Satellite-borne radar air target detection processing algorithm DSP engineering design system

Abstract

The invention discloses a satellite-borne radar air target detection processing algorithm DSP engineering design system, which is realized by adopting a TI C66x series multi-core DSP, and is used for receiving input radar system parameters, receiving an input processing mode configuration instruction, entering a specified processing mode according to the processing mode configuration instruction, and after entering the specified processing mode, carrying out digital processing on the satellite-borne radar air target detection processing algorithm DSP engineering design system. Acquiring stored original echo data of the target, and performing collaborative parallel processing on the original echo data by utilizing a plurality of cores in the multi-core DSP according to an algorithm code corresponding to a specified processing mode to obtain a target detection result; the system includes six processing modes for executing six different overhead target detection processing algorithms. According to the invention, a multi-algorithm engineering implementation scheme for detecting the free target by the spaceborne radar is realized through the multi-core DSP single board, the limitations of insufficient motion compensation precision and incomplete processing flow in the prior art can be overcome, and the resource consumption and the operation time can be reduced.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, specifically relating to a DSP engineering design system for a spaceborne radar near-air target detection and processing algorithm. Background Technology

[0002] As a key component of space situational awareness and national security defense systems, spaceborne radar bears the important task of effectively detecting, tracking, and identifying high-speed targets in the airspace. However, the extremely high relative motion between the radar platform and the target causes the target echo signal received by the radar to not only exhibit significant cross-range cell movement in the range dimension but also to produce Doppler frequency shift, severely disrupting the coherence between radar pulses. The target's echo energy cannot be effectively focused in the range-Doppler two-dimensional plane, resulting in a huge loss of signal-to-noise ratio and a sharp decline in the radar's detection probability, or even complete undetectability. Therefore, motion compensation of the echo signal through an improved coherent accumulation algorithm is a crucial step in achieving effective accumulation detection of targets in the airspace. While existing typical algorithms are theoretically complete and have been proven in academic research to effectively solve the detection problem of highly maneuverable targets, they often suffer from computational efficiency issues in practical engineering applications. For example, the Keystone transform requires intensive interpolation operations in high-dimensional space, while the computational complexity of GRFT increases exponentially with the parameter search dimension and search range. Especially with matrix operations and data caching under massive amounts of observation data, the real-time requirements of signal processing pose a severe challenge. On the one hand, researchers are improving algorithm performance to reduce computational complexity; on the other hand, existing technologies are also improving computational efficiency by enhancing high-speed hardware platforms. Among these, DSPs are optimized for signal processing operations and, while possessing advantages in high efficiency, low power consumption, radiation resistance, and the implementation of complex, flexible, and high-precision algorithms, the key factor restricting the breakthrough in the ability of spaceborne radar to detect near-air targets remains how to reasonably port the coherent accumulation algorithm and deeply optimize hardware platform resources to enable the algorithm to run efficiently and stably on the spaceborne DSP platform, given the severely limited resources of spaceborne radar platforms (including computing resources and memory bandwidth) and the stringent real-time requirements for target detection.

[0003] When space-based early warning radars detect nearby targets (such as hypersonic vehicles and cruise missiles), the extremely high relative velocity and high-order motion between the target and the radar platform lead to severe range travel and Doppler frequency spread phenomena during long-term coherent accumulation. This causes the accumulated target energy to spread along both the range and Doppler dimensions, reducing the output signal-to-noise ratio and degrading the moving target detection (MTD) performance of the space-based radar. To address these issues, complex coherent accumulation algorithms are commonly used in the field for target motion compensation and energy accumulation. Typical algorithms include the Keystone Transform (KT) based on linear scaling transformation and the Generalized Radon Fourier Transform (GRFT) based on three-dimensional parameter joint search. However, while these methods are theoretically effective, practical engineering applications often require frequent large-scale matrix operations and data exchanges, resulting in high computational complexity. Currently, due to stringent requirements for reliability, power consumption, radiation resistance, and complex multi-algorithm tasks, space-based radar signal processing hardware platforms generally employ high-performance multi-core digital signal processors (DSPs) as the core computing unit. While DSPs possess powerful signal processing capabilities, their computational resources, memory bandwidth, and storage capacity are still severely limited. Implementing the aforementioned algorithms directly on a spaceborne DSP would present serious engineering challenges, including insufficient processing timeliness, excessive memory resource consumption, and high power consumption, making it difficult to meet the demands of real-time onboard processing and severely restricting the space-based radar's ability to detect high-speed, weak targets.

[0004] In summary, the engineering implementation techniques of existing radar motion compensation algorithms are mostly based on FPGA platforms for optimization, and they have shortcomings, such as using a single algorithm or insufficient accuracy. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a DSP engineering design system for a spaceborne radar near-air target detection and processing algorithm. The technical problem to be solved by this invention is achieved through the following technical solution: This invention provides a DSP engineering design system for a spaceborne radar near-field target detection processing algorithm, implemented using a TIC66x series multi-core DSP. The system is used to: receive input radar system parameters; receive input processing mode configuration instructions and enter a specified processing mode according to the instructions; after entering the specified processing mode, acquire stored raw echo data of the target, and, according to the algorithm code corresponding to the specified processing mode, utilize multiple cores in the multi-core DSP to perform collaborative parallel processing of the raw echo data to obtain the target detection result; wherein, the system includes six processing modes: a first processing mode is used to determine the target detection result based on the MTD method and the raw echo data when the target has not moved; a second processing mode is used to determine the target detection result based on a one-dimensional... The target detection result is determined by the filter bank and the target's raw echo data. The third processing mode is used to determine the target detection result based on the first-order KT transform and the target's raw echo data when the target exhibits linear range movement. The fourth processing mode is used to determine the target detection result based on the two-dimensional filter bank and the target's raw echo data when the target exhibits linear range movement, range curvature, and Doppler frequency spread. The fifth processing mode is used to determine the target detection result based on the first-order KT transform, the matching quadratic term method, and the target's raw echo data when the target exhibits linear range movement, range curvature, and third-order Doppler frequency spread. The sixth processing mode is used to determine the target detection result based on the frequency domain GRFT method, the MTD method, and the target's raw echo data when the target exhibits linear range movement, range curvature, and third-order Doppler frequency spread.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) Based on a multi-core DSP board, this invention provides a multi-algorithm engineering implementation scheme for detecting airborne targets by spaceborne radar. For airborne targets with different motion characteristics, corresponding accumulation detection algorithms can be selected, including velocity fuzzy compensation, range movement correction, phase spread compensation and CFAR detection, thereby overcoming the limitations of insufficient motion compensation accuracy and incomplete processing flow in the existing technology. 2) This invention employs the CZT fast transformation method for the KT module, further reducing computational complexity; 3) This invention adopts a word length optimization scheme for the GRFT method, which reduces resource consumption and computation time as much as possible while ensuring accuracy, thereby improving computational efficiency.

[0007] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0008] Figure 1This is an exemplary workflow diagram of the DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm provided in this embodiment of the invention. Figure 2 This is a flowchart illustrating the implementation of the MTD method that ignores distance travel using a multi-core DSP, as provided in this embodiment of the invention. Figure 3A and Figure 3B These are schematic flowcharts illustrating the implementation of two MTD methods that only compensate for linear distance travel using a multi-core DSP, as provided in the embodiments of the present invention. Figure 4A and Figure 4B These are schematic flowcharts illustrating the implementation of two MTD methods for compensating linear distance travel and matching second-order phase using a multi-core DSP, as provided in this embodiment of the invention. Figure 5 This is a flowchart illustrating the implementation of the GRFT-based MTD method using a multi-core DSP according to an embodiment of the present invention. Figure 6 These are the algorithm performance test results of the standard MTD method in MATLAB that ignores distance movement, as provided in the embodiments of this invention; Figure 7 The results are the algorithm performance test results based on the one-dimensional filter bank method in MATLAB, provided by the embodiments of the present invention. Figure 8 The results are the algorithm performance test results based on the first-order KT transform method in MATLAB, provided by the embodiments of the present invention. Figure 9 These are the algorithm performance test results based on the two-dimensional filter bank method in MATLAB, provided by the embodiments of the present invention. Figure 10 These are the algorithm performance test results provided by the embodiments of the present invention in MATLAB based on the first-order KT transform + matching quadratic term method; Figure 11 These are the algorithm performance test results of the GRFT-based MTD method in MATLAB provided in the embodiments of the present invention; Figure 12 This is a comparison of the distance dimension data profile results of the DSP processing results and the MATLAB processing results of the MTD method based on GRFT provided in this embodiment of the invention; Figure 13 This refers to the program execution time in CCS for the complete process of the six algorithms provided in this embodiment of the invention; Figure 14 This is a statistical result of storage resource consumption in CCS provided by an embodiment of the present invention. Detailed Implementation

[0009] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0010] This invention provides a DSP engineering design system for a spaceborne radar near-field target detection processing algorithm. The system is implemented using any model of the TI C66x series multi-core DSP, for example, using a TMS320C6678 digital signal processor, which is a multi-core DSP. Exemplarily, the system's workflow includes the following steps: receiving input radar system parameters; receiving input processing mode configuration instructions and entering the specified processing mode according to the instructions; after entering the specified processing mode, acquiring stored raw echo data of the target, and using multiple cores in the multi-core DSP to perform collaborative parallel processing of the raw echo data according to the algorithm code corresponding to the specified processing mode, to obtain the target detection result. The system includes six processing modes. The first processing mode is used to determine the target detection result based on the MTD method and the target's raw echo data when the target has not moved. The second processing mode is used to determine the target detection result based on a one-dimensional filter bank and the target's raw echo data when the target has moved linearly. The third processing mode is used when the target exhibits linear range movement, determining the target detection result based on the first-order KT transform and the target's raw echo data. The fourth processing mode is used when the target exhibits linear range movement, range curvature, and Doppler frequency spread, determining the target detection result based on a two-dimensional filter bank and the target's raw echo data. The fifth processing mode is used when the target exhibits linear range movement, range curvature, and Doppler frequency spread, determining the target detection result based on the first-order KT transform, a matched quadratic term method, and the target's raw echo data. The sixth processing mode is used when the target exhibits linear range movement, range curvature, and third-order Doppler frequency spread, determining the target detection result based on the frequency domain GRFT method, the MTD method, and the target's raw echo data. The specified processing mode can be any one of the above six processing modes.

[0011] The core function of the cache is to bridge the speed gap between the high-speed processing units of the DSP core and the relatively slow off-chip memory (such as DDR3), and to improve the data access speed between the two. For example, the algorithm code corresponding to the six processing modes mentioned above is stored in the on-chip memory L2 of the multi-core DSP, while the raw echo data is stored in the off-chip memory DDR3. Furthermore, this invention sets both the on-chip memories L1D and L1P of the multi-core DSP as 32K caches. Considering that the echo data from DDR3 needs to be moved to L2 during operation, this step requires requesting dynamic memory space during runtime and cannot be directly run in external DDR3 or shared memory. Therefore, this invention sets L2 as SRAM to improve data access efficiency. When performing large-scale data computations in the C6678 multi-core processor, data transfer is required between the processor core and external DDR3 and MSM SRAM memories. The C6678 provides a high-speed data transfer module: Direct Memory Access (DMA). DMA includes IDMA and EDMA. This invention uses EDMA based on AB synchronous transfer mode to realize data transfer (DDR3, MSM SRAM, L2) between two memory-mapped slave terminals on the device, thereby improving data transfer efficiency.

[0012] For example, when the multi-core DSP is a TMS320C6678 digital signal processor, this multi-core DSP contains eight cores: Core0, Core1, Core2, Core3, Core4, Core5, Core6, and Core7. After entering a specified processing mode, each core acquires a portion of the raw echo data and processes it according to the algorithm corresponding to the specified processing mode to obtain a CFAR detection result. Then, any one of the eight cores calculates the target parameters using the CFAR detection results of the eight cores according to the algorithm corresponding to the specified processing mode, thus obtaining the target detection result. The portion of raw echo data acquired by each core is one-eighth of the target's raw echo data. In other words, when executing the algorithm corresponding to any one of the six processing modes mentioned above, after these eight cores divide the target's raw echo data into eight equal parts, each core simultaneously processes its own portion of the raw echo data to obtain its own CFAR detection result. Then, one core performs a unified calculation of the target parameters using the CFAR detection results obtained by the eight cores to obtain the final target detection result.

[0013] In this invention, each core corresponds to a semaphore, and the eight semaphores corresponding to the eight cores are different and independent of each other. The eight cores perform multi-core synchronization through these eight semaphores. These eight semaphores are arbitrarily selected from 64 independent hardware semaphores provided by the multi-core DSP. There is no necessary connection between these 64 semaphores and hardware resources or cores. When the eight cores perform each multi-core computation task in each processing mode, each core uses a semaphore request function (i.e., ... The function requests its own semaphore, and when its own semaphore is in an idle state, marks it as occupied. It then continuously checks whether the semaphores of the remaining cores are marked as occupied. When all eight cores detect that all eight semaphores are marked as occupied, it indicates that the eight cores have completed the multi-core computation task. Afterwards, each core releases the semaphore through the semaphore release function (i.e., ...). The function releases its corresponding semaphore, resetting it from an occupied state to an idle state. Each core continuously checks whether the semaphores associated with the remaining cores have been reset to an idle state. Once all eight cores have detected that all eight semaphores have been reset to an idle state, each core proceeds to the next multi-core computation task. Specifically, this invention achieves multi-core synchronization by accessing hardware semaphores, using a combination of direct access and polling access. The process is as follows: ① After system power-on, the eight semaphores associated with these eight cores are first released, initializing them all to an idle state; ② When these eight cores execute multi-core tasks, each core... The function itself, marked as occupied, enters a loop to check if the remaining 7 semaphores are all occupied. When all 8 semaphores are occupied, the multi-core task of this stage is considered complete and the subsequent process begins; ③ After the multi-core task of this stage ends, each core calls... The function releases its own semaphore to reset its corresponding semaphore to the idle state. Then it loops again to check until all 8 semaphores are reset to the idle state before each core can start the next stage of the task.

[0014] To address the range and Doppler dimension characteristics of radar echo data, this invention employs a contiguous storage data layout and achieves batch data transmission by configuring the source address step size and offset. In this invention, when each core performs matrix transpose operations, for each element in each row of the matrix to be transposed, the source address step size is set to the size of the element, and the destination address step size is set to the product of the row length and the element size. A single DMA operation completes the transpose of an entire row of elements. Similarly, for each element in each column of the matrix to be transpose, the source address step size is set to the product of the column length and the element size, and the destination address step size is set to the element size. A single DMA operation completes the transpose of an entire column of elements. These two transpose modes improve the efficiency of transpose operations. Furthermore, in this invention, 64-byte alignment is used during EDMA transmission, allowing for one-time transmission without splitting into multiple smaller transmissions, thus improving transmission efficiency. Additionally, this invention uses single-precision floating-point numbers in the C6678. and double-precision floating-point numbers Perform the calculation.

[0015] Software pipelining is an instruction scheduling technique that leverages instruction-level parallelism by overlapping the execution of multiple iterations of a loop. This invention decomposes the loop body into multiple functional stages, allowing different stages of different iterations to execute simultaneously. Enabling the -o2 and -o3 compiler options assists the compiler in effective software pipelining scheduling, and using loop counter registers to manage the pipelining loop results in higher computer resource utilization. To fully utilize the high-speed computing and multi-core parallelism advantages of the C66x processor, TI has developed numerous functions specifically for high-speed data reading and processing, improving the processor's core performance. This invention extensively uses common inline functions, thereby improving program integration and execution efficiency.

[0016] In this invention, among the six processing modes mentioned above, the algorithm corresponding to the first processing mode belongs to the MTD method that ignores distance travel; the two different algorithms corresponding to the second and third processing modes belong to two different MTD methods that only compensate for linear distance travel; the two different algorithms corresponding to the fourth and fifth processing modes belong to two different MTD methods that compensate for linear distance travel and match second-order phase; and the algorithm corresponding to the sixth processing mode belongs to the GRFT-based MTD method. Figure 1 This is an exemplary workflow diagram of the DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm provided in this embodiment of the invention. Figure 1 As shown, by configuring the processing mode, the system can choose to execute any one of the six different algorithms represented by the above four categories.

[0017] In this invention, the MTD method ignoring range travel is used to determine the target detection result based on the MTD method and the target's original echo data when the target has not moved. Specifically, the MTD method ignoring range travel includes steps S10-S15: S10, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data of the target without any linear range travel or Doppler frequency spread; S11, performing FFT transformation on each pulse of the acquired original echo data to obtain pulse-compressed data. The pulse-compressed data is multiplied by the conjugate reference signal to achieve matched filtering, resulting in matched-filtered data; S12, the matched-filtered data is subjected to pulse-by-pulse IFFT transformation to obtain range-azimuth time domain data; S13, the range-azimuth time domain data is subjected to range-gate-by-range FFT transformation on the range-azimuth time domain data Sftm to obtain range-Doppler data; S14, the range-Doppler data is subjected to range-gate-by-range CFAR detection to obtain CFAR detection results; S15, the target parameters are calculated based on the CFAR detection results to obtain the target detection results.

[0018] In this invention, the algorithm code of the MTD method that ignores distance travel includes... Modules and These two processing modules The module is used to perform FFT transformation on each pulse of the echo data to the range frequency domain, and then multiply each pulse point by the conjugate reference signal and perform IFFT transformation to return to the time domain, thereby obtaining range frequency domain-azimuth domain data. This module performs range-gated FFT transform on the range-azimuth time-domain data to obtain range-Doppler data. Then, it performs range-gated CFAR detection on the range-Doppler data to obtain CFAR detection results. Finally, it calculates target parameters on the CFAR detection results to obtain the target detection results (i.e., target estimation parameters). By executing these two modules, the final target estimation parameters can be obtained. For example, Figure 2 It is a multi-core DSP that is utilizing Modules and This is a flowchart illustrating the implementation of the MTD method that ignores distance movement in the module. Sftm represents the generated range-azimuth time-domain data (i.e., range-azimuth time-domain signal), and Stfm represents the generated target detection result. For example... Figure 2 As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal (i.e., n Pulse / 8 (equally divided pulse)), they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 2 As shown, each core needs to use the process outlined in the dashed box when executing the process. function, functions and function, where, The function is used to perform an IFFT transform on each pulse of the echo data. The function is used to perform a distance-gated FFT transform on the echo data. The function is used to perform CFAR detection on the echo data using distance-by-distance gate.

[0019] In this invention, there are two MTD methods that only compensate for linear distance movement. The first method is used to determine the target detection result based on a one-dimensional filter bank and the original echo data of the target when the target only generates linear distance movement. The second method is used to determine the target detection result based on a first-order KT transform and the original echo data of the target when the target only generates linear distance movement.

[0020] Specifically, the first type of MTD method that only compensates for linear distance movement, i.e., the method used to determine the target detection result based on a one-dimensional filter bank and the target's original echo data when the target moves at a linear distance, includes steps S20-S26: S20, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data when the target moves at a linear distance; S21, performing an FFT transform on each pulse of the acquired original echo data to obtain pulse-compressed data, and multiplying the pulse-compressed data by the conjugate reference signal to achieve matched filtering, obtaining the matched-filtered data. According to the steps: S22, perform ambiguity compensation on the matched-filtered data, and perform first-order linear range travel correction on the ambiguity-compensated data for each range gate to obtain range-frequency domain-azimuth domain data; S23, perform IFFT transform on the range-frequency domain-azimuth domain data pulse by pulse to obtain range-azimuth time domain data; S24, perform FFT transform on the range-azimuth time domain data for each range gate to obtain range-Doppler domain data; S25, perform CFAR detection on the range-Doppler domain data for each range gate to obtain CFAR detection results; S26, calculate target parameters on the CFAR detection results to obtain target detection results.

[0021] In this invention, the algorithm code of the first MTD method that only compensates for linear distance travel includes... Module, Modules and Module. The module is used to perform FFT transformation on each pulse of the acquired raw echo data to obtain pulse compressed data, and then multiply the pulse compressed data by the conjugate reference signal to achieve matched filtering and obtain matched filtered data. The module performs ambiguity compensation on the matched-filtered data, and then performs first-order linear range-travel correction on the ambiguity-compensated data step-by-step using range gates to obtain range-frequency domain-azimuth domain data. Finally, it performs an IFFT transform on the range-frequency domain-azimuth domain data pulse-by-pulse to obtain range-azimuth time domain data. By executing these three modules, the final target estimation parameters can be obtained. For example, Figure 3A It is a multi-core DSP that is utilizing Module, Modules and This is a flowchart illustrating the implementation of the first MTD method that only compensates for linear distance travel. Sftm represents the generated matched-filtered data, Sttm_1DML represents the generated range-azimuth time-domain data, and Stfm_focus represents the final target detection result. Figure 3A As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal, they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 3A As shown, each core needs to use the process outlined in the dashed box when executing the process. function, function, function, functions and function, where, The function is used to perform ambiguity number compensation on distance-by-distance gate of echo data. The function is used to perform distance travel correction on the echo data using distance gates.

[0022] Specifically, the second type of MTD method, which only compensates for linear distance movement, is used to determine the target detection result based on the first-order KT transform and the target's original echo data when the target only moves linearly. The method includes steps S30-S36: S30, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data when the target only moves linearly; S31, performing an FFT transform on each pulse of the acquired original echo data to obtain pulse-compressed data, and multiplying the pulse-compressed data by the conjugate reference signal to achieve matched filtering, obtaining the matched-filtered data; S 32. Perform ambiguity compensation on the matched-filtered data, and then perform CZT-based Keystone transform on the ambiguity-compensated data for each range gate to obtain range-frequency domain-azimuth domain data; S33. Perform IFFT transform on the range-frequency domain-azimuth domain data for each pulse to obtain range-azimuth time domain data; S34. Perform FFT transform on the range-azimuth time domain data for each range gate to obtain range-Doppler domain data; S35. Perform CFAR detection on the range-Doppler domain data for each range gate to obtain CFAR detection results; S36. Calculate target parameters on the CFAR detection results to obtain target detection results.

[0023] In this invention, the algorithm code for the second MTD method that only compensates for linear distance travel includes... Module, Modules and Module. The module performs ambiguity compensation on the matched-filtered data, and then performs a CZT-based Keystone transform on the ambiguity-compensated data pulse-by-pulse to obtain range-frequency domain-azimuth domain data. Finally, it performs an IFFT transform on the range-frequency domain-azimuth domain data pulse-by-pulse to obtain range-azimuth time domain data. By executing these three modules, the final target estimation parameters can be obtained. For example, Figure 3B It is a multi-core DSP that is utilizing Module, Modules and This is a flowchart illustrating the second MTD method that compensates only for linear distance travel. Sftm represents the generated matched-filtered data, Sttm_Keystone represents the generated range-azimuth time-domain data, and Stfm_focus represents the final target detection result. Figure 3B As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal, they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 3BAs shown, each core needs to use the process outlined in the dashed box when executing the process. function, function, function, functions and function, where, The function is used to perform a CZT-based Keystone transform on the echo data through distance-gated transitions.

[0024] In this invention, there are two MTD methods for compensating for linear range travel and matching second-order phase. The first method is used to determine the target detection result based on a two-dimensional filter bank and the original echo data of the target when the target exhibits linear range travel, range curvature, and Doppler frequency spread. The second method is used to determine the target detection result based on a first-order KT transform, a method for matching quadratic terms, and the original echo data of the target when the target exhibits linear range travel, range curvature, and Doppler frequency spread.

[0025] Specifically, the first MTD method that compensates for linear range travel and matches second-order phase, i.e., the method used to determine the target detection result based on the two-dimensional filter bank and the target's original echo data when the target exhibits linear range travel, range curvature, and Doppler frequency spread, includes steps S40~S47: S40, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data under the condition that the target exhibits linear range travel, range curvature, and Doppler frequency spread; S41, performing FFT transformation on each pulse of the acquired original echo data to obtain pulse-compressed data, and multiplying the pulse-compressed data by the conjugate reference signal to achieve matched filtering, obtaining the matched-filtered data; S42, applying the matched filter... The data is then subjected to ambiguity compensation, and the ambiguity-compensated data is subjected to first-order linear range travel correction by range gate to obtain range-frequency domain-azimuth domain data; S43, the range-frequency domain-azimuth domain data is subjected to pulse-by-pulse IFFT transformation to obtain range-azimuth time domain data; S44, the quadratic term coefficients of the range-azimuth time domain data are compensated by range gate to obtain echo data after quadratic phase matching; S45, the echo data after quadratic phase matching is subjected to range gate FFT transformation to obtain echo data after azimuth FFT; S46, the echo data after azimuth FFT is subjected to CFAR detection by range gate to obtain CFAR detection results; S47, the target parameters are calculated from the CFAR detection results to obtain the target detection results.

[0026] In this invention, the algorithm code of the first MTD method that compensates for linear distance travel and matches second-order phase includes... Module, Modules and Module. This module is used to perform range-gate compensation of the quadratic term coefficients on the range-azimuth time-domain data to obtain echo data after quadratic phase matching. It then performs a range-gate FFT transform on the echo data after quadratic phase matching to obtain echo data after azimuth FFT. Finally, it performs CFAR detection on the echo data after azimuth FFT, obtaining the CFAR detection result. Finally, it calculates the target parameters on the CFAR detection result to obtain the target detection result. By executing these three modules, the final target estimation parameters can be obtained. For example, Figure 4A It is a multi-core DSP that is utilizing Module, Modules and This is a flowchart illustrating the first MTD method that compensates for linear distance travel and matches second-order phase. Sftm represents the generated matched-filtered data, Sttm_1DML represents the generated range-azimuth time-domain data, and Stfm_focus represents the final target detection result. Figure 4A As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal, they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 4A As shown, each core needs to use the process outlined in the dashed box when executing the process. function, function, function, function, functions and function, where, The function is used to compensate the quadratic coefficients of the echo data by distance gate.

[0027] Specifically, the second type of MTD method, which compensates for linear range travel and matches second-order phase, is used to determine the target detection result based on the first-order KT transform, the matching quadratic term, and the target's original echo data when the target exhibits linear range travel, range curvature, and Doppler frequency spread. The method includes steps S50-S57: S51, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data under the condition that the target exhibits linear range travel, range curvature, and Doppler frequency spread; S52, performing an FFT transform on each pulse of the acquired original echo data to obtain pulse-compressed data, and multiplying the pulse-compressed data by the conjugate reference signal to achieve matched filtering, obtaining the matched-filtered data; S53, performing matched... The filtered data undergoes ambiguity compensation, and the ambiguity-compensated data is then subjected to a CZT-based Keystone transform on a range-gate-by-range basis to obtain range-frequency domain-azimuth domain data; S54, the range-frequency domain-azimuth domain data is subjected to an IFFT transform on a pulse-by-pulse basis to obtain range-azimuth time domain data; S55, the quadratic term coefficients of the range-azimuth time domain data are compensated on a range-gate-by-range basis to obtain echo data after quadratic phase matching; S56, the echo data after quadratic phase matching is subjected to an FFT transform on a range-gate-by-range basis to obtain echo data after azimuth FFT; S57, the echo data after azimuth FFT is subjected to CFAR detection on a range-gate-by-range basis to obtain CFAR detection results; S58, the target parameters are calculated based on the CFAR detection results to obtain the target detection results.

[0028] In this invention, the algorithm code for the second MTD method that compensates for linear distance travel and matches second-order phase includes... Module, Modules and Modules. By executing these three modules, the final target estimation parameters can be obtained. For example, Figure 4B It is a multi-core DSP that is utilizing Module, Modules and This is a flowchart illustrating the second MTD method that compensates for linear distance travel and matches second-order phase. Sftm represents the generated matched-filtered data, Sttm_Keystone represents the generated range-azimuth time-domain data, and Stfm_focus represents the final target detection result. Figure 4B As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal, they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 4BAs shown, each core needs to use the process outlined in the dashed box when executing the process. function, function, Function (also known as) function), function, functions and function.

[0029] In this invention, the GRFT-based MTD method is used to determine the target detection result based on the frequency domain GRFT method, the MTD method, and the target's original echo data when the target exhibits linear range movement, range curvature, and third-order Doppler frequency spread. Specifically, the GRFT-based MTD method includes steps S60-S70: S60, acquiring the original echo data and generating a conjugate reference signal; wherein, the original echo data is the echo data under the conditions of linear range movement, range curvature, and third-order Doppler frequency spread; S61, processing the acquired original echo... Each pulse of the data undergoes an FFT transform to obtain pulse-compressed data. The compressed data is then multiplied by the conjugate reference signal to achieve matched filtering, resulting in matched-filtered data. S62. Range curvature correction is applied to the matched-filtered data pulse by pulse, yielding range-curvature-corrected data. S63. An IFFT transform is performed on each pulse of the range-curvature-corrected data to obtain range-azimuth time-domain data. The range-azimuth time-domain data is in complex form. S64. The amplitude and phase of the complex number are calculated, and the calculated phase is modulo 2π to obtain a new phase. After multiplying the magnitude of the bit and the complex number, the result is converted to float data to obtain the initially optimized range-azimuth time domain data; S65, the phase of the quadratic term coefficient of the range-gate compensation of the initially optimized range-azimuth time domain data is defined as double type. Then, the phase defined as double type is modulo 2π to make the phase range between [-π,π], and then the modulo phase is converted to float data to obtain the word-length optimized range-azimuth time domain data; S66, the word-length optimized range-azimuth time domain data is obtained by using the determined float type compensation function to optimize the word length. S67. The range-azimuth time domain data after long-term optimization is compensated for by range-gate quadratic coefficients to obtain echo data after quadratic phase matching; S68. The echo data after quadratic phase matching is compensated for by range-gate cubic coefficients to obtain echo data after tertiary phase matching; S69. The echo data after tertiary phase matching is subjected to range-gate FFT transformation to obtain range-Doppler domain data; S70. The range-Doppler domain data is subjected to range-gate CFAR detection to obtain CFAR detection results; S71. The target parameters are calculated from the CFAR detection results to obtain the target detection results.

[0030] The word length optimization scheme described above can improve computational efficiency. Although it increases the time consumption in numerical conversion, it can ensure data processing accuracy while meeting real-time requirements.

[0031] In this invention, the algorithm code of the GRFT-based MTD method includes... Module, Modules and Module. The module is used to perform range curvature correction on the matched-filtered data one-to-one by range gate to obtain the range curvature corrected data. Then, the range curvature corrected data is subjected to IFFT transformation pulse by pulse to obtain range-azimuth time domain data. The range-azimuth time domain data is in complex form. Next, the amplitude and phase of the complex number are calculated, and the calculated phase is modulo 2π to obtain a new phase. The new phase is multiplied by the amplitude of the complex number, and the result is converted into float data to obtain the initially optimized range-azimuth time domain data. The module is used to perform range-gate compensation on the quadratic coefficients of the initially optimized range-azimuth time-domain data, defining the phase as a double type. Then, the double-type phase is modulo 2π to ensure the phase range is between [-π, π]. The modulo-modified phase is then converted to float data to obtain the word-length optimized compensation function. The selected float-type compensation function is then used to perform range-gate compensation on the word-length optimized range-azimuth time-domain data, obtaining echo data after quadratic phase matching. The echo data after quadratic phase matching is then compensated for with cubic coefficients using range-gate compensation, obtaining echo data after cubic phase matching. The echo data after cubic phase matching is then subjected to range-gate FFT transformation to obtain range-Doppler domain data. CFAR detection is then performed on the range-Doppler domain data using range-gate compensation to obtain the CFAR detection results. Finally, target parameters are calculated from the CFAR detection results to obtain the target detection results. By executing these three modules, the final target estimation parameters can be obtained. For example, Figure 5 It is a multi-core DSP that is utilizing Module, Modules and This is a flowchart illustrating the implementation of the GRFT-based MTD method in the module. Sftm represents the generated matched-filtered data, Sttm_2DML(float) represents the generated word-length optimized range-azimuth time-domain data (i.e., float-type range-azimuth time-domain data), and Stfm_focus represents the final target detection result. Figure 5As shown, after the eight cores (Core0 to Core7) equally divide the raw echo signal, they process their respective raw echo signals in parallel by executing the process shown in the dashed box until they obtain their respective CFAR detection results. Then, one core calculates the target parameters based on the CFAR detection results obtained from the eight cores, thereby obtaining the target detection result. Figure 5 As shown, each core needs to use the process outlined in the dashed box when executing the process. function, Function (also known as) function), function, function, functions and function. The function is used to perform distance curvature correction on the echo data using distance gates. The function is used to compensate the cubic term coefficients of the echo data by distance gate.

[0032] It should be noted that in DSP multi-core parallel design, a task needs to be divided into multiple subtasks and assigned to multiple cores. These cores then execute the subtasks in parallel to improve processing efficiency. This invention primarily processes echo matrices (i.e., echo data). The matrix can be divided into blocks based on the number of rows and columns to execute tasks, and the tasks between these blocks are independent and do not require inter-core communication. Therefore, this invention chooses a data parallel approach for engineering design. Thus, the matrix block processing methods for the aforementioned functions that require parallel execution are shown in Table 1 below: Table 1

[0033] This invention, based on a multi-core DSP board, provides a multi-algorithm engineering implementation scheme for detecting nearby targets using spaceborne radar. For targets with different motion characteristics, appropriate accumulation detection algorithms can be selected, including velocity ambiguity compensation, range migration correction, phase spread compensation, and CFAR detection, thus overcoming the limitations of insufficient motion compensation accuracy and incomplete processing flow in existing technologies. When performing parameter search, this invention adopts a "coarse-to-fine" approach—that is, a coarse search first determines the approximate range of target parameters, and then a fine search is performed. Simultaneously, the KT module employs a CZT fast transformation method, further reducing computational complexity. This invention uses a word length optimization scheme for the GRFT method, minimizing resource consumption and computation time while ensuring accuracy, thereby improving computational efficiency. Furthermore, by optimizing other aspects of the DSP performance, this invention further reduces resource occupancy and improves program running efficiency compared to existing technologies.

[0034] The effects of the present invention will be further illustrated below through simulation experiments.

[0035] Simulation conditions: The simulation experiment parameters for this example are configured as shown in the table below: Table 2 Simulation Experiment Parameter Configuration

[0036] For the standard MTD method that ignores distance travel, the target has only a radial velocity of 50 m / s; for the MTD method that compensates for linear distance travel, the target has only a radial velocity of 5000 m / s; for the MTD method that compensates for distance travel and matches quadratic terms, the target has a radial velocity of 5000 m / s and a radial acceleration of 300 m / s²; for the GRFT method (i.e., the MTD method based on GRFT), the target has a radial velocity of 5000 m / s, a radial acceleration of 300 m / s², and a radial jerk of 2000 m / s³ (the radial jerk is set to a larger value for simulation verification, but this situation will not exist in reality).

[0037] Simulation Content: This invention presents a DSP engineering design system for a spaceborne radar near-field target detection and processing algorithm provided in the above embodiments. First, echo data is generated in MATLAB based on radar system parameters and target motion parameters. Then, the data is written to a DDR address for algorithm processing. Finally, the DSP processing results are compared with the simulation results obtained in MATLAB to verify the effectiveness of the DSP engineering design. Furthermore, in CCS, the total program execution time and memory configuration of each method are statistically analyzed based on the number of CPU clock cycles to verify the real-time performance and memory resource usage of the engineering design.

[0038] Results analysis: See details Figure 6 , Figure 7 , Figure 8 , Figure 9 , Figure 10 , Figure 11 , Figure 12 , Figure 13 and Figure 14 .

[0039] Specifically, Figure 6 This is the result obtained by following the processing flow of this invention, inputting radar system parameters and target parameters that do not involve range movement into MATLAB to generate echo signals, importing them into CCS, and running the standard MTD processing program that ignores range movement. At this point, the standard MTD method can be directly used for target accumulation, such as... Figure 6 As shown in Figure A, the target was focused at a Doppler center frequency of 500km and 0Hz, achieving a good focusing effect. According to... Figure 6In the equation B, the signal-to-noise ratio of the accumulated signal is approximately 33.17 dB. Ignoring the effects of discrete sampling, limitations on the number of pulses accumulated, and numerical accuracy, this result is very close to the theoretical value. It meets the performance requirements. Figure 6 C and Figure 6 The "D" in the figure compares the DSP algorithm implementation of this invention with the results obtained from direct simulation in MATLAB in the Doppler dimension and the distance dimension. It can be seen that the scheme of this invention is highly consistent with the simulation results of MATLAB, which further verifies the excellent performance of the engineering scheme. Figure 7 The result is obtained by running a processing program based on a one-dimensional filter bank method, following the processing flow of this invention, by inputting radar system parameters and target parameters that only involve linear range movement. Figure 7 In the case of A, the signal after pulse compression exhibits linear distance travel, requiring compensation for the echo envelope, such as... Figure 7 As shown in B, target signal accumulation is then performed, such as... Figure 7 As shown in C, a good focusing effect can be obtained. According to... Figure 7 In the D region, the signal-to-noise ratio of the accumulated signal is approximately 33.19 dB, which meets the performance requirements. Similarly, Figure 7 The letter E indicates that the DSP algorithm implementation scheme of this invention is highly consistent with the results obtained by direct simulation in MATLAB. Figure 8 Is it importing the same Figure 7 The echo signal used is the result of a processing program based on the first-order KT transform method. According to... Figure 8 In the case of A, the signal after pulse compression exhibits linear distance shift. This linear shift needs to be compensated for using a first-order KT transform based on the CZT fast transform, such as... Figure 8 As shown in B, target signal accumulation is then performed, such as... Figure 8 As shown in C, a good focusing effect can be obtained. According to... Figure 8 The signal-to-noise ratio of the signal after D accumulation is approximately 33.17 dB, which meets the performance requirements. Similarly, Figure 8 The letter E indicates that the DSP algorithm implementation scheme of this invention is highly consistent with the results obtained by direct simulation in MATLAB. Figure 9 This is the result obtained by running a processing program based on a two-dimensional filter bank method, following the processing flow of this invention, inputting radar system parameters and target parameters showing range movement and second-order phase spread. Figure 9 In A, the signal after pulse compression exhibits distance travel and Doppler frequency spread. First, compensate for the distance travel, as shown below. Figure 9 As shown in B, the echo signal still exhibits symmetrical Doppler frequency spread. In this case, second-order phase compensation is performed, followed by target signal accumulation, as shown in Figure B. Figure 9 As shown in C, a good focusing effect can be obtained. According to... Figure 9In the D region, the signal-to-noise ratio of the accumulated signal is approximately 33.12 dB, which meets the performance requirements. Similarly, Figure 9 E and Figure 9 The F in the figure indicates that the DSP algorithm implementation scheme of the present invention is highly consistent with the results obtained by direct simulation in MATLAB. Figure 10 Is it importing the same Figure 9 The echo signal used was obtained by running a processing program based on the first-order KT transform + matched quadratic term method. According to... Figure 10 In the A-wave pulse compression signal, distance travel and Doppler frequency spread are observed. Firstly, a first-order KT transform based on CZT fast transform is used to compensate for the distance travel. Figure 10 As shown in B, second-order phase compensation is performed, followed by target signal accumulation, as follows. Figure 10 As shown in C, a good focusing effect can be obtained. According to... Figure 10 In the D region, the signal-to-noise ratio of the accumulated signal is approximately 33.08 dB, which meets the performance requirements. Similarly, Figure 10 E and Figure 10 The F in the figure indicates that the DSP algorithm implementation scheme of the present invention is highly consistent with the results obtained by direct simulation in MATLAB. Figure 11 The result is obtained by running a GRFT-based processing program, which inputs radar system parameters and target parameters that have experienced range movement and second- and third-order phase spread, according to the processing flow of this invention. Figure 12 This is a comparison of the distance dimension data profile results between DSP processing and MATLAB processing results using the GRFT-based MTD method. According to... Figure 11 In the A-axis signal, distance travel and asymmetric Doppler frequency spread are observed. After compensating for the distance travel, as shown below... Figure 11 As shown in B, the echo still has asymmetric phase extension, which needs to be compensated for with second-order phase, as shown in Figure B. Figure 11 As shown in C, at this point only the third-order phase spread caused by the target's radial jerk remains. Finally, target signal accumulation is performed, as shown in... Figure 11 As shown in D, a good focusing effect can be obtained. According to... Figure 11 The signal-to-noise ratio (SNR) of the accumulated signal (E) is approximately 32.97 dB, which meets the performance requirements. Similarly, Figure 11 F and Figure 12 This indicates that the DSP algorithm implementation scheme of the present invention is highly consistent with the results obtained by direct simulation in MATLAB. Figure 13 This represents the total program execution time of each algorithm in this invention within CCS. Figure 13Statistical results show that the total runtime of the six methods in this invention is approximately 45.315ms, 99.428ms, 128.650ms, 153.762ms, 162.526ms, and 263.467ms, respectively, all meeting the real-time requirements for hypersonic target signal accumulation. In the existing technology "Research on Real-time Imaging Processing Technology for Space Targets by Spaceborne Radar", completing the Keystone transform and azimuth FFT for 256×1024 points requires 178ms, while this invention requires only 128.65ms under the same conditions, improving computational efficiency by approximately 27.7%. Figure 14 This refers to the storage resource usage information of the .map file output by the CCS compiler in this invention. , , The maximum available space is 512KB, 4MB, and 2GB respectively, while the storage resource consumption of each level in the C6678 of this invention is 362.10KB, 1.86MB, and 0.16GB respectively, with corresponding utilization rates of approximately 70.72%, 46.50%, and 8.00%. The memory space utilization rate is less than 75% for all levels. Therefore, the kernel operation still has a certain degree of scalability and can meet the needs of more scenarios. From the above results, it can be concluded that the DSP engineering design scheme of the spaceborne radar airborne target detection and processing algorithm described in this invention designs corresponding signal processing modules and algorithm implementation schemes for different targets, and further improves the processing efficiency of C6678 through various software and hardware optimizations. At the same time, it meets the engineering performance indicators and real-time processing requirements, and solves the shortcomings of insufficient motion compensation accuracy, incomplete signal processing, and insufficient optimization of DSP performance in existing technologies. It can effectively accumulate and detect airborne hypersonic targets.

[0040] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A DSP engineering design system for a spaceborne radar near-air target detection and processing algorithm, characterized in that, Implemented using a TIC66x series multi-core DSP, the system is used for: Receive input radar system parameters; Receive the input processing mode configuration instruction, and enter the specified processing mode according to the processing mode configuration instruction; After entering the specified processing mode, the stored raw echo data of the target is obtained, and according to the algorithm code corresponding to the specified processing mode, multiple cores in the multi-core DSP are used to perform collaborative parallel processing of the raw echo data to obtain the target detection result. The system includes six processing modes: The first processing mode determines the target detection result based on the MTD method and the target's raw echo data when the target has not moved distance. The second processing mode determines the target detection result based on a one-dimensional filter bank and the target's raw echo data when the target moves linearly. The third processing mode determines the target detection result based on a first-order KT transform and the target's raw echo data when the target moves linearly. The fourth processing mode determines the target detection result based on a two-dimensional filter bank and the target's raw echo data when the target moves linearly, exhibits distance curvature, and experiences Doppler frequency spread. The fifth processing mode determines the target detection result based on a first-order KT transform, a matched quadratic term method, and the target's raw echo data when the target moves linearly, exhibits distance curvature, and experiences Doppler frequency spread. The sixth processing mode determines the target detection result based on a frequency domain GRFT method, the MTD method, and the target's raw echo data when the target moves linearly, exhibits distance curvature, and experiences third-order Doppler frequency spread.

2. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, The multi-core DSP contains eight cores, Core0 to Core7. After entering the specified processing mode, each core acquires a portion of the raw echo data and processes it according to the algorithm corresponding to the specified processing mode to obtain the CFAR detection result. Then, any one of the eight cores calculates the target parameters using the CFAR detection results of the eight cores according to the algorithm corresponding to the specified processing mode to obtain the target detection result. The portion of raw echo data acquired by each core is one-eighth of the target's raw echo data.

3. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, The process of determining the target detection result based on a one-dimensional filter bank and the target's raw echo data when the target moves a linear distance includes: Acquire raw echo data and generate a conjugate reference signal; wherein the raw echo data is the echo data under the condition that the target moves a linear distance; Perform an FFT transform on each pulse of the acquired raw echo data to obtain pulse-compressed data, and multiply the pulse-compressed data by the conjugate reference signal to achieve matched filtering, thereby obtaining matched-filtered data; The matched-filtered data is subjected to ambiguity compensation, and the ambiguity-compensated data is subjected to first-order linear distance travel correction for each distance gate to obtain range-frequency domain-azimuth domain data; Perform IFFT transform on the range-frequency domain-azimuth domain data pulse by pulse to obtain range-azimuth time domain data; Perform a range-gate-by-range FFT transform on the range-azimuth time domain data to obtain range-Doppler domain data; CFAR detection is performed on the range-Doppler domain data with a range gate to obtain CFAR detection results; The target parameters are calculated based on the CFAR detection results to obtain the target detection results.

4. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, In the case where the target moves a linear distance, the target detection result is determined based on the first-order KT algorithm and the target's original echo data, including: Acquire raw echo data and generate a conjugate reference signal; wherein the raw echo data is the echo data under the condition that the target moves a linear distance; Perform an FFT transform on each pulse of the acquired raw echo data to obtain pulse-compressed data, and multiply the pulse-compressed data by the conjugate reference signal to achieve matched filtering, thereby obtaining matched-filtered data; The matched-filtered data is subjected to ambiguity compensation, and the ambiguity-compensated data is subjected to CZT-based Keystone transform on a range-gate-by-range basis to obtain range-frequency domain-azimuth domain data. Perform IFFT transform on the range-frequency domain-azimuth domain data pulse by pulse to obtain range-azimuth time domain data; Perform a range-gate-by-range FFT transform on the range-azimuth time domain data to obtain range-Doppler domain data; CFAR detection is performed on the range-Doppler domain data with a range gate to obtain CFAR detection results; The target parameters are calculated based on the CFAR detection results to obtain the target detection results.

5. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, The determination of target detection results based on a two-dimensional filter bank and the target's raw echo data, under conditions of linear range travel, range curvature, and Doppler frequency spread, includes: Acquire raw echo data and generate a conjugate reference signal; wherein, the raw echo data is the echo data under the condition that the target generates linear distance movement, distance curvature and Doppler frequency spread; Perform an FFT transform on each pulse of the acquired raw echo data to obtain pulse-compressed data, and multiply the pulse-compressed data by the conjugate reference signal to achieve matched filtering, thereby obtaining matched-filtered data; The matched-filtered data is subjected to ambiguity compensation, and the ambiguity-compensated data is subjected to first-order linear distance travel correction for each distance gate to obtain range-frequency domain-azimuth domain data; Perform IFFT transform on the range-frequency domain-azimuth domain data pulse by pulse to obtain range-azimuth time domain data; The range-azimuth time domain data is compensated for quadratic term coefficients by range gate-by-range gating to obtain echo data after quadratic phase matching; The echo data after the second phase matching is subjected to FFT transformation by range gate to obtain the echo data after azimuth FFT. CFAR detection is performed on the echo data after azimuth FFT by range gate to obtain CFAR detection results; The target parameters are calculated based on the CFAR detection results to obtain the target detection results.

6. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, The method for determining the target detection result based on the first-order KT transform, a matching quadratic term, and the target's original echo data when the target exhibits linear range movement, range curvature, and Doppler frequency spread includes: Acquire raw echo data and generate a conjugate reference signal; wherein, the raw echo data is the echo data under the condition that the target generates linear distance movement, distance curvature and Doppler frequency spread; Perform an FFT transform on each pulse of the acquired raw echo data to obtain pulse-compressed data, and multiply the pulse-compressed data by the conjugate reference signal to achieve matched filtering, thereby obtaining matched-filtered data; The matched-filtered data is subjected to ambiguity compensation, and the ambiguity-compensated data is subjected to CZT-based Keystone transform on a range-gate-by-range basis to obtain range-frequency domain-azimuth domain data. Perform IFFT transform on the range-frequency domain-azimuth domain data pulse by pulse to obtain range-azimuth time domain data; The range-azimuth time domain data is compensated for quadratic term coefficients by range gate-by-range gating to obtain echo data after quadratic phase matching; The echo data after the second phase matching is subjected to FFT transformation by range gate to obtain the echo data after azimuth FFT. CFAR detection is performed on the echo data after azimuth FFT by range gate to obtain CFAR detection results; The target parameters are calculated based on the CFAR detection results to obtain the target detection results.

7. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, The determination of target detection results based on the frequency domain GRFT method, MTD method, and the target's original echo data, in the case of target range movement and Doppler frequency spread including secondary and tertiary phases, includes: Acquire raw echo data and generate a conjugate reference signal; wherein, the raw echo data is the echo data under the condition that the target generates linear distance movement, distance curvature, and third-order Doppler frequency spread; Perform an FFT transform on each pulse of the acquired raw echo data to obtain pulse-compressed data, and multiply the pulse-compressed data by the conjugate reference signal to achieve matched filtering, thereby obtaining matched-filtered data; The matched-filtered data is subjected to distance curvature correction through each distance gate to obtain the distance curvature-corrected data. The distance-bending correction data is subjected to pulse-by-pulse IFFT transformation to obtain distance-azimuth time domain data; the distance-azimuth time domain data is in complex form. Calculate the amplitude and phase of the complex number, and take the remainder of 2π from the calculated phase to obtain a new phase. Multiply the new phase by the amplitude of the complex number, and then convert the multiplication result into float data to obtain the initially optimized range-azimuth time domain data. The phase of the initially optimized range-azimuth time domain data is defined using the double type. Then, the phase defined using the double type is modulo 2π to make the phase range between [-π, π]. The moduloed phase is then converted into float data to obtain the word length optimized range-azimuth time domain data. The quadratic term coefficients of the range-azimuth time domain data after word length optimization are compensated by a predetermined float type compensation function to obtain echo data after quadratic phase matching; The echo data after the second phase matching is compensated for the coefficients of the cubic term by a range gate to obtain the echo data after the third phase matching. Perform FFT transformation on the echo data after the three phase matches, range-gated, to obtain range-Doppler domain data; CFAR detection is performed on the range-Doppler domain data with a range gate to obtain CFAR detection results; The target parameters are calculated based on the CFAR detection results to obtain the target detection results.

8. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 1, characterized in that, When performing matrix transpose operations, each core sets the source address step size to the size of the element for each element in each row of the matrix to be transposed, and sets the destination address step size to the product of the row length and the element size. The transpose of an entire row of elements is completed in a single DMA operation. Similarly, for each element in each column of the matrix to be transposed, the source address step size is set to the product of the column length and the element size, and the destination address step size is set to the element size. The transpose of an entire column of elements is completed in a single DMA operation.

9. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 2, characterized in that, Each core corresponds to a semaphore, and the eight semaphores corresponding to the eight cores are different and independent of each other. The eight cores use the eight semaphores for multi-core synchronization. The eight semaphores are arbitrarily selected from the 64 independent hardware semaphores of the multi-core DSP. When the eight cores perform each multi-core operation task in each processing mode, each core requests its own semaphore through the semaphore request function. When its own semaphore is in an idle state, it marks its own semaphore as occupied and continuously checks whether the semaphores corresponding to the remaining cores are marked as occupied. When all eight cores detect that all eight semaphores are marked as occupied, it indicates that the eight cores have completed the current multi-core operation task. Afterward, each core releases its own semaphore through the semaphore release function to reset its own semaphore from occupied to idle. Each core continuously checks whether the semaphores corresponding to the remaining cores are reset to idle. When all eight cores detect that all eight semaphores are reset to idle, each core proceeds to the next multi-core operation task.

10. The DSP engineering design system for the spaceborne radar near-air target detection and processing algorithm according to claim 2, characterized in that, The on-chip memory L1D and L1P of the multi-core DSP are both 32K caches. The code corresponding to the six processing modes is stored in the on-chip memory L2, and L2 is set as SRAM. The raw echo data is stored in the off-chip memory DDR3.

Images