Millimeter wave imaging sparse sampling method and device, storage medium and processor

By constructing the sparse sampling minimization objective function and using binary differential evolution algorithm optimization, the problems of low intensive sampling efficiency and fixed guilt rate in millimeter wave imaging technology are solved, and an efficient and low aliased rare guilt sampling scheme is realized, suitable for humans. security inspection and industrial inspection and other fields.

CN118897330APending Publication Date: 2024-11-05SOUTHWEST JIAOTONG UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410800477.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-20
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

When the existing millimeter wave imaging technology meets high-quality imaging effects, dense full sampling results in a significant increase in data acquisition volume and acquisition time. The sparse sampling method has aliasing problems and fixed limitations of sparse rate, which is difficult to ensure the imaging quality. Improve data acquisition efficiency.

Method used

The sparse sampling design method based on intelligent optimization is adopted to construct the sparse sampling minimization objective function and optimize the solution using a binary differential evolution algorithm with no mutated parameters to optimize the millimeter wave imaging quality under sparse sampling conditions.

Benefits of technology

显著降低了点扩散函数旁瓣幅值,减少了成像结果中的混叠现象,提高了毫米波图像的重建质量,提供了灵活的稀疚率选择,减少了数据采集量和时间,提高了系统 real-time and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118897330A_ABST
    Figure CN118897330A_ABST
Patent Text Reader

Abstract

The invention discloses a millimeter wave imaging sparse sampling method and device, a storage medium and a processor. The method comprises the following steps: constructing a sparse sampling minimization objective function; rearranging the sparse sampling position matrix into a column vector form as an individual in a differential evolution population; after the individuals in the population are initialized, iterative solution is carried out on the sparse sampling minimization objective function, variation, crossover and selection operations are carried out in sequence in each iteration, and the individuals are updated according to the operations; when the number of iterations reaches a preset value, arranging optimal individuals in the current population in a matrix form to obtain an optimal sampling matrix; and acquiring echo data of the target object according to the optimal sampling matrix, and completing millimeter wave image reconstruction of the target object. According to the method, the sparse sampling minimization objective function is constructed, and the binary differential evolution algorithm without variation parameters is designed for optimization solution, so that the sidelobe amplitude of the point spread function is effectively reduced, the aliasing phenomenon in the imaging result is reduced, and the reconstruction quality of the millimeter wave image is remarkably improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of millimeter wave imaging technology, and in particular to a millimeter wave imaging sparse sampling method, device, storage medium and processor. Background Art

[0002] Millimeter Wave (MMW) imaging technology is widely used in human security inspection, industrial inspection, military reconnaissance and other fields due to its unique penetrating ability and high resolution. Compared with X-ray imaging technology, millimeter wave imaging is non-ionizing and will not cause ionizing radiation damage to the human body, so it is more suitable for human security inspection applications. Millimeter waves can penetrate media such as clothing, textiles and wrapping paper, provide high-resolution images, and detect metal and non-metal contraband hidden under clothing, such as plastic pistols, liquid explosives, etc.

[0003] At present, active millimeter wave security inspection imaging systems mainly use holographic imaging technology. In order to obtain high-quality imaging effects, dense full sampling is usually required to meet the Nyquist sampling theorem. However, dense full sampling will lead to a significant increase in data collection volume and collection time, which poses a challenge to real-time performance and efficiency in practical applications. To solve this problem, sparse sampling methods are considered to be an effective solution.

[0004] Traditional sparse sampling methods such as random sparse sampling can reduce the amount of data and acquisition time, but the corresponding point spread function (PSF) has a high sidelobe amplitude, which leads to serious aliasing problems in the imaging results. Another sparse sampling method based on Barker code, although it overcomes the defects of random sparse sampling to a certain extent, has certain limitations in practical applications due to the limited number of Barker codes and the fixed sparse rate. The sparse sampling method based on the Archimedean spiral curve can improve the problem of uneven distribution of sampling points to a certain extent through equal-angle sampling and equal-arc length interval sampling, but its sampling scheme still has room for further optimization.

[0005] Therefore, how to design a more optimized sparse sampling method to further improve data acquisition efficiency while ensuring imaging quality is still a technical problem that needs to be solved urgently. Summary of the invention

[0006] In order to solve the technical problems existing in the above-mentioned prior art, the present invention aims to provide a sparse sampling design method for millimeter-wave imaging based on intelligent optimization. By constructing a sparse sampling minimization objective function and adopting a binary differential evolution algorithm with no mutation parameters for optimization and solution, the shortcomings of the traditional method are effectively overcome, and the quality of millimeter-wave imaging under sparse sampling conditions is significantly improved.

[0007] In order to achieve the above-mentioned invention object, the technical solution provided by the present invention includes:

[0008] The millimeter wave imaging sparse sampling method comprises the following steps:

[0009] Construct a sparse sampling minimization objective function f(a) = [λ1·N(a)+λ2·S(a)+λ3·W(a)+λ4·P(a)], where a is a sparse sampling position matrix represented in binary, N(a) is the sampling ratio, S(a) is the average value of the ratio of the area with an amplitude greater than -40 dB in the point spread function in the x-direction and y-direction excluding the main lobe area to the area outside the main lobe, W(a) is the average value of the larger value of the half-wavelength values ​​on the left and right sides of the main lobe peak of the point spread function in the x-direction and y-direction, and P(a) is the average value of the amplitude difference between the peak side lobe of the point spread function in the x-direction and the adjacent main lobe; λ1, λ2, λ3, and λ4 are weighting coefficients of various parameters respectively;

[0010] The sparse sampling position matrix a is rearranged into a column vector form as the individual X in the differential evolution population i , i = 1, 2, ... NP, NP is the population size;

[0011] After initializing the individuals in the population, the sparse sampling minimization objective function f(a) is iteratively solved, and each iteration sequentially performs mutation, crossover, and selection operations to update the individuals accordingly;

[0012] When the number of iterations reaches the preset value N max When the optimal individual X in the current population is best Arrange in matrix form to obtain the optimal sampling matrix;

[0013] The echo data of the target object is collected according to the optimal sampling matrix to complete the millimeter wave image reconstruction of the target object.

[0014] In some preferred embodiments, the method for initializing individuals in the population includes:

[0015] Initializing at least one individual in the population using a spiral equiangular sparse sampling scheme;

[0016] Initializing at least one individual in the population using a spiral equal arc length interval sparse sampling scheme;

[0017] The remaining individuals in the population are initialized randomly.

[0018] In some preferred embodiments, the random initialization method includes:

[0019] The initial population The jth gene in the i-th individual Generate it as follows:

[0020] Among them, rand i,j is a uniformly distributed random number generated between [0,1], and Z is a real number between [0,1] set according to the sparse sampling ratio.

[0021] In some preferred embodiments, the mutation operation method includes:

[0022] Randomly select three individuals from the current population and in, represents the i-th individual in the g-th generation population, and r1, r2, r3∈{1,2,…,NP}, i≠r1≠r2≠r3;

[0023] The new target individual obtained by the mutation operation without mutation parameters The j-th gene of is represented by:

[0024]

[0025] In some preferred embodiments, the crossover operation method includes:

[0026] The new target individual obtained after mutation With the parent individual Perform crossover to generate new test individuals Test individual The j-th gene of is represented by:

[0027] Where CR is the crossover probability and rand(D) is a random integer between [1,D].

[0028] In some preferred embodiments, the method of selecting an operation includes:

[0029] The new test individuals obtained by the crossover operation With the parent individual Compare and select according to the formula:

[0030]

[0031] The present invention also provides a device, which is used to implement the above method.

[0032] The present invention also provides a storage medium, the storage medium includes a stored program, wherein the program executes the above method when it is run.

[0033] The present invention also provides a processor, which is used to run a program, wherein the program executes the above method when it is run.

[0034] Beneficial Effects

[0035] 1. Improving imaging quality: The present invention constructs a sparse sampling minimization objective function and designs a binary differential evolution algorithm without variable parameters for optimization and solution, which effectively reduces the point spread function (PSF) sidelobe amplitude, reduces the aliasing phenomenon in the imaging results, and significantly improves the reconstruction quality of millimeter wave images.

[0036] 2. Flexible sparse rate selection: The present invention is solved through an optimization algorithm and is not limited to a fixed sparse rate. The sparse rate can be flexibly adjusted according to actual needs, providing greater design freedom to meet the needs of different application scenarios.

[0037] 3. Reduce the amount and time of data collection: Compared with the dense full sampling method, the sparse sampling design of the present invention greatly reduces the amount and time of data collection, improves the real-time performance and efficiency of the system, and is suitable for occasions requiring rapid detection and real-time imaging.

[0038] 4. Improve computational efficiency: By using a binary differential evolution algorithm with no mutation parameters, an efficient optimization process is achieved by initializing the population and iteratively solving the problem, which reduces computational complexity and improves the convergence speed of the algorithm.

[0039] 5. Broad application prospects: The present invention is applicable to various millimeter wave imaging systems, especially in application fields such as human body security inspection, industrial inspection and military reconnaissance that require efficient and rapid imaging, and has broad application prospects and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flowchart of the sparse sampling method for millimeter wave imaging in some preferred embodiments of the present invention;

[0041] Figure 2 Schematic diagram of imaging scene of plane scanning three-dimensional active millimeter wave imaging system in some experimental examples of the present invention;

[0042] Figure 3 Schematic diagram of metal strip imaging target in some experimental examples of the present invention;

[0043] Figure 4 Schematic diagram for comparing target reconstruction results of various comparison methods in some experimental examples of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described below in conjunction with the accompanying drawings. In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside" and the like indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.

[0045] Example

[0046] like Figure 1 As shown, this embodiment provides a millimeter wave imaging sparse sampling method, including the steps of:

[0047] S1. Construct a sparse sampling minimization objective function f(a) = [λ1·N(a)+λ2·S(a)+λ3·W(a)+λ4·P(a)], where a is a sparse sampling position matrix represented in binary, N(a) is the sampling ratio, S(a) is the average value of the ratio of the area with an amplitude greater than -40dB in the point spread function in the x- and y-directions excluding the main lobe area to the area outside the main lobe, W(a) is the average value of the larger value of the half-wavelength on the left and right sides of the main lobe peak of the point spread function in the x- and y-directions, and P(a) is the average value of the amplitude difference between the sidelobe of the point spread function peak and the adjacent main lobe in the x- and y-directions; λ1, λ2, λ3, and λ4 are the weighting coefficients of each parameter, respectively.

[0048] Those skilled in the art should know that based on the sparse sampling millimeter wave imaging system, the point target imaging result after normalization in the xy plane is obtained, and then the one-dimensional point spread function PSF in the x direction (horizontal direction) and y direction (vertical direction) of the point target is obtained. x With PSF y . The value of the normalized point spread function is expressed in dB, and the maximum value of the main lobe is normalized to 1, corresponding to 0dB. A binary matrix a composed of "0" and "1" is used to describe the sampling point positions of the antenna array scanning plane under the sparse sampling mode. The number of rows and columns of the matrix a is the same as the number of rows and columns of the sampling matrix under uniform full sampling. The sampling points corresponding to the position of the element "0" are not sampled, and the sampling points corresponding to the position of the element "1" are sampled.

[0049] The above N(a), S(a), W(a), and P(a) are all evaluation indicators of sparse sampling schemes, where N(a) represents the ratio of the number of sampling points in the sampling point set a under sparse sampling to the number of sampling points under full sampling (i.e., the sampling ratio under sparse sampling conditions). For the one-dimensional point spread function PSF d(a), where the subscript d can be x or y, and S d (a) represents the ratio of the area with amplitude greater than -40 dB outside the main lobe to the area outside the main lobe (unit: dB), S(a) is S x (a) With S y The average value of (a); W d (a) The larger value of the point spread function half a wavelength to the left and right of the main lobe peak is taken (in dB) to measure the main lobe width, W d The smaller the (a) value, the smaller the main lobe width. W(a) is W x (a) With W y The average value of (a); P d (a) represents the amplitude difference between the peak side lobe of the point spread function and the adjacent main lobe (unit: dB), P(a) is P x (a) With P y (a) The average value of

[0050] The larger the weighting coefficients λ1, λ2, λ3, and λ4 are, the more important the corresponding indicators are in the constructed minimization objective function. In some preferred embodiments, the default values ​​of the weighting coefficients λ1, λ2, λ3, and λ4 are 5.0, 5.0, 0.1, and 0.1, respectively.

[0051] Next, a binary differential evolution algorithm with no mutation parameters is designed to solve the above sparse sampling optimization problem.

[0052] It should be noted that the control parameters in the binary differential evolution algorithm without mutation parameters designed in the present invention mainly include: population size NP, maximum population evolution generation number N max And the crossover probability CR. The crossover probability CR is used to determine the gene position of the individual that undergoes crossover during the crossover operation, and its value is usually between (0,1).

[0053] S2, rearrange the sparse sampling position matrix a into a column vector form as the individual X in the differential evolution population i , i = 1, 2, ... NP, NP is the population size. It should be understood that the population consists of NP individuals, where the i-th individual is represented by X i , i = 1, 2, ... NP. Individual X i In vector form, corresponding to the binary matrix a, the encoding dimension of the individual is D, and its value is the product of the number of rows and columns of the sampling matrix corresponding to full sampling. Rearranging the binary matrix a into a column vector form gives the individual X i , otherwise individual X i Arrange in matrix form to get the corresponding a. Population evolution generation g=0,1,…,N max , where N maxis the maximum number of evolutionary generations of the population.

[0054] S3. After initializing the individuals in the population, the sparse sampling minimization objective function f(a) is iteratively solved, and each iteration performs mutation, crossover, and selection operations in turn to update the individuals accordingly.

[0055] Those skilled in the art will know that there are many methods for initializing the population. The present invention does not intend to specify a particular method, and those skilled in the art can choose it by themselves. After the population is initialized, each individual in the initial population is substituted into the objective function f(a) to calculate the initial population of each individual Fitness Then perform iterative calculations based on this.

[0056] In some preferred embodiments, the method for initializing individuals in the population includes:

[0057] Initialize at least one individual in the population using a spiral equiangular sparse sampling scheme. The spiral equiangular sparse sampling scheme is a spatial sampling strategy for sparse sampling. The scheme is based on an Archimedean spiral curve and achieves a sparse sampling effect by sampling at equiangular intervals.

[0058] Initialize at least one individual in the population using a spiral equal arc length interval sparse sampling scheme. The spiral equal arc length interval sparse sampling scheme is a sparse sampling strategy based on an Archimedean spiral curve, on which sampling points are arranged at equal arc length intervals.

[0059] The remaining individuals in the population are randomly initialized. The random initialization means that the initial population or initial solution is set by randomly generating. There are many specific implementation methods. In some preferred embodiments, a method of random initialization based on sparse sampling ratio is given, which specifically includes:

[0060] The initial population The jth gene in the i-th individual Generate it as follows:

[0061] Among them, rand i,j is a uniformly distributed random number generated between [0,1], and Z is a real number between [0,1] set according to the sparse sampling ratio.

[0062] The mutation is to randomly change some parts of the individual to increase the diversity of the population and avoid falling into the local optimal solution. For real-coded individuals, the mutation operation is usually a small random perturbation of some genes of the individual. The general operation method is as follows:

[0063] Assume that an individual is represented by a vector [x1,x2,x3,...,x n ], the mutation operation can be performed as follows:

[0064] 1. Randomly select a gene position, for example, select the i-th position.

[0065] 2. Make a small perturbation to the gene value, such as adding a random number: x i ←x i +δ, where δ is a decimal number randomly drawn from some distribution (such as the normal distribution).

[0066] 3. The individual after mutation is [x1,x2,...,x i +δ,...,x n ].

[0067] In some preferred embodiments, a mutation operation without mutation parameters is also provided, specifically including:

[0068] S301. Randomly select three individuals from the current population. and in, represents the i-th individual in the g-th generation population, and r1, r2, r3∈{1,2,…,NP}, i≠r1≠r2≠r3;

[0069] S302: New target individuals obtained through mutation operation without mutation parameters The j-th gene of is represented by:

[0070]

[0071] The crossover operation is to exchange some genes from two individuals to generate a new individual, simulating the reproduction process of organisms. For individuals coded with real numbers, commonly used crossover methods include arithmetic crossover and linear crossover. The conventional operation method is as follows:

[0072] Suppose there are two parent individuals Parent1=[x1,x2,...,x n ] and Parent2=[y1,y2,...,y n ], using arithmetic crossover:

[0073] 1. Select an intersection point, such as the jth position.

[0074] 2. Generate a random weight α∈[0,1].

[0075] 3. After crossover, two new individuals are generated:

[0076] Wherein, j is the gene position index, j = 1, 2, ..., n.

[0077] In some preferred embodiments, the following crossover operation method is provided, specifically including:

[0078] S311, the new target individual obtained after mutation With the parent individual Perform crossover to generate new test individuals Test individual The j-th gene of is represented by:

[0079] Where CR is the crossover probability and rand(D) is a random integer between [1,D].

[0080] The selection is to select excellent individuals to enter the next generation according to the fitness value of the individuals, simulating the process of natural selection. Commonly used selection methods include roulette selection, tournament selection and sorting selection.

[0081] In some preferred embodiments, the following selection operation method is provided, specifically including:

[0082] S321, the new test individual obtained by the crossover operation With the parent individual Compare and select according to the formula:

[0083]

[0084] S4. When the number of iterations reaches the preset value N max When the optimal individual X in the current population is best Arrange them in matrix form to get the optimal sampling matrix.

[0085] S5. Collect the echo data of the target object according to the optimal sampling matrix to complete the reconstruction of the millimeter wave image of the target object. Specifically, the system for millimeter wave imaging in the present invention is a plane scanning system, and the target echo collected by the receiving antenna at the position (x', y') on the scanning plane is expressed as s(x', y', k), where k = 2πf / c represents the wave number, f is the system operating frequency, and c is the speed of light. The received echo data of the target object is processed using a conventional sparse millimeter wave imaging algorithm to obtain the corresponding scattering coefficient of the three-dimensional target point. Reconstruct the millimeter wave image based on the obtained target scattering coefficient. Reconstruct the millimeter wave image based on the obtained target scattering coefficient. Specifically, the scattering coefficient of the target point is modulo and normalized, and the normalized value is corresponded to the grayscale value of the grayscale image, so that the millimeter wave image corresponding to the target scattering coefficient can be obtained.

[0086] The present invention also provides a device, which is used to implement the above method.

[0087] The present invention also provides a storage medium, the storage medium includes a stored program, wherein the program executes the above method when it is run.

[0088] The present invention also provides a processor, which is used to run a program, wherein the program executes the above method when it is run.

[0089] Experimental example

[0090] 1. Imaging scene.

[0091] This experimental example uses a plane scanning three-dimensional active millimeter wave imaging system, and its imaging scene is as follows Figure 2 As shown, it is assumed that each pair of transmitting antennas and receiving antennas are approximately located at the same position on the z = Z0 plane (in practice, the midpoint position of the receiving antenna and the transmitting antenna is used as the equivalent position of the transmitting and receiving antenna pair), the target is located in front of the antenna array plane, and the scattering coefficient of the point target located at (x, y, z) is represented by g(x, y, z). When the imaging system is working, the transmitting antenna radiates electromagnetic waves to the target, and the receiving antenna receives the scattered echo of the target. The echo collected by the receiving antenna at the position (x', y') on the scanning plane is shown as follows: s(x', y', k) = ∫∫∫g(x, y, z)·e -j2kR / R·dx·dy·dz; where represents the distance between the transmitting and receiving antenna pair and the target point, k=2πf / c represents the wave number, f is the system operating frequency, and c is the speed of light.

[0092] 2. System parameters.

[0093] The transmission signal of the millimeter wave imaging system used in this experimental example is a linear frequency modulation signal with a frequency range of 27 to 32.8 GHz and a bandwidth of B = 5.8 GHz. The number of demultiplexed frequency modulation signal data collected at each equivalent sampling point position (corresponding to the number of frequency samples) is 220. The full sampling sampling intervals in the row direction (x direction) and the column direction (y direction) are both 5 mm, the plane imaging area is 1.895 m × 0.785 m, and the number of sampling points under full sampling conditions is 379 × 157.

[0094] When using the millimeter wave imaging sparse sampling method proposed in the present invention, the weighted coefficients λ1, λ2, λ3, and λ4 of each indicator in the sparse sampling minimization objective function are 5.0, 5.0, 0.1, and 0.1 respectively. Furthermore, the parameter settings for iterative solution are as follows: the population NR contains 100 individuals; the encoding dimension D of each individual is the product of the number of sampling points in the horizontal direction and the number of sampling points in the vertical direction, that is, D = 379 × 157 = 59503; the crossover probability CR is 0.05; the maximum evolutionary generation N of the population max For the 50th generation.

[0095] When initializing the population individuals, six different sparse sampling methods, including spiral equal arc length interval sparse sampling method with sampling ratios of 20%, 25%, and 30%, and traditional spiral equal angle sparse sampling method with sampling ratios of 20%, 25%, and 30%, are used as some individuals in the initial population. In addition to the above six individuals, the remaining 94 individuals in the initial population are generated randomly.

[0096] 3. Imaging target object.

[0097] This experiment uses the Figure 3 The metal strips shown are used as imaging targets. The target consists of multiple metal strips, each of which is 100 mm long. The metal strips are placed in three different directions: row (x direction), column (y direction) and 45° inclined direction. Among them, three metal strips form a group, and the spacing between any two adjacent metal strips in the same group is the same as the width of the metal strip. In different directions, the spacing between two adjacent groups of metal strips is 50 mm. Five groups of metal strips are placed in the row direction and the 45° inclined direction. The widths of the five groups of metal strips are 10 mm, 7 mm, 6 mm, 5 mm and 4 mm respectively, and the stripe spacing within the group is the same as the stripe width. In addition, the six groups of metal strips placed in the column direction additionally include a group of metal strips with a width of 20 mm.

[0098] 4. Comparison method.

[0099] a. Full sampling method (as reference image); b. Spiral equal arc length interval sparse sampling method with a sampling ratio of 30%; c. Random sparse sampling method.

[0100] 5. Experimental results

[0101] like Figure 4 As shown, the maximum projection two-dimensional reconstructed image under full sampling conditions is as follows Figure 4 As shown in (a), the reconstructed image under the spiral equal arc length interval sparse sampling condition when the sampling ratio is 30% is as follows Figure 4 As shown in (b), the reconstructed images under random sparse sampling and the sampling of the present invention are respectively as follows Figure 4 (c) and Figure 4 As shown in (d), the sampling ratio of the random sparse sampling and the sparse sampling solution obtained by the present invention is 32%. Figure 4 (c) It can be seen that the aliasing is more serious in the reconstruction of random sparse sampling, and the metal stripe target image is greatly disturbed. Figure 4 (b) with Figure 4 (d) It can be seen that compared with the reconstruction result of spiral equal arc length interval sparse sampling at a sampling ratio of 30%, the reconstructed image result obtained by the present invention has significantly less aliasing phenomenon, the interference to the metal stripe target is reduced, and the imaging quality is relatively higher.

[0102] 6. Image quality evaluation

[0103] by Figure 4 The reconstruction result of full sampling in (a) is the reference image, and the commonly used image quality evaluation indicators such as peak signal-to-noise ratio (PSNR), structural similarity (SSIM) and root mean square error (RMSE) are used to evaluate the spiral equal arc length interval sparse sampling at a sampling ratio of 30%, the random sparse sampling at a sampling ratio of 32%, and the target reconstructed image results of the present invention. Among the above three indicators, the larger the values ​​of PSNR and SSIM, the better, and the smaller the value of RMSE, the better. The evaluation results of the reconstructed image quality under the above three sparse sampling methods are shown in Table 1.

[0104] Table 1 Evaluation table of reconstructed image quality under different sparse sampling methods

[0105] Sampling method PSNR SSIM RMSE Sampling ratio Equal arc length interval sparse sampling 24.9844 0.6176 0.0563 30% Random sparse sampling 21.4481 0.4640 0.0846 32% Original call 25.9839 0.6295 0.0502 32%

[0106] It can be seen from the above table that compared with random sparse sampling and spiral equal arc length interval sparse sampling, the PSNR and SSIM of the reconstructed image of the present invention are significantly larger, the RMSE value is also significantly smaller, and the image quality is relatively better.

[0107] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A millimeter wave imaging sparse sampling method, characterized in that: Includes steps: Construct a sparse sampling minimization objective function f(a) = [λ1·N(a)+λ2·S(a)+λ3·W(a)+λ4·P(a)], where a is a sparse sampling position matrix represented in binary, N(a) is the sampling ratio, S(a) is the average value of the ratio of the area with an amplitude greater than -40 dB in the point spread function in the x-direction and y-direction excluding the main lobe area to the area outside the main lobe, W(a) is the average value of the larger value of the half-wavelength values ​​on the left and right sides of the main lobe peak of the point spread function in the x-direction and y-direction, and P(a) is the average value of the amplitude difference between the peak side lobe of the point spread function in the x-direction and the adjacent main lobe; λ1, λ2, λ3, and λ4 are weighting coefficients of various parameters respectively; The sparse sampling position matrix a is rearranged into a column vector form as the individual X in the differential evolution population i , i = 1, 2, ... NP, NP is the population size; After initializing the individuals in the population, the sparse sampling minimization objective function f(a) is iteratively solved, and each iteration sequentially performs mutation, crossover, and selection operations to update the individuals accordingly; When the number of iterations reaches the preset value N max When the optimal individual X in the current population is best Arrange in matrix form to obtain the optimal sampling matrix; The echo data of the target object is collected according to the optimal sampling matrix to complete the millimeter wave image reconstruction of the target object.

2. The millimeter wave imaging sparse sampling method according to claim 1, characterized in that: The method for initializing individuals in the population includes: Initializing at least one individual in the population using a spiral equiangular sparse sampling scheme; Initializing at least one individual in the population using a spiral equal arc length interval sparse sampling scheme; The remaining individuals in the population are initialized randomly.

3. The millimeter wave imaging sparse sampling method according to claim 2, characterized in that: The random initialization method includes: The initial population The jth gene in the i-th individual Generate it as follows: Among them, rand i,j is a uniformly distributed random number generated between [0,1], and Z is a real number between [0,1] set according to the sparse sampling ratio.

4. The millimeter wave imaging sparse sampling method according to claim 1, characterized in that: The mutation operation method includes: Randomly select three individuals from the current population and in, represents the i-th individual in the g-th generation population, and r1, r2, r3∈{1,2,…,NP}, i≠r1≠r2≠r3; The new target individual obtained by the mutation operation without mutation parameters The j-th gene of is represented by:

5. The millimeter wave imaging sparse sampling method according to claim 1, characterized in that: The crossover operation method comprises: The new target individual obtained after mutation With the parent individual Perform crossover to generate new test individuals Test individual The j-th gene of is represented by: Where CR is the crossover probability and rand(D) is a random integer between [1,D].

6. The millimeter wave imaging sparse sampling method according to claim 1, characterized in that: The method for selecting an operation includes: The new test individuals obtained by the crossover operation With the parent individual Compare and select according to the formula:

7. A device, characterized in that: The device is used to implement the method according to any one of claims 1 to 6.

8. A storage medium, characterized in that: The storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6 when executed.

9. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 6 when running.

Citation Information

Cited By

  • Echo data sparse acquisition method for microwave near-field three-dimensional imaging

    CN121208759A

  • Echo data sparse acquisition method for microwave near-field three-dimensional imaging

    CN121208759B