …, K; Sample the echo of each channel, t=[t1,t2,…,t T ], thus obtaining the preliminary echo signal vector At the same time, the observation vector Then we get the fine detection model y = S·β; Step 2: According to the finely divided angle grid and the spatial position relationship between the set apertures L2 and L1, a modulation interval Φ is set, and T groups of mutually independent white noise are randomly generated within the modulation interval Φ according to a uniformly distributed probability criterion to achieve phase modulation. The modulation coefficient matrix composed of the generated multiple groups of modulation coefficients is used to process the echo signal vector y and the observation vector S of the Q channels of the real aperture array in parallel to obtain a new fine detection model y′=S′·β; wherein, Step 3: Solve the new fine detection model y′=S′·β to obtain the target scattering coefficient vector β, thereby completing the super-resolution reconstruction of the angle information in the target area.

2. The array radar super-resolution detection method according to claim 1, characterized in that: Also includes: Step 4. First, update the target scattering coefficient vector β obtained in step 3 according to the following method: according to the actual aperture of the array, use the traditional array radar detection method to detect the fine angle grid, find the pixel units in the defined grid whose energy is less than the preset energy threshold value, set the grid where these pixels are located as the non-target area, and assign its scattering coefficient to 0; then, substitute the updated target scattering coefficient vector β' as prior information into the new fine detection model, so as to achieve the optimization update of the new fine detection model, and then obtain a low-noise super-resolution reconstruction result.

3. The array radar super-resolution detection method as claimed in claim 2, characterized in that: The preset energy threshold value is the average energy value of the fine angle grid detection result according to the traditional array radar detection method.

4. The array radar super-resolution detection method according to claim 1, characterized in that: The solution is performed using a direct space matched filtering method.

5. The array radar super-resolution detection method according to claim 1, characterized in that: The modulation interval Φ is set based on an improved Gerchberg-Saxton algorithm, and specifically includes the following steps: a) Initialize a set of ideal scattering points with the number K; initialize a small-size real aperture array with the aperture size L1 and the number of channels Q; initialize a two-dimensional random phase matrix Initialize a large-size real aperture array with an aperture size of L2 and a number of channels of Q, and generate the target information reconstructed by the matching method under the conditions of the array. y2 and S2 are the de-slanted target reflection echo and observation vector received by the large-size real aperture array after detecting the set ideal scattering point; the initialization convergence error Δ, the maximum number of cycles i max ; b) Using a two-dimensional random phase matrix The echo and observation matrix of the small-size real aperture array are phase modulated to obtain a new detection model y1 and S1 are respectively the de-slanted target reflection echo and observation vector received by the small-size real aperture array after detecting the set ideal scattering point; c) Reconstruct the target information through matching method based on the new detection model Preserve Phase Replace the magnitude with the magnitude of the reference information |β ref |, synthesize new target information d) for β i Perform inverse solution transformation to obtain the corresponding echo information y i , retaining its phase angle(y i ), calculate the additional phase matrix of this cycle, and update the operation function for the next iteration; e) Calculate the root mean square error between the target reconstruction information and the reference value in the current cycle. If the preset iteration stop condition is met, the algorithm stops, outputs the phase distribution that meets the condition and sets the interval to which it belongs as the modulation interval Φ; otherwise, enter step b) and start the next iteration.

Images