A software-based frequency-domain matched filtering method based on the VSIPL standard

Abstract

The present invention discloses a software-based frequency domain matched filtering method based on the VSIPL standard. The method comprises the following steps: initializing a VSIPL library, binding a complex block object to a user data array, and binding a matrix view to the complex block object to convert the complex block object into a permitted state; calling a vsip_ccfftmop_create_f function to create a pointer to a one-dimensional multi-FFT object; performing a fast Fourier transform on a radar echo matrix using the vsip_ccfftmop_f function to obtain first data transformed into a frequency domain; calling a vsip_fftm_destroy_f function to destroy the one-dimensional multi-FFT object; cyclically calling a vsip_cvmul_f function to multiply Nrfft data points in each row of the first data by a matched filtering coefficient to obtain second data; and calling a vsip_ccfftmop_create_f function, a vsip_ccfftmop_f function, and a vsip_fftm_destroy_f function to convert the second data into a time domain to obtain a matched filtering result. The present invention not only ensures the portability of the program, but also improves the operating efficiency of the VSIPL middleware on the Feiteng platform.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and in particular relates to a software-based frequency domain matched filtering method based on the VSIPL standard. Background Art

[0002] Software-based radar has an open architecture with features such as hardware reconfiguration, demand customization, and software reconfiguration. It can effectively solve the problems of poor versatility and low reuse rate of current radar signal processing platforms designed based on high-performance DSP (Digital Signal Processing) and FPGA (Field Programmable Gate Array).

[0003] VSIPL (Vector Signal Image Processing Library) is an algorithm library for vector and signal processing development. It defines an application programming interface with a unified industry standard and can support applications for vector, signal, and image processing in real-time embedded signal processing systems. It has the advantages of portability and high performance. While being portable, it maintains high software performance. The same software can be expanded to other hardware platforms. There is no need to rewrite the software when upgrading the system, which reduces development costs and extends the software lifecycle.

[0004] Prior art computing middleware based on the VSIPL standard features a unified VSIPL API, simplifying the development of cross-platform algorithms. JVSIPL is an open-source library based on the VSIPL 1.3 standard that implements basic mathematical operations, vector operations, matrix operations, and signal processing. However, it is entirely written in C and not optimized for any specific platform, resulting in low program execution efficiency.

[0005] Furthermore, algorithm components are functionally complete, independent, reusable, and collaborative programs. They are standardized, modular, and universal, allowing for independent development, compilation, debugging, and testing. In existing technologies, developers encapsulate algorithms using standard interfaces, designing both standard and custom components, and visualizing these components using a graphical development environment. This allows for rapid construction of software-based radar system solutions and object-oriented system design. However, algorithm components currently lack a unified interface across platforms. While implementations can call upon underlying high-efficiency libraries, they lack the VSIPL computing middleware to shield against underlying hardware differences, making cross-platform code portability impossible. Summary of the Invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a software-based frequency domain matched filtering method based on the VSIPL standard. The technical problem to be solved by the present invention is achieved by the following technical solutions:

[0007] The present invention provides a software-based frequency domain matched filtering method based on the VSIPL standard, comprising:

[0008] Initialize the VSIPL library, bind a complex block object to a user data array, and bind a matrix view to the complex block object, so that the complex block object is converted to a permitted state; the user data array includes a radar echo matrix;

[0009] Call the vsip_ccfftmop_create_f function to create a pointer to a one-dimensional multi-FFT object, where the one-dimensional multi-FFT object includes: the number of rows Nafft, the number of columns Nrfft, and complex matrix FFT related information of the radar echo matrix;

[0010] Calling the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain;

[0011] Call the vsip_fftm_destroy_f function to destroy the one-dimensional multi-FFT object;

[0012] The vsip_cvmul_f function is called cyclically to multiply Nrfft data points in each row of the first data by the matched filter coefficient to obtain the second data;

[0013] The vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function are called to convert the second data into the time domain to obtain a matched filtering result.

[0014] In one embodiment of the present invention, the complex matrix FFT related information includes at least: fast Fourier transform by row or by column, scale factor, transform direction, and estimated number of times the one-dimensional multi-FFT object is called.

[0015] In one embodiment of the present invention, the step of calling the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain includes:

[0016] Based on the number of rows Nafft, the number of columns Nrfft of the radar echo matrix and related information for calculating the complex matrix FFT, the radar echo matrix is ​​fast Fourier transformed row by row by calling the vsip_ccfftmop_f function to obtain first data after the radar echo matrix is ​​transformed into the frequency domain.

[0017] In one embodiment of the present invention, based on the number of rows Nafft and the number of columns Nrfft of the radar echo matrix and related information for calculating the complex matrix FFT, the step of performing a fast Fourier transform on the radar echo matrix row by row by calling the vsip_ccfftmop_f function to obtain first data after the radar echo matrix is ​​transformed into the frequency domain includes:

[0018] Initialize the number of cycles M = 0;

[0019] Performing a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function, and the vsip_fft_destroy_f function;

[0020] After setting M = M + 1, determine whether M is less than Nafft; if so, return to the step of performing a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_f function; if not, obtain the first data after the radar echo matrix is ​​transformed into the frequency domain.

[0021] In one embodiment of the present invention, the step of performing a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function, and the vsip_fft_destroy_f function includes:

[0022] Call the vsip_ccfftop_create_f function to create a pointer to a one-dimensional FFT object; the one-dimensional FFT object includes complex vector FFT related information;

[0023] Call the vsip_ccfftop_f function to determine whether to perform a forward transform or an inverse transform on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix based on the complex vector FFT related information, and further call the ne10_fft_c2c_1d_float42_neon function to calculate the FFT forward transform or FFT inverse transform of the complex vector composed of Nrfft data in the Mth row of the radar echo matrix;

[0024] Call the vsip_fft_destroy_f function to destroy the one-dimensional FFT object.

[0025] In one embodiment of the present invention, when a forward transform is performed on a complex vector consisting of Nrfft data in the Mth row of the radar echo matrix, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 0; conversely, when an inverse transform is performed on a complex vector consisting of Nrfft data in the Mth row of the radar echo matrix, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 1.

[0026] In one embodiment of the present invention, after calling the vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function to convert the second data into the time domain and obtain the matched filtering result, the method further includes:

[0027] Destroys the complex block object and matrix view.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention provides a software-based frequency-domain matched filtering method based on the VSIPL standard. In the process of implementing the middleware, the NE10 and OpenBLAS underlying high-efficiency libraries on the ARM platform are called. Complex vector FFT operations are implemented by calling the NE10 library. Compared with the FFT function in the open source VSIPL library, the efficiency is increased by more than 10 times, which can effectively improve the operating efficiency of the VSIPL middleware on the Feiteng platform.

[0030] In the process of implementing the signal processing component, the present invention optimizes the middleware in the process of implementing the middleware by calling the VSIPL middleware. On the other hand, since the VSIPL standard has a unified interface, there is no need to modify the code when calling the signal processing component implemented on Feiteng by calling the VSIPL middleware and transplanting it to a new hardware platform, thereby improving the portability of the program.

[0031] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of a software-based frequency domain matched filtering method based on the VSIPL standard provided by an embodiment of the present invention;

[0033] Figure 2This is a flow chart of a frequency domain pulse compression algorithm provided by an embodiment of the present invention;

[0034] Figure 3a This is a schematic diagram of the results of matlab frequency domain matched filtering provided by an embodiment of the present invention;

[0035] Figure 3b 2. It is a schematic diagram of the result of frequency domain matched filtering of the VSIPL component provided by an embodiment of the present invention;

[0036] Figure 3c yes Figure 3a and Figure 3b The error comparison diagram of frequency domain matched filtering is shown. DETAILED DESCRIPTION

[0037] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0038] Software-based radar features an open architecture that enables flexible expansion of hardware resources, device reconfiguration, and replacement and upgrade of standard hardware platforms. It also decouples hardware and software, shortening the development process. VSIPL is an algorithm library for vector and signal processing development, featuring a unified industry-standard application programming interface (API). Components are independent, deployable, and assembled software units with fixed interface protocols. They are independent of specific hardware platforms and operating systems, making them highly reusable.

[0039] When implementing algorithm components on different hardware platforms, chip manufacturers can utilize efficient computing libraries specifically provided for their processors, such as Intel's MKL (Math Kernel Library) optimized for Intel processors, AMD's AOCL-LibM library optimized for AMD processors, and Apple's Accelerate Framework optimized for Apple processors. However, the inconsistent interface standards for similar functions within these libraries result in poor cross-platform portability of algorithm components. Cross-platform use requires modification or rewriting of the algorithm implementation code, which in turn lengthens the development cycle and increases complexity.

[0040] In view of this, the present invention provides a software-based frequency domain matched filtering method based on the VSIPL standard.

[0041] Figure 1 This is a flow chart of a software-based frequency domain matched filtering method based on the VSIPL standard provided by an embodiment of the present invention. Figure 1 As shown, the embodiment of the present invention provides a software-based frequency domain matched filtering method based on the VSIPL standard, including:

[0042] S1. Initialize the VSIPL library, bind the complex block object to the user data array, and bind the matrix view to the complex block object, so that the complex block object is converted to the permitted state; the user data array includes the radar echo matrix;

[0043] S2. Call the vsip_ccfftmop_create_f function to create a pointer to a one-dimensional multi-FFT object. The one-dimensional multi-FFT object includes: the number of rows Nafft, the number of columns Nrfft of the radar echo matrix, and complex matrix FFT related information;

[0044] S3. Call the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain;

[0045] S4. Call the vsip_fftm_destroy_f function to destroy the one-dimensional multi-FFT object;

[0046] S5. Loop calling the vsip_cvmul_f function to multiply Nrfft data points in each row of the first data by the matched filter coefficient to obtain the second data;

[0047] S6. Call the vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function to convert the second data into the time domain to obtain a matched filtering result.

[0048] It should be understood that when receiving echo signals, radar receivers inevitably encounter noise and various interference sources, such as clutter generated by distributed objects and noise modulation interference from enemy sources. To select useful targets and suppress various noise and interference sources, filters are required to achieve frequency selection, and filters are crucial components for this task. The filter's bandwidth and frequency characteristics affect filtering effectiveness and are directly related to key radar receiver performance indicators such as sensitivity and waveform distortion. To maximize the signal-to-noise ratio (SNR) at the receiver output and minimize waveform distortion for varying input signals and noise interference, the filter requires optimal bandwidth and frequency characteristics for optimal filtering. A matched filter maximizes the filter output SNR at a specific moment.

[0049] In traditional pulse radar systems, the range resolution is calculated as follows:

[0050]

[0051] Where τ represents the pulse width and B represents the pulse bandwidth.

[0052] As can be seen, improving radar resolution requires reducing τ. However, reducing τ also reduces the radar's average transmit power, leading to a decrease in radar detection capability. Clearly, these two important indicators, radar detection capability and radar resolution, are in conflict. Using matched filtering technology can ensure that the radar achieves high range resolution while maintaining a high average transmit power.

[0053] In the specific implementation of matched filtering, the calculation methods are divided into time domain pulse compression and frequency domain pulse compression. The matched filter coefficient can be obtained by taking the conjugate and flipping the transmitted signal.

[0054] The time-domain pulse compression method is relatively simple. Taking a discrete digital signal as an example, let y(n) be the output of the matched filter, x(n) be the input signal of the matched filter, and h(n) be the pulse compression coefficient. Its calculation method is to perform a convolution operation, as shown in the following formula:

[0055]

[0056] like Figure 2 As shown, for the frequency domain pulse compression method, the input signal x(n) needs to be converted into the frequency domain through fast Fourier transform to obtain X(k), as shown in the following formula:

[0057] X(k)=FFT[x(n)];

[0058] Then perform a dot multiplication operation on the transformed signal X(k) and the frequency domain pulse compression coefficient, as shown in the following formula:

[0059] Y(k)=X(k)·H(k);

[0060] Where H(k) is the result of transforming the time-domain pulse compression coefficient into the frequency domain through FFT calculation, i.e., the transfer function of the matched filter, which is as follows:

[0061]

[0062] Finally, the dot product result Y(k) is transformed into the time domain by IFFT, as shown in the following formula:

[0063] y(n)=IFFT[Y(k)].

[0064] In this embodiment, the complex matrix FFT related information includes at least: fast Fourier transform by row or by column, scale factor, transform direction, and estimated number of times the one-dimensional multi-FFT object is called.

[0065] Specifically, in step S3, the step of calling the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain includes:

[0066] Based on the number of rows Nafft, the number of columns Nrfft of the radar echo matrix and the related information of calculating the complex matrix FFT, the radar echo matrix is ​​fast Fourier transformed row by row by calling the vsip_ccfftmop_f function to obtain the first data after the radar echo matrix is ​​transformed into the frequency domain.

[0067] In this embodiment, the vsip_ccfftmop_f function is used to implement the fast Fourier transform of the radar echo matrix, wherein the FFT calculated row by row can be expressed as follows:

[0068]

[0069] Calculating FFT by column can be expressed as follows:

[0070]

[0071] Among them, W P =e signj2π / Nafft ,W T =e signj2π / Nrfft , scale is the scaling factor, which needs to be selected according to the type of change, including 1, 1 / Nrfft, and When calculating FFT, sign=-1, and when calculating IFFT, sign=1.

[0072] When performing fast Fourier transform on the radar echo matrix row by row, based on the number of rows Nafft, the number of columns Nrfft of the radar echo matrix, and related information for calculating the complex matrix FFT, the step of performing fast Fourier transform on the radar echo matrix row by row by calling the vsip_ccfftmop_f function to obtain first data after the radar echo matrix is ​​transformed into the frequency domain includes:

[0073] Initialize the number of cycles M = 0;

[0074] By calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function and the vsip_fft_destroy_f function, the complex vector composed of Nrfft data in the Mth row of the radar echo matrix is ​​fast Fourier transformed;

[0075] After setting M = M + 1, determine whether M is less than Nafft. If so, return to the step of calling the vsip_ccfftop_f function to perform a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix. If not, obtain the first data after the radar echo matrix is ​​transformed into the frequency domain.

[0076] Since the NE10 library does not directly implement a complex matrix FFT function, this embodiment requires the introduction of the vsip_ccfftop_f function for implementing the fast Fourier transform of a complex vector when performing the fast Fourier transform of the radar echo matrix. In other words, the functionality implemented by the vsip_ccfftop_f function is expressed as follows:

[0077]

[0078] Where W = e signj2π / Nrfft , scale is the scaling factor, which needs to be selected according to the type of change, including 1, 1 / Nrfft, and When calculating FFT, sign=-1, and when calculating IFFT, sign=1.

[0079] Specifically, the steps of performing fast Fourier transform on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function, and the vsip_fft_destroy_f function include:

[0080] Call the vsip_ccfftop_create_f function to create a pointer to a one-dimensional FFT object; the one-dimensional FFT object contains complex vector FFT related information;

[0081] Call the vsip_ccfftop_f function to determine whether to perform a forward transform or an inverse transform on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix based on the complex vector FFT related information, and further call the ne10_fft_c2c_1d_float42_neon function to calculate the FFT forward transform or FFT inverse transform of the complex vector composed of Nrfft data in the Mth row of the radar echo matrix;

[0082] Call the vsip_fft_destroy_f function to destroy the one-dimensional FFT object.

[0083] Among them, when the complex vector composed of Nrfft data in the Mth row of the radar echo matrix is ​​forward transformed, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 0; conversely, when the complex vector composed of Nrfft data in the Mth row of the radar echo matrix is ​​inverse transformed, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 1.

[0084] In this embodiment, implementing complex vector FFT based on VSIPL requires calling multiple functions, such as: vsip_ccfftop_create_f function, vsip_ccfftop_f function and vsip_fft_destroy_f function, where calling vsip_ccfftop_create_f can create a pointer to a one-dimensional FFT object. The FFT object encapsulates complex vector FFT related information, including: the complex vector length of the FFT transformation, the scaling factor, the transformation direction, the estimated number of times the one-dimensional FFT object is called, the algorithm selection and the configuration structure.

[0085] Next, the vsip_ccfftop_f function is called. In the vsip_ccfftop_f function, the first step is to determine whether to perform a forward or inverse transform on the complex vector based on the transform direction (dir parameter) in the one-dimensional FFT object. The dir parameter is set by the vsip_ccfftop_create_f function. If the dir parameter is set to VSIP_FWD, it means that a forward FFT transform is required on the complex vector. If it is a forward transform, the inverse_fft parameter in the ne10_fft_c2c_1d_float32_neon function must be set to 0. Conversely, if the dir parameter is set to VSIP_INV, it means that an inverse FFT transform is required on the complex vector. In this case, the inverse_fft parameter in the ne10_fft_c2c_1d_float32_neon function is set to 1. The ne10_fft_c2c_1d_float32_neon function is then called to calculate the forward or inverse FFT transform of the complex vector.

[0086] For step S5, a for loop may be used to call the vsip_cvmul_f function in each loop, for a total of Nafft times. Each time, the result of multiplying Nrfft data in each row of the first data by the matched filter coefficient is calculated. For example, the function of the vsip_cvmul_f function is to calculate the product of two vectors element by element, which is expressed as follows:

[0087] r j =a j b j ,j=0,1,...,Nrfft-1;

[0088] The second data can be obtained by using a for loop to traverse each element in the two vectors (matching filter coefficients and Nrfft data in each row of the first data).

[0089] In addition, it should be noted that after obtaining the second data, the second data can also be subjected to IFFT calculation and transformed into the time domain by calling the three functions vsip_ccfftmop_create_f, vsip_ccfftmop_f and vsip_fftm_destroy. In this process, the dir parameter of the vsip_ccfftmop_create_f function is set to VSIP_INV, which means performing an inverse FFT transform to obtain the matched filtering result.

[0090] Optionally, after calling the vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function to convert the second data into the time domain and obtain the matched filtering result, the method further includes:

[0091] Destroys the complex block object and matrix view.

[0092] In the vsip_fft_destroy_f function, the NE10_FREE function and the ne10_fft_destroy_c2c_float32 function are included.

[0093] The following is a simulation experiment to further illustrate the software-based frequency domain matched filtering method based on the VSIPL standard provided by the present invention.

[0094] Figure 3a This is a schematic diagram of the result of matlab frequency domain matched filtering provided by an embodiment of the present invention. Figure 3b Schematic diagram of the result of frequency domain matched filtering of the VSIPL component provided by an embodiment of the present invention. Figure 3c yes Figure 3a and Figure 3b The error comparison diagram of the second frequency domain data is shown in FIG. Figure 3a-3b As shown, both the Matlab program and the VSIPL component program can detect the target. Figure 3c It can be seen that the error between the two test results is less than 1×10 -5 , indicating that the VSIPL component functions correctly.

[0095] Table 1 shows the performance test results of the vsip_ccfftop_f function provided by an embodiment of the present invention. As shown in Table 1, after the NE10 library is called on the Feiteng CPU to optimize the vsip_ccfftop_f function, the function's running efficiency is much improved compared to the vsip_ccfftop_f function in Jvsipl.

[0096] Table 1 Performance test results of vsip_ccfftop_f function

[0097] Points Call NE10 library optimization (unit: μs) JVSIPL (unit: μs) Speedup 128 0.90 16.10 17.89 256 2.10 31.79 15.14 512 4.10 64.30 15.68 1024 9.50 167.10 17.59 2048 20.20 335.60 16.61 4096 45.90 721.09 15.71 8192 113.59 1805.59 15.89 16384 292.79 3815.30 13.03 32768 725.90 8014.50 11.04

[0098] It can be seen from the above embodiments that the beneficial effects of the present invention are:

[0099] The present invention provides a software-based frequency-domain matched filtering method based on the VSIPL standard. In the process of implementing the middleware, the NE10 and OpenBLAS underlying high-efficiency libraries on the ARM platform are called. Complex vector FFT operations are implemented by calling the NE10 library. Compared with the FFT function in the open source VSIPL library, the efficiency is increased by more than 10 times, thereby improving the operating efficiency of the VSIPL middleware on the Feiteng platform.

[0100] In the process of implementing the signal processing component, the present invention optimizes the middleware in the process of implementing the middleware by calling the VSIPL middleware. On the other hand, since the VSIPL standard has a unified interface, there is no need to modify the code when calling the signal processing component implemented on Feiteng by calling the VSIPL middleware and transplanting it to a new hardware platform, thereby improving the portability of the program.

[0101] In the description of the present invention, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0102] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A software-based frequency domain matched filtering method based on the VSIPL standard, characterized in that: include: Initialize the VSIPL library, bind a complex block object to a user data array, and bind a matrix view to the complex block object, so that the complex block object is converted to a permitted state; the user data array includes a radar echo matrix; Call the vsip_ccfftmop_create_f function to create a pointer to a one-dimensional multi-FFT object. The one-dimensional multi-FFT object includes: the number of rows Nafft, the number of columns Nrfft of the radar echo matrix, and complex matrix FFT related information; the complex matrix FFT related information includes at least: fast Fourier transform by row or column, scale factor, transformation direction, and an estimated number of times the one-dimensional multi-FFT object is called; Calling the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain; Call the vsip_fftm_destroy_f function to destroy the one-dimensional multi-FFT object; The vsip_cvmul_f function is called cyclically to multiply Nrfft data points in each row of the first data by the matched filter coefficient to obtain the second data; Calling the vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function to convert the second data into the time domain to obtain a matched filtering result; The step of calling the vsip_ccfftmop_f function to perform a fast Fourier transform on the radar echo matrix to obtain first data after the radar echo matrix is ​​transformed into the frequency domain includes: initializing the number of cycles M=0; Performing a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function, and the vsip_fft_destroy_f function; After setting M = M + 1, determine whether M is less than Nafft; if so, return to the step of performing a fast Fourier transform on the complex vector consisting of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_f function; if not, obtain the first data after the radar echo matrix is ​​transformed into the frequency domain.

2. A software-based frequency domain matched filtering method based on the VSIPL standard according to claim 1, characterized in that: The step of performing fast Fourier transform on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix by calling the vsip_ccfftop_create_f function, the vsip_ccfftop_f function, and the vsip_fft_destroy_f function includes: Call the vsip_ccfftop_create_f function to create a pointer to a one-dimensional FFT object; the one-dimensional FFT object includes complex vector FFT related information; Call the vsip_ccfftop_f function to determine whether to perform a forward transform or an inverse transform on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix based on the complex vector FFT related information, and further call the ne10_fft_c2c_1d_float42_neon function to calculate the FFT forward transform or FFT inverse transform of the complex vector composed of Nrfft data in the Mth row of the radar echo matrix; Call the vsip_fft_destroy_f function to destroy the one-dimensional FFT object.

3. A software-based frequency domain matched filtering method based on the VSIPL standard according to claim 2, characterized in that: When a forward transform is performed on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 0; conversely, when an inverse transform is performed on the complex vector composed of Nrfft data in the Mth row of the radar echo matrix, the parameter inverse_fft in the ne10_fft_c2c_1d_float32_neon function is 1.

4. The software-based frequency domain matched filtering method based on the VSIPL standard according to claim 1, characterized in that: After calling the vsip_ccfftmop_create_f function, the vsip_ccfftmop_f function, and the vsip_fftm_destroy_f function to convert the second data into the time domain and obtain a matched filtering result, the method further includes: Destroys the complex block object and matrix view.

Images