Strong real-time radar signal processing system

Abstract

The invention provides a strong real-time radar signal processing system, and the system comprises the steps: receiving strong real-time beam data and radar working parameters through a front-end processing module, processing the strong real-time beam data to obtain first beam data, and storing the first beam data and the radar working parameters in a CPU blade; the CPU blade receives the first beam data and the radar working parameters, and sends the first beam data and the radar working parameters to the GPU blade; the GPU blade receives the first beam data and the radar working parameters, processes the first beam data and the radar working parameters to obtain multiple groups of beam detection data, and sends the multiple groups of beam detection data to the CPU blade; the CPU blade receives and processes multiple groups of beam detection data to obtain a trace point report, and sends the trace point report to the target equipment, compared with a traditional FPGA + DSP processing architecture, the method has the advantages of being fast in development process, high in program transportability, low in cost and the like.

Description

Technical Field

[0001] The invention belongs to the technical field of radar signal processing, and in particular relates to a strong real-time radar signal processing system. Background Art

[0002] Field radars are highly maneuverable and have the ability to be deployed quickly. They need to work around the clock and have the ability to automatically track targets and transmit data. They need to achieve high-precision positioning and complete related processing tasks. The characteristics of strong real-time radar signal processing are strong real-time, parameter agility, high precision, and complex working modes. Traditional signal processing is mostly completed using the FPGA+DSP architecture. Generally, the FPGA part completes related data preparation and signal processing detection functions, and then the DSP completes the report processing. Although FPGA has parallel processing capabilities and low latency characteristics, it requires highly customized development features and is difficult to develop, resulting in a long development cycle and difficulty in transplantation. In addition, FPGA has high power consumption and high cost, which also brings certain limitations to system design. Although FPGA can be reprogrammed, each reprogramming takes a certain amount of time, resulting in poor reconfigurability. Summary of the invention

[0003] In view of this, the embodiments of the present application propose a strong real-time radar signal processing system, which aims to significantly shorten the research and development cycle, improve software reusability, and have high scalability while meeting the real-time processing requirements.

[0004] To achieve the above-mentioned purpose, an embodiment of the present application provides a strong real-time radar signal processing system, comprising a front-end processing module and a back-end processing module respectively located on two strong real-time radar signal processing nodes and communicatively connected, the back-end processing module having a CPU blade and a GPU blade communicatively connected, the CPU blade being connected to the front-end processing module: wherein the front-end processing module is used to receive strong real-time beam data and the radar operating parameters belonging to the same distance point, process the strong real-time beam data based on a digital beamforming process algorithm to obtain first beam data, and use a coherent processing interval as a beat to store the first beam data and the radar operating parameters arranged according to a pulse repetition interval in a DMA buffer area of ​​the CPU blade; the CPU blade is used to use a preset data receiving thread to receive each of the pulse repetitions The first beam data and the radar operating parameters in each pulse repetition interval are received by the GPU blade, and the first beam data and the radar operating parameters in each pulse repetition interval are sent to the GPU blade by using a preset detection processing thread; the GPU blade is used to receive the first beam data and the radar operating parameters in each pulse repetition interval, and sequentially process the first beam data and the radar operating parameters in each pulse repetition interval according to the preset detection processing thread, correspondingly obtain multiple groups of beam detection data, and send the multiple groups of beam detection data to the CPU blade according to the preset detection processing thread; the CPU blade is also used to receive the multiple groups of beam detection data, process the multiple groups of beam detection data to obtain a dot trace report, and send the dot trace report to the target device by using a preset report packaging thread.

[0005] Optionally, the front-end processing module is also used to generate a doorbell value after storing each frame of the first beam data and / or each frame of the radar operating parameters in the DMA buffer area of ​​the CPU blade, and store each doorbell value in the DMA buffer area of ​​the CPU blade to obtain a doorbell queue.

[0006] Optionally, the CPU blade is also used to use the preset data receiving thread to periodically monitor the doorbell queue. If the doorbell queue is not empty, the storage address of the first beam data and the radar operating parameters in the DMA buffer area is queried using the doorbell value as an index, and the first beam data and the radar operating parameters corresponding to the storage address are sent to the GPU blade.

[0007] Optionally, the GPU blade is also used to parse the first beam data within each pulse repetition interval according to the preset detection processing thread and the radar operating parameters to obtain multiple distance point data, and fill the multiple distance point data with zeros to a power of 2 to obtain second beam data.

[0008] Optionally, the GPU blade is also used to: arrange the second beam data in each pulse repetition interval according to matrix transposition to obtain third beam data; perform frequency domain pulse compression processing on the third beam data based on fast Fourier transform to obtain fourth beam data, wherein the fourth beam data includes shadowless beam data; select the fast Fourier transform or FIR filter according to the radar working parameters to perform moving target detection processing on the fourth beam data to obtain IQ data, perform modulus processing on the IQ data to obtain MTD plane modulus data; perform clutter map processing and constant false alarm rate processing on the fourth beam data respectively to obtain a clutter map and a CFAR matrix correspondingly, and obtain an index vector based on the CFAR matrix; determine each group of the beam detection data according to the IQ data, the MTD plane modulus data, the clutter map, the CFAR matrix and the index vector in each pulse repetition interval; and send each group of the beam detection data to the CPU blade.

[0009] Optionally, the GPU blade is also used to obtain multiple beam numbers based on the radar operating parameters, arrange the second beam data based on the same beam number to obtain multiple beam data groups, and arrange each beam data group based on the pulse repetition interval to obtain third beam data.

[0010] Optionally, the GPU blade is further configured to determine an index vector according to a threshold point in the CFAR matrix.

[0011] Optionally, the CPU blade is further configured to retrieve the CFAR matrix based on the retrieval vector to obtain each threshold-crossing point, and determine point trace data according to a comparison result between each threshold-crossing point and the blanking beam data.

[0012] Optionally, the CPU blade also has a preset report packaging thread, and the preset report packaging thread is used to respond to a search mode instruction, perform condensation processing and angle measurement processing on the point trace data to obtain a search report; or respond to a confirmation mode instruction, wherein the confirmation mode instruction carries a first command parameter, and confirms the confirmation gate data in the point trace data according to the command parameter to obtain a confirmation report; or respond to a tracking mode instruction, wherein the tracking mode instruction carries a second command parameter, and tracks the tracking gate data in the point trace data according to the second command parameter to obtain a tracking report.

[0013] Optionally, the CPU blade further comprises an SRIO thread and a network sending thread, and the SRIO thread and the network sending thread are used to send the search report, the confirmation report and the tracking report to the target device through a RapidIO interface and a network interface respectively.

[0014] The strong real-time radar signal processing system proposed in the embodiment of the present application is used to receive strong real-time beam data and radar operating parameters belonging to the same distance point through a front-end processing module, and processes the strong real-time beam data based on a digital beamforming process algorithm to obtain first beam data, and uses a coherent processing interval as a beat to store the first beam data and radar operating parameters arranged according to the pulse repetition interval in the DMA buffer area of ​​the CPU blade; the CPU blade is used to use a preset data receiving thread to receive the first beam data and radar operating parameters in each pulse repetition interval, and use a preset detection processing thread to send the first beam data and radar operating parameters in each pulse repetition interval to the GPU blade; the GPU blade is used to receive the first beam data and radar operating parameters in each pulse repetition interval. The invention relates to a method for processing a plurality of beam data and radar operating parameters, and processes the first beam data and radar operating parameters in each pulse repetition interval in sequence according to a preset detection processing thread, and obtains multiple groups of beam detection data, and sends multiple groups of beam detection data to the CPU blade according to the preset detection processing thread; the CPU blade is also used to receive multiple groups of beam detection data, and processes multiple groups of beam detection data to obtain a trace report, and uses a preset report packaging thread to send the trace report to the target device. The present invention makes a unified standard for the CPU blade and the GPU blade, unifies the input and output interfaces, and all processing is parameterized. According to the needs of the project, the sub-modules of the CPU blade and the GPU blade can be quickly added or removed at any time to realize the rapid expansion of program functions. For the needs of new projects, most of the necessary processing modules can be directly reused. Through standard modular design and parameterized processing, the past highly customized design ideas are abandoned, and the new project back-end program can be quickly built and debugged, shortening the research and development cycle. Therefore, compared with the traditional FPGA+DSP processing architecture, this method has the advantages of fast development process, strong program portability, and low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a structural block diagram of a strong real-time radar signal processing system provided in one embodiment of the present application;

[0016] Figure 2 is a processing thread relationship diagram of a strong real-time radar signal processing system provided in one embodiment of the present application;

[0017] Figure 3 It is a flowchart of strong real-time radar signal processing of a strong real-time radar signal processing system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. However, it will be appreciated by those skilled in the art that in the present application, many technical details are proposed in order to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The division of the following embodiments is for the convenience of description, and the specific implementation of the present application should not be construed as any limitation, and the various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0019] An embodiment of the present application proposes a strong real-time radar signal processing system, the radar signal processing system 10 includes a front-end processing module 101 and a back-end processing module 104 respectively located on two strong real-time radar signal processing nodes and communicatively connected, wherein the back-end processing module 104 includes a CPU blade 102 and a GPU blade 103, and the CPU blade 102 is communicatively connected to the front-end processing module 101: wherein the front-end processing module 101 is used to receive strong real-time beam data and radar operating parameters belonging to the same distance point, process the strong real-time beam data based on the digital beamforming process algorithm to obtain first beam data, and use the coherent processing interval as the beat to store the first beam data and radar operating parameters arranged according to the pulse repetition interval in the DMA buffer area of ​​the CPU blade 102; the CPU blade Blade 102 is used to receive the first beam data and radar operating parameters in each pulse repetition interval by using a preset data receiving thread, and send the first beam data and radar operating parameters in each pulse repetition interval to GPU blade 103 by using a preset detection processing thread; GPU blade 103 is used to receive the first beam data and radar operating parameters in each pulse repetition interval, and sequentially process the first beam data and radar operating parameters in each pulse repetition interval according to the preset detection processing thread, correspondingly obtain multiple groups of beam detection data, and send the multiple groups of beam detection data to CPU blade 102 according to the preset detection processing thread; CPU blade 102 is also used to receive multiple groups of beam detection data, process the multiple groups of beam detection data to obtain a dot trace report, and send the dot trace report to the target device by using a preset report packaging thread.

[0020] In summary, in this embodiment, the front-end processing module 101 is used to receive strong real-time beam data and radar operating parameters belonging to the same distance point, and the strong real-time beam data is processed based on the digital beamforming process algorithm to obtain the first beam data, and the coherent processing interval is used as the beat to store the first beam data and radar operating parameters arranged according to the pulse repetition interval in the DMA buffer area of ​​the CPU blade 102; the CPU blade 102 is used to use a preset data receiving thread to receive the first beam data and radar operating parameters in each pulse repetition interval, and use a preset detection processing thread to send the first beam data and radar operating parameters in each pulse repetition interval to the GPU blade 103; the GPU blade 103 is used to receive the first beam data and radar operating parameters in each pulse repetition interval As parameters, and according to the preset detection processing thread, the first beam data and radar working parameters in each pulse repetition interval are processed in sequence, and multiple groups of beam detection data are obtained accordingly, and multiple groups of beam detection data are sent to the CPU blade 102 according to the preset detection processing thread; the CPU blade 102 is also used to receive multiple groups of beam detection data, and process multiple groups of beam detection data to obtain point trace reports, and use the preset report packaging thread to send the point trace report to the target device. The present invention makes a unified standard for the CPU blade 102 and the GPU blade 103, unifies the input and output interfaces, and all processing is parameterized. According to the needs of the project, the sub-modules of the CPU blade 102 and the GPU blade 103 can be quickly added or removed at any time to realize the rapid expansion of program functions. For the needs of new projects, most of the necessary processing modules can be directly reused. Through standard modular design and parameterized processing, the past highly customized design ideas are abandoned, and the new project back-end program can be quickly built and debugged, shortening the research and development cycle. Therefore, compared with the traditional FPGA+DSP processing architecture, this method has the advantages of fast development process, strong program portability, and low cost.

[0021] In an embodiment of the present application, the front-end processing module 101 also includes a doorbell generating submodule and a doorbell sending submodule, wherein the doorbell generating submodule is used to generate a doorbell value after storing each frame of first beam data and / or each frame of radar operating parameters in the DMA buffer area of ​​the CPU blade 102, and the doorbell sending submodule is used to store each doorbell value in the DMA buffer area of ​​the CPU blade 102 to obtain a doorbell queue.

[0022] In the specific implementation process, the front-end processing module 101 is also used to generate a doorbell value after storing each frame of the first beam data and / or each frame of the radar operating parameters in the DMA buffer area of ​​the CPU blade 102, and the front-end processing module 101 and the back-end server use RapidIO to realize the transmission of the first beam data, wherein the back-end server is the CPU blade 102. The front-end processing module 101 uses CPI (Coherent Processing Interval) as the beat to write the strong real-time beam data that has completed DBF (Digital Beam Formin, digital beam forming), that is, the first beam data into the cache space allocated by the CPU. When receiving data, the CPU blade 102 uses the RapidIO interface to meet the needs of fast downlink of large amounts of data, and at the same time designs an 8-level cache queue to cache the first beam data and radar operating parameters. The CPU blade 102 receives data through RapidIO with CPI as the beat. This is because the interrupt response time of a non-real-time system such as Linux cannot be met when receiving data with PRI as the beat.

[0023] It should be noted that digital beamforming is a key technology. It forms a receiving beam with specific directivity and beam shape by digitizing and weighting the signals received by multiple sensors (such as antenna array elements). This process involves complex signal processing algorithms, including adjusting the phase and amplitude of the signal to enhance the signal in the desired direction and suppress interference signals in other directions. For example, in phased array radars, digital beamforming can flexibly control the direction of the beam and quickly scan different directions to search and track targets.

[0024] Completing DBF means that the digital beamforming processing of the original received signal has been completed, and the beam data after beamforming processing has been obtained. These data are ready to be written into the cache space allocated by the CPU by the front end with CPI (coherent processing interval) as the beat, and can be used for further analysis, target detection and other processing.

[0025] After that, the data is ready to be written into the cache space allocated by the CPU by the front end with CPI (Coherent Processing Interval) as the beat, and is written in units of frames. After storing each frame of the first beam data and radar operating parameters in the DMA buffer area of ​​the CPU blade 102, the front-end processing module 101 generates a doorbell value, and stores each doorbell value in the DMA buffer area of ​​the CPU blade 102 to obtain a doorbell queue.

[0026] In an embodiment of the present application, the CPU blade 102 also includes a doorbell monitoring submodule, which is used to periodically monitor the doorbell queue using a preset data receiving thread. If the doorbell queue is not empty, the storage address of the first beam data and radar operating parameters in the DMA buffer area is queried using the doorbell value as an index, and the first beam data and radar operating parameters corresponding to the storage address are sent to the GPU blade 103.

[0027] like Figure 3 As shown, it should be pointed out that after the strong real-time radar signal processing program is running, the main function of the CPU blade 102 completes the relevant initialization of the CPU and GPU, and then starts the data receiving thread. The data receiving thread continuously monitors the doorbell queue. If the doorbell queue is not empty, the detection processing thread is started to complete the CPI level parameter parsing and report the data to the target device, and then returns to continue monitoring the doorbell queue.

[0028] In other words, the main task of the data receiving thread is to constantly check the status of the doorbell receiving queue. This receiving queue is a buffer specifically used to store doorbell-related data. It's like a doorman who keeps watching the mailbox (the mailbox here is analogous to the receiving queue) to see if there are new letters (data) arriving. When the data receiving thread finds that there is data in the doorbell receiving queue (that is, "the doorbell is not empty"), it starts the processing flow. This means that once new doorbell-related information arrives in the queue, the system will trigger subsequent operations to ensure that the data will not be missed and can be processed in time.

[0029] It should be pointed out that CPI is an important concept in the field of signal processing. Analyzing CPI-level parameters involves analyzing various characteristics of the signal within the coherent processing interval, such as the phase and amplitude of the signal, and extracting valuable parameters from it. The above steps can better analyze the characteristics of the target echo signal to achieve functions such as target detection and tracking.

[0030] refer to Figure 2 In an embodiment of the present application, the GPU blade 103 also includes a zero-filling module, which is used to parse the first beam data in each pulse repetition interval according to a preset detection processing thread and radar operating parameters to obtain multiple distance point data, and fill the multiple distance point data with zeros to a power of 2 to obtain second beam data.

[0031] It should be noted that the first beam data is transferred to the GPU (graphics processing unit) memory after previous processing (such as acquisition from the doorbell receiving queue, CPI level parameter parsing, etc.). GPU memory, as a dedicated storage area of ​​the GPU, can provide fast data access for subsequent high-speed data processing. This step is similar to moving raw materials to the warehouse (video memory) of the production workshop, ready for the next step of processing.

[0032] Specifically, each pulse repetition interval contains a certain number of distance point data, which are related to the transmission and reception of radar pulses. During a pulse repetition interval, the radar transmits pulses and receives signals reflected by the target, which are processed to form distance point data. In order to meet the requirements of subsequent processing, the number of distance points in each pulse repetition interval needs to be adjusted to the closest power of 2 by padding with zeros to obtain the second beam data. For example, if the original number of distance points of a PRI is 100, and the closest power of 2 is 128, 28 zeros need to be added to the data. The power of 2 as the data length can improve the computational efficiency and simplify the algorithm implementation in certain calculations (such as fast Fourier transform, etc.). Doing so can make subsequent calculations on the GPU more efficient and convenient, and help to better utilize the parallel computing capabilities of the GPU to process the second beam data.

[0033] Continue to refer Figure 2 In an embodiment of the present application, the GPU blade 103 further includes a beam arrangement submodule, a pulse compression submodule, a first data processing module, a second data processing module, a detection data determination module, and a sending module, wherein the beam arrangement submodule is used to arrange the second beam data in each pulse repetition interval according to matrix transposition to obtain third beam data;

[0034] Specifically, the GPU blade 103 can rearrange the second beam data after zero padding is completed. The second beam data is first arranged by beam number, and then the data is arranged at each pulse repetition interval. The second beam data arranged by beam number obtains the third beam data.

[0035] Continue to refer Figure 2 In a specific implementation process, the second beam data stored in a matrix form may be rearranged by matrix transposition.

[0036] The pulse compression submodule is used to perform frequency domain pulse compression processing on the third beam data based on fast Fourier transform to obtain fourth beam data, wherein the fourth beam data includes the shadowless beam data;

[0037] After the rearrangement is completed, pulse compression processing is performed. The GPU blade 103 can use FFT and IFFT to perform frequency domain pulse compression on the third beam data to obtain fourth beam data.

[0038] It should be noted that pulse compression is an important technology. Radar needs to emit wide pulses to obtain sufficient energy when detecting distant targets, but wide pulses have limitations in range resolution. Pulse compression technology can obtain the range resolution of narrow pulses at the receiving end through methods such as matched filtering while emitting wide pulses, thereby improving the radar's ability to resolve target distances.

[0039] The first data processing module is used to select a fast Fourier transform or an FIR filter according to the radar working parameters to perform moving target detection processing on the fourth beam data to obtain IQ data, and perform modulus processing on the IQ data to obtain MTD plane modulus data;

[0040] Specifically, the GPU blade 103 can perform MTD processing after completing the pulse compression processing. During the MTD processing, the GPU blade 103 can select to use the FFT method or the FIR filter method to process the fourth beam data according to the radar working parameters to obtain IQ data, and perform modulus processing on the IQ data to obtain MTD plane modulus data.

[0041] It should be pointed out that MTD (Moving Target Detection) is a key link in radar signal processing for distinguishing moving targets from fixed clutter. Clutter refers to other unnecessary echoes received by the radar except for the target reflection signal, such as the signals reflected from the ground and sea surface. The echo signals of moving targets and fixed clutter have differences in characteristics such as spectrum, and MTD uses these differences to detect moving targets.

[0042] The second data processing module is used to perform clutter map processing and constant false alarm rate processing on the fourth beam data, respectively, to obtain a clutter map and a CFAR matrix accordingly, and to obtain an index vector based on the CFAR matrix;

[0043] Specifically, after completing the pulse compression, the GPU blade 103 may perform clutter map processing and CFAR processing to obtain a clutter map and a CFAR matrix, and obtain an index vector based on the CFAR matrix.

[0044] It should be noted that clutter is the echo signal generated by the reflection of the radar's surrounding environment (such as the ground, sea surface, buildings, etc.), which will interfere with the detection of target signals. Clutter map is a way of representing information used to describe the clutter environment. It forms a "map" of clutter intensity, distribution and other characteristics by performing long-term or short-term statistics and analysis on the clutter in the radar receiving area. Clutter map processing is performed after pulse compression processing, mainly using clutter map to identify and suppress clutter. By comparing the currently received signal with the information in the clutter map, it can be determined which parts are clutter and which parts may be target signals. For example, when a ground radar monitors an aerial target, the clutter reflected by the ground will be recorded accordingly in the clutter map. When a new signal comes in, if its characteristics match the ground clutter characteristics in the clutter map, it can be determined as clutter and suppressed, thereby improving the accuracy of target detection. Clutter map processing can adopt different update and matching strategies according to different radar working modes and environments to adapt to the dynamic changes of clutter.

[0045] In addition, in the process of radar detecting targets, false alarm refers to the situation where clutter or noise is misjudged as a target signal. A high false alarm rate will cause the radar system to frequently issue false alarms, reducing the reliability of the system. CFAR processing is a technology used to control the false alarm rate. It ensures that the false alarm probability when the radar detects a target remains at a constant and acceptable level under different clutter and noise environments. There are many methods for CFAR processing, such as unit average CFAR, ordered statistical CFAR, etc. Taking unit average CFAR as an example, it will set some reference units around the detection unit and set a suitable detection threshold by estimating the clutter or noise power in the reference unit. Only when the signal power in the detection unit exceeds this threshold is it determined to be a target signal. In this way, even if the clutter or noise intensity changes, the false alarm rate can be kept constant because the threshold is dynamically adjusted according to the surrounding environment. Performing CFAR processing on the signal after pulse compression can further improve the quality of target detection, reduce false alarms, and enable the radar to more accurately detect the real target in a complex environment.

[0046] The detection data determination module is used to determine each group of beam detection data according to the IQ data, MTD plane modulus data, clutter map, CFAR matrix and index vector in each pulse repetition interval;

[0047] The sending module is used to send each group of beam detection data to the CPU blade 102.

[0048] In an embodiment of the present application, the beam arrangement submodule is further configured to determine an index vector according to a threshold-crossing point in the CFAR matrix.

[0049] Specifically, the second data processing module is further used to retrieve the CFAR matrix based on the retrieval vector to obtain each threshold point, and determine the point trace data according to the comparison result of each threshold point and the blanking beam data.

[0050] In the specific execution process, in order to speed up the retrieval of points that pass the threshold in the CFAR matrix on the CPU, an index vector of the distance unit number where the threshold is located is generated on the GPU blade 103. Based on this vector, when searching, only the data of the Doppler channel at the distance unit number that passes the threshold needs to be retrieved, without searching the entire matrix, which reduces the number of loops and greatly speeds up the retrieval.

[0051] refer to Figure 2 and Figure 3In an embodiment of the present application, the CPU blade also includes a packing submodule, which has a preset report packing thread. The preset report packing thread is used to respond to a search mode instruction, perform condensation processing and angle measurement processing on the point trace data to obtain a search report; or respond to a confirmation mode instruction, wherein the confirmation mode instruction carries a first command parameter, and confirms the confirmation gate data in the point trace data according to the command parameter to obtain a confirmation report; or respond to a tracking mode instruction, wherein the tracking mode instruction carries a second command parameter, and tracks the tracking gate data in the point trace data according to the second command parameter to obtain a tracking report.

[0052] Specifically, after completing all GPU processing links, the modulus data, IQ data, clutter map, CFAR matrix and retrieval vector of the MTD plane are sent back to the CPU blade 102, and the CPU blade 102 performs point trace screening processing in combination with the shadowless beam to obtain beam detection data. The CPU blade 102 generates a response report according to the instruction received by the report packaging thread. The preset report packaging thread is used to respond to the search mode instruction, perform condensation processing and angle measurement processing on the point trace data to obtain a search report; or respond to the confirmation mode instruction, wherein the confirmation mode instruction carries a first command parameter, and confirms the confirmation gate data in the point trace data according to the command parameter to obtain a confirmation report; or respond to the tracking mode instruction, wherein the tracking mode instruction carries a second command parameter, and tracks the tracking gate data in the point trace data according to the second command parameter to obtain a tracking report.

[0053] The CPU blade also includes a data sending submodule, which has an SRIO thread and a network sending thread. The SRIO thread and the network sending thread are used to send a search report, a confirmation report and a tracking report to the target device through the RapidIO interface and the network interface respectively.

[0054] The present invention provides a specific embodiment below, which is implemented by using the CPU and GPU in a radar digital signal processing chassis. The main boards in the digital signal processing chassis are a CPU of Feiteng Company and a GPU of Tianshu Zhixin Company. The CPU is Feiteng D2000, which adopts ARM architecture, 64 bits, and has 8 cores and 8 threads. The GPU model is Tianshu Zhikai 100.

[0055] As attached Figure 2As shown in the figure, it is a relationship diagram of the processing threads, in which the main function performs initialization work, completes the allocation of memory space, loads and copies coefficients, and initializes threads. After completing the initialization work, the data receiving thread is started. The data receiving thread continues to run and monitors whether the data receiving queue is not empty. When the data receiving queue is not empty and the signal that the previous frame of data has been processed is received, the data is moved, the parameters are parsed, and the detection processing thread is started. After the detection is completed, the target report is packaged and sent to the report sending queue, which can be sent through RapidIO or the network. At the same time, the signal of processing completion is returned to start the next frame of data processing. The sending thread continuously monitors the sending queue and starts sending reports when it detects that the sending queue is not empty. Through the data receiving and report sending queues, the processing and data transmission speeds are matched.

[0056] like Figure 3 As shown, it is a processing flow chart of the present invention, wherein initialization work, data reception and parameter analysis are completed in the CPU, and then the beam data is copied to the video memory, and zero padding, rearrangement, pulse pressure, MTD, clutter map, CFAR, target distance index and other detection work are performed, and then the data is sent back to the CPU for point trace processing and report packaging and sending processing.

[0057] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A strong real-time radar signal processing system, characterized in that: It includes a front-end processing module and a back-end processing module respectively located on two strong real-time radar signal processing nodes and communicatively connected, wherein the back-end processing module has a CPU blade and a GPU blade communicatively connected, and the CPU blade is connected to the front-end processing module: The front-end processing module is used to receive strong real-time beam data and radar operating parameters belonging to the same distance point, process the strong real-time beam data based on the digital beamforming process algorithm to obtain first beam data, and use the coherent processing interval as the beat to store the first beam data and the radar operating parameters arranged according to the pulse repetition interval in the DMA buffer area of ​​the CPU blade; The CPU blade is used to receive the first beam data and the radar operating parameters in each pulse repetition interval by using a preset data receiving thread, and send the first beam data and the radar operating parameters in each pulse repetition interval to the GPU blade by using a preset detection processing thread; The GPU blade is used to receive the first beam data and the radar operating parameters in each pulse repetition interval, and process the first beam data and the radar operating parameters in each pulse repetition interval in sequence according to the preset detection processing thread, correspondingly obtain multiple groups of beam detection data, and send the multiple groups of beam detection data to the CPU blade according to the preset detection processing thread; The CPU blade is also used to receive multiple groups of the beam detection data, process the multiple groups of the beam detection data to obtain a point trace report, and send the point trace report to the target device using a preset report packaging thread.

2. The strong real-time radar signal processing system according to claim 1, characterized in that: The front-end processing module also includes: A doorbell generation submodule, configured to generate a doorbell value after storing each frame of the first beam data and the radar operating parameters in the DMA buffer area of ​​the CPU blade; The doorbell sending submodule is used to store each doorbell value in the DMA buffer area of ​​the CPU blade to obtain a doorbell queue.

3. The strong real-time radar signal processing system according to claim 2, characterized in that: The CPU blade also includes: The doorbell monitoring submodule is used to periodically monitor the doorbell queue using the preset data receiving thread. If the doorbell queue is not empty, the storage address of the first beam data and the radar operating parameters in the DMA buffer area is queried using the doorbell value as an index, and the first beam data and the radar operating parameters corresponding to the storage address are sent to the GPU blade.

4. The strong real-time radar signal processing system according to claim 1, characterized in that: The GPU blade further includes: The zero-filling module is used to parse the first beam data in each pulse repetition interval according to the preset detection processing thread and the radar working parameters to obtain a plurality of distance point data, and fill the plurality of distance point data with zeros to a power of 2 to obtain second beam data.

5. The strong real-time radar signal processing system according to claim 4 is characterized in that: The GPU blade further includes: A beam arrangement submodule, configured to arrange the second beam data in each pulse repetition interval according to matrix transposition to obtain third beam data; A pulse compression submodule, configured to perform frequency domain pulse compression processing on the third beam data based on a fast Fourier transform to obtain fourth beam data, wherein the fourth beam data includes an invisible beam data; A first data processing module is used to select the fast Fourier transform or the FIR filter according to the radar working parameters to perform moving target detection processing on the fourth beam data to obtain IQ data, and perform modulus processing on the IQ data to obtain MTD plane modulus data; A second data processing module is used to perform clutter map processing and constant false alarm rate processing on the fourth beam data, respectively, to obtain a clutter map and a CFAR matrix accordingly, and to obtain an index vector based on the CFAR matrix; A detection data determination module, configured to determine each group of the beam detection data according to the IQ data within each pulse repetition interval, the MTD plane modulus data, the clutter map, the CFAR matrix and the index vector; A sending module is used to send each group of beam detection data to the CPU blade.

6. The strong real-time radar signal processing system according to claim 5, characterized in that: The beam arrangement submodule is further used to obtain multiple beam numbers based on the radar operating parameters, arrange the second beam data based on the same beam number to obtain multiple beam data groups, and arrange each beam data group based on the pulse repetition interval to obtain third beam data.

7. The strong real-time radar signal processing system according to claim 5, characterized in that: The second data processing module is further used to determine an index vector according to a threshold point in the CFAR matrix.

8. The strong real-time radar signal processing system according to claim 5, characterized in that: The second data processing module is further used to retrieve the CFAR matrix based on the retrieval vector to obtain each threshold point, and determine the point trace data according to the comparison result of each threshold point and the blanking beam data.

9. The strong real-time radar signal processing system according to claim 1, characterized in that: The CPU blade also includes: A packaging submodule, comprising a report packaging thread, wherein the report packaging thread is used to perform agglomeration processing and angle measurement processing on the point trace data in response to a search mode instruction to obtain a search report; or in response to a confirmation mode instruction, wherein the confirmation mode instruction carries a first command parameter, confirming the confirmation gate data in the point trace data according to the command parameter to obtain a confirmation report; Or in response to a tracking mode instruction, wherein the tracking mode instruction carries a second command parameter, the tracking gate data in the point trace data is tracked according to the second command parameter to obtain a tracking report.

10. The strong real-time radar signal processing system according to claim 9, characterized in that: The CPU blade also includes: The data sending submodule comprises an SRIO thread and a network sending thread, wherein the SRIO thread and the network sending thread are used to send the search report, the confirmation report and the tracking report to the target device through the RapidIO interface and the network interface respectively.

Images