An array stream type compression method, system and device based on DCT domain sparse reconstruction

Abstract

The application discloses an array stream type compression method, system and device based on DCT domain sparse reconstruction, and the method comprises the following steps: processing an array stream type matrix according to frequency to obtain an amplitude matrix and a phase matrix; performing quantization coding on the amplitude matrix and the phase matrix to obtain an amplitude quantization matrix and a phase quantization matrix, and performing sub-block segmentation on the amplitude quantization matrix and the phase quantization matrix to generate a plurality of amplitude quantization matrix sub-blocks and a plurality of phase quantization matrix sub-blocks, and performing DCT transformation on the amplitude quantization matrix sub-blocks and the phase quantization matrix sub-blocks to obtain sparse matrices in the DCT domain; performing DCT inverse transformation on the sparse matrices in the DCT domain to realize sparse reconstruction of the amplitude quantization matrix sub-blocks and the phase quantization matrix sub-blocks in the DCT domain, and obtaining original sub-block matrix data; splicing all the original sub-block matrix data to obtain original amplitude matrices and phase matrices, calculating complex plane matrices of each frequency, and obtaining an original array stream type matrix. The application effectively reduces the storage space and significantly improves the data access efficiency.

Description

Technical Field

[0001] This application relates to the field of array signal processing technology, and in particular to an array manifold compression method, system and device based on DCT domain sparse reconstruction. Background Technology

[0002] In array signal processing, the wider the frequency band coverage, the broader the field of view, and the denser the spatial grid division, the larger the number of grid points, and the more dimensional the array manifold data becomes, resulting in a dramatic increase in the storage dimension. Traditional storage methods, such as... Figure 1 As shown, directly unfolding the raw 3D data creates a large contiguous storage space, which is then stored sequentially in memory according to its storage address. This consumes a significant amount of storage space and resources. Traditional storage uses a direct data storage method without considering data structure or extracting useful information from the data. This results in a large amount of redundancy in the stored data, high storage resource utilization, and a significant impact on the performance of high-speed data read / write caching operations. Summary of the Invention

[0003] To address the problem that high-density grid partitioning in wide-frequency, large-field-of-view array signal processing leads to array manifold data occupying a large amount of storage space and consuming a large amount of storage resources, this application utilizes the sparsity of array manifold data in the DCT (Discrete Cosine Transform) domain to provide an array manifold compression method, system, and device based on sparse reconstruction in the DCT domain.

[0004] This application discloses an array manifold compression method based on DCT domain sparse reconstruction, which includes:

[0005] Step 1: Collect array manifold data in a microwave anechoic chamber and construct an array manifold matrix. Process the array manifold matrix by frequency and obtain the amplitude matrix and phase matrix based on the steering vector of each frequency.

[0006] Step 2: Perform quantization encoding on the amplitude matrix and phase matrix to obtain the amplitude quantization matrix and phase quantization matrix, and then perform sub-block segmentation to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively;

[0007] Step 3: Perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain a sparse matrix in the DCT domain; the sparse coefficients in the sparse matrix are equivalent to all the information in the original data; the original data is array manifold data;

[0008] Step 4: Using the sparse coefficients in the sparse matrix of the DCT domain, perform inverse DCT transformation on the sparse matrix of the DCT domain, realize sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, and obtain the original sub-block matrix data.

[0009] Step 5: Arrange all the original sub-block matrix data in order and splice them to obtain the original amplitude matrix and phase matrix, thereby reconstructing the amplitude matrix and phase matrix; use the recovered original amplitude matrix and phase matrix to calculate the complex plane matrix of each frequency, and arrange them by page according to frequency to obtain the original array manifold matrix.

[0010] Further, step 1 includes:

[0011] The array manifold matrix contains full-band, full-field-of-view data, and each steering vector... , , , Represent arbitrary center frequency, elevation angle, and azimuth angle, respectively. , , ; The number of grids for the frequency. The number of grids for the pitch angle. The number of grids to be divided for the azimuth angle;

[0012] According to Euler's formula, the complex number of the steering vector can be converted into a two-dimensional planar matrix representation of amplitude and phase using the following formula:

[0013] (3)

[0014] In the formula, Indicates amplitude, Indicates phase, The imaginary unit;

[0015] They were obtained respectively The magnitude matrix and phase matrix of the dimension.

[0016] Furthermore, the array manifold matrix is ​​arranged in pages according to frequency, with each frequency occupying one page, for a total of [number missing]. Page; Two-dimensional data of each frequency at all angles constitute a After the complex matrix plane is transformed into a two-dimensional plane matrix representation of amplitude and phase according to Euler's formula, The plane decomposition of complex matrices is as follows dimensional magnitude matrix and A phase matrix of 3D; each frequency has a full range of angles including azimuth and elevation.

[0017] Further, step 2 includes:

[0018] right The amplitude and phase matrices are quantized and encoded to obtain amplitude quantization and phase quantization matrices, which are then segmented into sub-blocks to generate... Each amplitude quantization matrix sub-block and There are 1 phase quantization matrix sub-blocks, each sub-block matrix having a size of 1. .

[0019] Further, step 3 includes:

[0020] Performing a 2D DCT transformation on each sub-block is equivalent to projecting the sub-image onto... Transformation matrix of basis functions in the DCT domain As shown in equation (6), the sparse representation of the amplitude and phase subblocks in the DCT domain can be written in matrix form as shown in equation (7):

[0021] (6)

[0022] (7)

[0023] In the formula, and Representing amplitude and phase respectively Original sub-block matrix data, and Represent the amplitude and phase of the DCT transformation, respectively. The DCT domain matrix data of each sub-block is all 3D matrix The length of the sub-block's rows and columns; and As a sparse matrix, the information of the sub-image is preserved. Storing only the non-zero elements of the sparse matrix is ​​equivalent to storing the original information of the sub-image in the DCT domain, thus realizing data compression of the array manifold. .

[0024] Further, step 4 includes:

[0025] Using the sparsity coefficient of the sub-blocks in the compressed storage, the entire amplitude is processed according to equation (8). Individual blocks, phase Sparse matrix of sub-blocks and Perform inverse DCT transformation, and realize sparse reconstruction of sub-blocks in the DCT domain to sparsely reconstruct the original sub-block matrix data. and ;

[0026] (8)

[0027] in, The transformation matrix of the basis functions in the DCT domain The transpose of .

[0028] Furthermore, the kernel function of the DCT transform is a cosine function. After the image undergoes the DCT transform, the energy is concentrated on the transform coefficients and distributed in the lower left corner of the image; for each frequency, the amplitude and phase two-dimensional plane data... , , The DCT transform is defined as follows:

[0029] (4)

[0030]

[0031] (5)

[0032] In the formula, For DCT domain two-dimensional planar data, , These are the row and column transformation coefficients, respectively.

[0033] Furthermore, the method for acquiring the array stream data includes:

[0034] The signal generator produces signals of different frequencies and radiates electromagnetic signals through the transmitting antenna. The antenna array and data acquisition instrument are mounted on a turntable. The turntable rotates in two dimensions along the azimuth and elevation angles. At different angles, the electromagnetic signals are received through the antenna array. After being collected by the data acquisition instrument, the data is transmitted to the data processor to obtain the array stream data.

[0035] This application also discloses an array manifold compression system based on DCT domain sparse reconstruction, which implements the above-described method and includes:

[0036] The matrix acquisition module is used to acquire array manifold data in a microwave anechoic chamber and construct an array manifold matrix. The array manifold matrix is ​​processed by frequency pagination, and the amplitude matrix and phase matrix are obtained based on the steering vector of each frequency.

[0037] The matrix generation module is used to perform quantization encoding on the amplitude matrix and the phase matrix to obtain the amplitude quantization matrix and the phase quantization matrix, and then perform sub-block segmentation on them to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively.

[0038] The sparse matrix acquisition module is used to perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain a sparse matrix in the DCT domain; the sparse coefficients in the sparse matrix are equivalent to all the information in the original data;

[0039] The inverse transform module is used to perform an inverse DCT transform on the sparse matrix in the DCT domain using the sparse coefficients in the sparse matrix in the DCT domain. This enables sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, resulting in the original sub-block matrix data.

[0040] The matrix recovery module is used to arrange and splice all the original sub-block matrix data in order to reconstruct the original amplitude matrix and phase matrix, thereby realizing the reconstruction of the amplitude matrix and phase matrix. Using the recovered original amplitude matrix and phase matrix, the complex plane matrix of each frequency is calculated and arranged by page according to frequency to obtain the original array manifold matrix.

[0041] This application also discloses an electronic device including a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the method described above.

[0042] Due to the adoption of the above technical solution, this application has the following advantages:

[0043] 1. Compared with the direct storage method of array manifold data, this application can compress and store large-capacity array manifold data with high-density grid division in wide-frequency and large-field-of-view array signal processing, effectively reducing the storage space occupied, and can reconstruct the original array manifold data. The storage resource consumption is much lower than the direct storage method, which significantly improves the data access efficiency.

[0044] 2. Under the condition of high-density grid division in wide-frequency, large-field-of-view array signal processing, this application realizes lightweight compression storage of array manifold data; avoids the direct storage process of high-dimensional data, solves the contradiction between the storage capacity, accuracy and efficiency of array manifold data, and saves storage space. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0046] Figure 1 This is a schematic diagram of the array manifold high-dimensional matrix, three-dimensional matrix, or multi-dimensional matrix structure according to an embodiment of this application;

[0047] Figure 2(a) is a schematic diagram of the original complex matrix data in an embodiment of this application;

[0048] Figure 2(b) is a schematic diagram of the process of converting complex numbers into amplitude and phase using Euler's formula in an embodiment of this application;

[0049] Figure 2(c) is a schematic diagram of the amplitude matrix data generated after conversion according to an embodiment of this application;

[0050] Figure 2(d) is a schematic diagram of the phase matrix data generated after conversion according to an embodiment of this application;

[0051] Figure 3 This is a block diagram illustrating the principle of array-based streaming data compression and reconstruction in an embodiment of this application.

[0052] Figure 4 This is a flowchart illustrating an array manifold compression method based on DCT domain sparse reconstruction according to an embodiment of this application.

[0053] Figures 5(a) and 5(b) show schematic diagrams of sub-block segmentation of the amplitude quantization matrix and the phase quantization matrix, respectively.

[0054] Figures 5(c) and 5(d) show schematic diagrams of the results of DCT transformation on the sub-blocks obtained by segmenting the amplitude quantization matrix and phase quantization matrix, respectively.

[0055] Figure 6 This is a schematic diagram of the root mean square error of the reconstructed array manifold according to an embodiment of this application;

[0056] Figure 7 This is a schematic diagram of the peak signal-to-noise ratio of the reconstructed array manifold according to an embodiment of this application;

[0057] Figure 8 This is a schematic diagram of the array streaming data compression ratio in an embodiment of this application. Detailed Implementation

[0058] The present application will be further described in conjunction with the accompanying drawings and embodiments. The described embodiments are only some embodiments of the present application, and not all embodiments. All other embodiments obtained by those skilled in the art should fall within the protection scope of the present application.

[0059] In existing technologies, array signal processing involves multiple antenna elements forming an array antenna to receive spatial electromagnetic signals. For signals in space... A far-field signal is incident on the antenna array. Array receiving data at any given time for:

[0060] (1)

[0061] in, Indicates the number of spatial signals. Indicates the first ( =1, 2, ... The steering vector of each signal; Indicates the center frequency. , Indicates pitch angle and azimuth angle. For signal vectors, For noise vectors, It is an array manifold.

[0062] Assuming the signal center frequency, the number of grids for elevation and azimuth are respectively... , , ,make , , Represent arbitrary center frequency, elevation, and azimuth angles, respectively. , , Then the steering vector can be expressed as All the guiding vectors form the array manifold matrix. As shown in equation (2).

[0063] (2)

[0064] As can be seen from equation (2), the array manifold matrix It contains all possible guidance vectors in space, array manifold matrix The CCP contains One guide vector, therefore for Dimension. After the array stream data is acquired and generated, it is stored in memory as known stream data. The array signal processing algorithm then uses this data to... and array manifold matrix Estimate the parameter information of the signal.

[0065] Will The original data is directly expanded, opening up a large contiguous storage space. It is then stored in memory sequentially according to its storage address, which will occupy a large amount of storage space and consume a large amount of storage resources.

[0066] In view of this, see Figure 3 This application provides an embodiment of an array manifold compression method based on DCT domain sparse reconstruction, which includes:

[0067] Step 1: Collect array manifold data in a microwave anechoic chamber and construct an array manifold matrix. Process the array manifold matrix by frequency and obtain the amplitude matrix and phase matrix based on the steering vector of each frequency.

[0068] Step 2: Perform quantization encoding on the amplitude matrix and phase matrix to obtain the amplitude quantization matrix and phase quantization matrix, and then perform sub-block segmentation to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively;

[0069] Step 3: Perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain a sparse matrix in the DCT domain; the sparse coefficients in the sparse matrix are equivalent to all the information in the original data; the original data is array manifold data;

[0070] Step 4: Using the sparse coefficients in the sparse matrix of the DCT domain, perform inverse DCT transformation on the sparse matrix of the DCT domain, realize sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, and obtain the original sub-block matrix data.

[0071] Step 5: Arrange all the original sub-block matrix data in order and splice them to obtain the original amplitude matrix and phase matrix, thereby reconstructing the amplitude matrix and phase matrix; use the recovered original amplitude matrix and phase matrix to calculate the complex plane matrix of each frequency, and arrange them by page according to frequency to obtain the original array manifold matrix.

[0072] Optionally, see Figure 1 The array manifold matrix contains data across the entire frequency band and field of view (angle), with each steering vector... Complex numbers are generally used for storage. , , Represent arbitrary center frequency, elevation angle, and azimuth angle, respectively. = 0,1,… , =0,1,… , =0,1,… According to Euler's formula, complex numbers can be expressed in amplitude and phase form, as shown in equation (3).

[0073] (3)

[0074] In the formula, Indicates amplitude, Indicates phase, It is the imaginary unit.

[0075] Storing data directly in the array manifold matrix according to the real and imaginary parts of complex numbers is redundant and will consume a lot of storage resources. , , These represent the number of grids for frequency, elevation angle, and azimuth angle, respectively. The array manifold matrix is ​​arranged in pages according to frequency, with each frequency on one page, for a total of [number missing]. Page. Two-dimensional data for each frequency across all angles (azimuth and elevation) constitute a... The complex matrix plane, after being converted into amplitude and phase representations using Euler's formula, The complex matrix plane can be decomposed into dimensional magnitude matrix plane and The phase matrix plane of dimensionality, the transformation process is as follows Figures 2(a) to 2(d) As shown, Figure 2(a) represents the original complex matrix data, Figure 2(b) represents the process of converting complex numbers into amplitude and phase using Euler's formula, and Figures 2(c) and 2(d) represent the amplitude and phase matrix data generated after the conversion, respectively. Then, the amplitude and phase are quantized and encoded to obtain the amplitude and phase quantization encoded matrix.

[0076] Amplitude and phase quantization coding matrices can be viewed as two-dimensional image information. For amplitude and phase images, the data in the two-dimensional image matrix are highly correlated, containing a large amount of redundant information. What truly describes the amplitude and phase characteristics is the amount of information contained in the image, not the amount of data. Generally, the amount of information contained in an image is finite; images are sparse within a specific domain, meaning sparsity is an inherent property of images. If an image is projected to a certain domain, it can be sparsely represented with less data in that domain, achieving data compression. When needed, the original array manifold data can be reconstructed from the data. The essence of this process is to remove data redundancy, minimizing the amount of stored data without losing information.

[0077] The kernel function of the DCT transform is a cosine function, which has energy concentration properties and information compression capabilities. After the image undergoes the DCT transform, the energy is concentrated on a few transform coefficients, distributed in the lower left corner of the image. This applies to the amplitude and phase two-dimensional plane data for each frequency. , =0,1,… , =0,1,… DCT transform is defined as

[0078] (4)

[0079] , (5)

[0080] In the formula, For DCT domain two-dimensional planar data, , These are the row and column transformation coefficients, respectively.

[0081] Typically, the amplitude image matrix and phase image matrix are processed using block-based DCT, with each sub-block being [size missing]. , Let be the row and column length of the sub-block. Then, the number of sub-blocks after dividing the amplitude image matrix and phase image matrix is: Each sub-block can be viewed as a sub-image. Performing a 2D DCT transformation on each sub-block is equivalent to projecting the sub-image onto a... Transformation matrix of basis functions in the DCT domain As shown in equation (6), the sparse representation of the amplitude and phase subblocks in the DCT domain can be written in matrix form as shown in equation (7).

[0082] (6)

[0083] , (7)

[0084] In the formula, and Representing amplitude and phase respectively Original matrix data of each sub-block and Represent the amplitude and phase of the DCT transformation, respectively. The DCT domain matrix data of each sub-block is all 3D matrix This represents the row and column length of the sub-block. Due to the sparsity of the sub-block image, and A sparse matrix has only a small number of non-zero elements. A sparse matrix preserves the information of the sub-image; storing only the non-zero elements of the sparse matrix is ​​equivalent to storing the original information of the sub-image in the DCT domain, thus achieving data compression of the array manifold.

[0085] During sparse reconstruction of the array manifold matrix, data is compressed for each amplitude and phase sub-block. and Perform inverse DCT transformation to sparsely reconstruct the original sub-block matrix data. and As shown in equation (8).

[0086] (8)

[0087] in, The transformation matrix of the basis functions in the DCT domain The transpose of .

[0088] For all the reconstructed original sub-block matrix data and The original amplitude and phase two-dimensional matrices are arranged and spliced ​​in order. Then, each element in the matrix is ​​converted into a complex number according to Euler's formula to obtain the original array manifold data complex matrix.

[0089] The array manifold compression method based on DCT domain sparse reconstruction represents the array manifold data as amplitude and phase matrices, divides them into sub-blocks, and performs sparse representation of each sub-block in the DCT domain, thus achieving compressed storage of the array manifold data. During sparse reconstruction, the original array manifold data is recovered by using the inverse DCT transform of each sub-block and sub-block concatenation. Its principle block diagram is shown below. Figure 3 As shown.

[0090] This application also provides an embodiment of an array manifold compression system based on DCT domain sparse reconstruction, which implements the method described in the above embodiment, and includes:

[0091] The matrix acquisition module is used to acquire array manifold data in a microwave anechoic chamber and construct an array manifold matrix. The array manifold matrix is ​​processed by frequency pagination, and the amplitude matrix and phase matrix are obtained based on the steering vector of each frequency.

[0092] The matrix generation module is used to perform quantization encoding on the amplitude matrix and the phase matrix to obtain the amplitude quantization matrix and the phase quantization matrix, and then perform sub-block segmentation on them to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively.

[0093] The sparse matrix acquisition module is used to perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain a sparse matrix in the DCT domain; the sparse coefficients in the sparse matrix are equivalent to all the information in the original data;

[0094] The inverse transform module is used to perform an inverse DCT transform on the sparse matrix in the DCT domain using the sparse coefficients in the sparse matrix in the DCT domain. This enables sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, resulting in the original sub-block matrix data.

[0095] The matrix recovery module is used to arrange and splice all the original sub-block matrix data in order to reconstruct the original amplitude matrix and phase matrix, thereby realizing the reconstruction of the amplitude matrix and phase matrix. Using the recovered original amplitude matrix and phase matrix, the complex plane matrix of each frequency is calculated and arranged by page according to frequency to obtain the original array manifold matrix.

[0096] This application also provides an electronic device including a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the methods described in the above embodiments.

[0097] For ease of understanding, this application provides a more specific embodiment:

[0098] Array manifold data was acquired in a microwave anechoic chamber, and data compression and reconstruction experiments were conducted on a data processor. A signal generator produced signals of different frequencies, which were then radiated through a transmitting antenna. The antenna array and data acquisition instrument were mounted on a two-dimensional turntable. The turntable rotated in two dimensions along the azimuth and elevation angles, receiving electromagnetic signals through the antenna array at different angles. After being acquired by the data acquisition instrument, the signals were transmitted to the data processor to obtain the array manifold data. The block diagram of the implementation example is shown below. Figure 4 As shown.

[0099] The method proposed in this application is used to compress and reconstruct the acquired array manifold data, which consists of 500 frequency points, each with an azimuth and elevation angle range of ±20°. The specific steps are as follows:

[0100] 1) The data processor processes the received array stream raw data in 500-frequency-point pagination to obtain the azimuth and elevation two-dimensional plane complex matrix data for each frequency;

[0101] 2) The azimuth and elevation two-dimensional plane complex matrix data of a single frequency are quantized and encoded in amplitude and phase, and then processed in blocks, with the row and column lengths of each block being... =8 (i.e., 8×8 sub-blocks), resulting in 25 amplitude and phase quantization coding matrix sub-blocks;

[0102] 3) Perform DCT transformation on the matrix sub-blocks according to equation (7) to obtain the sparse matrix of the DCT domain of the sub-blocks, and extract and save the sparse coefficients;

[0103] 4) Perform DCT inverse transformation on the sparse matrix of the sub-blocks formed by the sparse coefficients according to equation (8) to reconstruct the original sub-block matrix data;

[0104] 5) Combine all the original sub-block matrix data and arrange them in order to form the original amplitude and phase two-dimensional matrix, and recover the original array manifold complex matrix at that frequency point;

[0105] 6) Repeat steps 2) to 5) to obtain the original array manifold complex matrix for all frequency points;

[0106] 7) Statistically analyze the root mean square error, peak signal-to-noise ratio, compression ratio, and other indicators of the reconstructed array manifold data, and compare the compressed storage method in this application with the direct storage method.

[0107] Figures 5(a) and 5(b) illustrate the sub-block segmentation of the amplitude quantization matrix and phase quantization matrix, respectively. The matrices are divided into 25 sub-blocks, which are then subjected to block-based DCT transformation. Figures 5(c) and 5(d) show the results of DCT transformation on the 25 sub-blocks obtained from the amplitude and phase quantization matrix segmentation, resulting in 25 sub-blocks in the DCT domain. As can be seen from the figures, after the DCT transformation, the energy in each amplitude and phase sub-block is concentrated on a few transform coefficients, distributed in the lower left corner of the image. This preserves the information contained in the image, removes redundant data, and achieves data compression in the DCT domain.

[0108] like Figure 6 As shown, the array manifold data is reconstructed using the sparse coefficients after DCT domain compression and compared with the original array manifold data. The root mean square error (coded value) of the reconstruction of all frequency points (500 in total) is statistically analyzed. The average value is 0.75, indicating that the reconstructed array manifold data has high accuracy and approximates the original array manifold data.

[0109] like Figure 7 As shown, the array manifold data at each frequency point is treated as a two-dimensional image, and the peak signal-to-noise ratio (PSNR) is used to evaluate the compression and reconstruction effect of the array manifold data. PSNR is a widely used objective indicator for evaluating image quality, mainly used to measure the degree of distortion between the reconstructed image and the original image. The higher the PSNR value, the smaller the distortion, and the closer the reconstructed image is to the original image. A PSNR greater than 40dB indicates high reconstruction quality, close to high-precision lossless reconstruction. The PSNR of the reconstructed array manifold data at all frequency points ranges from 59dB to 62dB, with an average peak signal-to-noise ratio of 60.38dB. The statistical results show that the data compression and reconstruction method proposed in this application has high reconstruction accuracy and closely approximates the original array manifold data.

[0110] like Figure 8 As shown, the storage space required for compressed data is statistically analyzed and compared with the storage space required by direct storage, and the data compression rate is calculated. Statistical results show that the array stream data compression rate is better than 60% across all frequency points, with an average compression rate of 78%. Compared with the traditional direct storage method, the storage resource occupancy rate is reduced by nearly 80%, thus verifying that the method of this application significantly outperforms the direct storage method in terms of storage space resource consumption.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and not to limit them. Although this application has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of this application. Any modifications or equivalent substitutions that do not depart from the spirit and scope of this application should be covered within the protection scope of the claims of this application.

Claims

1. An array manifold compression method based on DCT domain sparse reconstruction, characterized in that, include: Step 1: Collect array manifold data in a microwave anechoic chamber and construct an array manifold matrix. Process the array manifold matrix by frequency and obtain the amplitude matrix and phase matrix based on the steering vector of each frequency. Step 2: Perform quantization encoding on the amplitude matrix and phase matrix to obtain the amplitude quantization matrix and phase quantization matrix, and then perform sub-block segmentation to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively; Step 3: Perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain a sparse matrix in the DCT domain; The sparse coefficients in a sparse matrix are equivalent to all the information in the original data; the original data is array-type data. Step 4: Using the sparse coefficients in the sparse matrix of the DCT domain, perform inverse DCT transformation on the sparse matrix of the DCT domain, realize sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, and obtain the original sub-block matrix data. Step 5: Arrange all the original sub-block matrix data in order and splice them to obtain the original amplitude matrix and phase matrix, thereby reconstructing the amplitude matrix and phase matrix; use the recovered original amplitude matrix and phase matrix to calculate the complex plane matrix of each frequency, and arrange them by page according to frequency to obtain the original array manifold matrix.

2. The method according to claim 1, characterized in that, Step 1 includes: The array manifold matrix contains full-band, full-field-of-view data, and each steering vector... , , , Represent arbitrary center frequency, elevation angle, and azimuth angle, respectively. , , ; The number of grids for the frequency. The number of grids for the pitch angle. The number of grids to be divided for the azimuth angle; According to Euler's formula, the complex number of the steering vector can be converted into a two-dimensional planar matrix representation of amplitude and phase using the following formula: (3) In the formula, Indicates amplitude, Indicates phase, The imaginary unit; They were obtained respectively The magnitude matrix and phase matrix of the dimension.

3. The method according to claim 2, characterized in that, The array manifold matrix is ​​arranged in pages according to frequency, with each frequency occupying one page, for a total of [number missing]. Page; Two-dimensional data of each frequency at all angles constitute a After the complex matrix plane is transformed into a two-dimensional plane matrix representation of amplitude and phase according to Euler's formula, The plane decomposition of complex matrices is as follows dimensional magnitude matrix and A phase matrix of 3D; each frequency has a full range of angles including azimuth and elevation.

4. The method according to claim 2, characterized in that, Step 2 includes: right The amplitude and phase matrices are quantized and encoded to obtain amplitude quantization and phase quantization matrices, which are then segmented into sub-blocks to generate... Each amplitude quantization matrix sub-block and There are 1 phase quantization matrix sub-blocks, each sub-block matrix having a size of 1. .

5. The method according to claim 1, characterized in that, Step 3 includes: Performing a 2D DCT transformation on each sub-block is equivalent to projecting the sub-image onto... Transformation matrix of basis functions in the DCT domain As shown in equation (6), the sparse representation of the amplitude and phase subblocks in the DCT domain can be written in matrix form as shown in equation (7): (6) (7) In the formula, and Representing amplitude and phase respectively Original sub-block matrix data, and Represent the amplitude and phase of the DCT transformation, respectively. The DCT domain matrix data of each sub-block is all 3D matrix The length of the sub-block's rows and columns; and As a sparse matrix, the information of the sub-image is preserved. Storing only the non-zero elements of the sparse matrix is ​​equivalent to storing the original information of the sub-image in the DCT domain, thus realizing data compression of the array manifold. .

6. The method according to claim 4, characterized in that, Step 4 includes: Using the sparsity coefficient of the compressed storage sub-blocks, the entire amplitude is processed according to equation (8). Individual blocks, phase Sparse matrix of sub-blocks and Perform inverse DCT transformation, and realize sparse reconstruction of sub-blocks in the DCT domain to sparsely reconstruct the original sub-block matrix data. and ; (8) in, The transformation matrix of the basis functions in the DCT domain The transpose of .

7. The method according to claim 1, characterized in that, The kernel function of the DCT transform is a cosine function. After the image undergoes the DCT transform, the energy is concentrated on the transform coefficients and distributed in the lower left corner of the image; for each frequency, the amplitude and phase are two-dimensional plane data. , , , The number of grids for the pitch angle. The number of grids for the azimuth angle; the DCT transform is defined as: (4) (5) In the formula, For DCT domain two-dimensional planar data, , These are the row and column transformation coefficients, respectively.

8. The method according to claim 1, characterized in that, The method for acquiring the array stream data includes: The signal generator produces signals of different frequencies and radiates electromagnetic signals through the transmitting antenna. The antenna array and data acquisition instrument are mounted on a turntable. The turntable rotates in two dimensions along the azimuth and elevation angles. At different angles, the electromagnetic signals are received through the antenna array. After being collected by the data acquisition instrument, the data is transmitted to the data processor to obtain the array stream data.

9. An array manifold compression system based on DCT domain sparse reconstruction, implementing the method described in any one of claims 1-8, characterized in that, include: The matrix acquisition module is used to acquire array manifold data in a microwave anechoic chamber and construct an array manifold matrix. The array manifold matrix is ​​processed by frequency pagination, and the amplitude matrix and phase matrix are obtained based on the steering vector of each frequency. The matrix generation module is used to perform quantization encoding on the amplitude matrix and the phase matrix to obtain the amplitude quantization matrix and the phase quantization matrix, and then perform sub-block segmentation on them to generate multiple amplitude quantization matrix sub-blocks and multiple phase quantization matrix sub-blocks respectively. The sparse matrix acquisition module is used to perform DCT transformation on each amplitude quantization matrix sub-block and each phase quantization matrix sub-block respectively to obtain the sparse matrix in the DCT domain. The sparse coefficients in a sparse matrix are equivalent to all the information in the original data. The inverse transform module is used to perform an inverse DCT transform on the sparse matrix in the DCT domain using the sparse coefficients in the sparse matrix in the DCT domain. This enables sparse reconstruction of the amplitude quantization matrix sub-blocks and phase quantization matrix sub-blocks in the DCT domain, resulting in the original sub-block matrix data. The matrix recovery module is used to arrange and splice all the original sub-block matrix data in order to reconstruct the original amplitude matrix and phase matrix, thereby realizing the reconstruction of the amplitude matrix and phase matrix. Using the recovered original amplitude matrix and phase matrix, the complex plane matrix of each frequency is calculated and arranged by page according to frequency to obtain the original array manifold matrix.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the method of any one of claims 1-8.

Images