FT-M6678 half-precision access airborne SAR imaging method and device
By adopting semi-precision access and OpenMP multi-core parallel processing methods on FT-M6678, the problem of huge data volume in large-scene SAR imaging is solved, the hardware resource requirements are reduced, and efficient SAR imaging effect and processing time are achieved.
Patent Information
- Application Number
- CN202411316887.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2024-12-27
AI Technical Summary
In SAR imaging with large scenarios, large sampling rates and large wide sampling rates, the amount of data required by signal processing software is huge, resulting in a very large demand for hardware resources, making it difficult to effectively utilize the processing capabilities of FT-M6678.
The on-board SAR imaging method adopts the FT-M6678 half-accurate access, and uses distance FFT and azimuth interpolation processing to convert the data into a half-accurate format, and stores and processes it in OpenMP multi-core parallel processing to reduce storage Requirements, improve processing efficiency.
The imaging storage requirements and hardware scale are reduced, imaging effects and processing time similar to single-precision access are achieved, and the scale of hardware resources is reduced. SAR imaging processing is completed by only one piece of FT-M6678.
Smart Images
Figure CN119199856A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of airborne synthetic aperture radar imaging, and in particular to an airborne SAR imaging method and device with FT-M6678 half-precision access. Background Art
[0002] Currently, multi-core digital signal processors (DSPs) have become the mainstream processors for radar signal processing. One of the representatives is the FT-M6678, which integrates 8 C66x cores and has excellent computing and storage performance. The OpenMP parallel programming protocol provides a multi-core programming solution to help developers automatically complete master-slave core communication and parallel processing allocation. In order to fully utilize the multi-core processing power of the FT-M6678 and maximize the multi-core computing power, the application of OpenMP is implemented in the FT-M6678.
[0003] At present, drones are widely used in the military field, and drone radar has become the fastest-growing and most active branch of modern military radar. Among them, the use of airborne radar for SAR imaging is a very important reconnaissance means. With the changes in the needs of military reconnaissance, the requirements for SAR imaging indicators in practical applications are getting higher and higher, and the requirements for imaging scenes are getting larger and larger, so the computing power and storage requirements of signal processing software are getting higher and higher. When airborne synthetic aperture radar performs long-distance, high-resolution, and large-scene imaging, the polar coordinate format PFA algorithm is generally used. For large-scene, large-sampling-rate, and wide-width SAR imaging, and the storage requirement is the core contradiction, for example, when the radar system requires the SAR imaging width to be 65532 in the range direction and 32760 in the azimuth direction, the azimuth interpolation output data is a single-precision complex number of 65532*32760, and no more than 16GB of data needs to be stored. According to the use of 1.6GB of external memory per FT-M6678, 10 FT-M6678s are required to store. In this way, the scale of hardware resources required to complete the entire SAR imaging processing is very large. Summary of the invention
[0004] In order to solve the above problems, the present application provides an FT-M6678 half-precision access airborne SAR imaging method and device to solve the problem of huge amount of data access required by signal processing software under large scene, large sampling rate and wide bandwidth SAR imaging.
[0005] The first aspect of the present application provides an airborne SAR imaging method for FT-M6678 half-precision access, mainly comprising:
[0006] Step S1, performing range FFT processing on the received echo data, multiplying it by a given reference function and a motion compensation factor, and then performing range IFFT processing;
[0007] Step S2, intercepting data of a set length at the echo data center, performing range FFT processing and range interpolation, converting the processing results into a half-precision format to form SAR half-precision data, and storing them in an external memory;
[0008] Step S3, for the SAR half-precision data of the entire large frame accumulated in the external memory, obtain the half-precision data of the azimuth line length in sections, convert it into single-precision data, perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0009] Step S4, obtaining the half-precision data of the range gate length from the external memory, converting it into single-precision data, and then performing range IFFT processing, converting the processed single-precision data into half-precision data and storing it in the external memory;
[0010] Step S5, obtaining half-precision data of the azimuth line length from the external memory, converting it into single-precision data, and then performing whole-block focusing processing, and calculating the single-precision real power image after the whole-block focusing;
[0011] Step S6: geometrically correct and quantize the single-precision real power image to obtain a SAR grayscale image.
[0012] Preferably, step S5 further comprises:
[0013] Step S51, obtaining half-precision data of the azimuth line length from the external memory in sections, converting it into single-precision data and performing contrast processing;
[0014] Step S52, performing focus error iterative processing on the single-precision data after contrast processing to obtain error data;
[0015] Step S53, re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
[0016] Preferably, in step S3 and step S5, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-core, and in step S4, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-core.
[0017] Preferably, in step S3 to step S5, the single-precision data and the processing of the single-precision data use the exclusive L2 space in the high-bandwidth core of the FT-M6678 and the multi-core shared space SL2 as intermediate cache.
[0018] The second aspect of the present application provides an airborne SAR imaging device with FT-M6678 half-precision access, mainly comprising:
[0019] The distance processing module is used to perform distance FFT processing on the received echo data, multiply it by the given reference function and motion compensation factor, and then perform distance IFFT processing; intercept the data of the set length at the echo data center, perform distance FFT processing and distance interpolation, convert the processing results into a semi-precision format to form SAR semi-precision data, and store it in the external memory;
[0020] The post-accumulation azimuth processing module is used to obtain the half-precision data of the azimuth line length in segments for the entire large-frame SAR half-precision data accumulated in the external memory, convert it into single-precision data, and then perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0021] The post-accumulation distance processing module is used to obtain the half-precision data of the range gate length from the external memory, convert it into single-precision data, perform distance IFFT processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0022] The whole block focusing processing module after accumulation is used to obtain the half-precision data of the azimuth line length from the external memory, convert it into single-precision data, and then perform whole block focusing processing, and calculate the single-precision real power image after the whole block is focused;
[0023] The geometric correction and image quantization processing module is used to perform geometric correction and image quantization processing on the single-precision real power image to obtain a SAR grayscale image.
[0024] Preferably, the post-accumulation whole-block focusing processing module comprises:
[0025] A contrast processing unit, used for acquiring half-precision data of the azimuth line length from the external memory in segments, converting it into single-precision data and performing contrast processing thereafter;
[0026] An iterative processing unit, used for performing focus error iterative processing on the single-precision data after contrast processing to obtain error data;
[0027] The error compensation unit is used to re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
[0028] Preferably, in the post-accumulation azimuth processing module and the post-accumulation whole-block focusing processing module, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-cores, and in the post-accumulation distance processing module, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-cores.
[0029] Preferably, single-precision data and processing of single-precision data use the exclusive L2 space in the high-bandwidth core of FT-M6678 and the multi-core shared space SL2 as intermediate cache.
[0030] A third aspect of the present application provides an FT-M6678 digital signal processor, comprising a storage unit and a processing unit, wherein the processing unit is configured to execute the FT-M6678 half-precision access airborne SAR imaging method as described above.
[0031] This application reduces the imaging storage requirements and hardware scale. Compared with SAR imaging with single-precision access, the SAR imaging effect is the same and the processing time is almost the same. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 This is a flow chart of a preferred embodiment of the airborne SAR imaging method with FT-M6678 half-precision access of the present application. DETAILED DESCRIPTION
[0033] In order to make the purpose, technical scheme and advantages of the implementation of this application clearer, the technical scheme in the implementation of this application will be described in more detail in combination with the drawings in the implementation of this application. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described implementation is a part of the implementation of this application, not all of the implementations. The implementation described below with reference to the drawings is exemplary and is intended to be used to explain this application, and cannot be understood as a limitation on this application. Based on the implementation in this application, all other implementations obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. The implementation of this application is described in detail below in combination with the drawings.
[0034] The first aspect of the present application provides an airborne SAR imaging method with FT-M6678 half-precision access, such as Figure 1 As shown, it mainly includes:
[0035] Step S1, performing range FFT processing on the received echo data, multiplying it by a given reference function and a motion compensation factor, and then performing range IFFT processing;
[0036] Step S2, intercepting data of a set length at the echo data center, performing range FFT processing and range interpolation, converting the processing results into a half-precision format to form SAR half-precision data, and storing them in an external memory;
[0037] Step S3, for the SAR half-precision data of the entire large frame accumulated in the external memory, obtain the half-precision data of the azimuth line length in sections, convert it into single-precision data, perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0038] Step S4, obtaining the half-precision data of the range gate length from the external memory, converting it into single-precision data, and then performing range IFFT processing, converting the processed single-precision data into half-precision data and storing it in the external memory;
[0039] Step S5, obtaining half-precision data of the azimuth line length from the external memory, converting it into single-precision data, and then performing whole-block focusing processing, and calculating the single-precision real power image after the whole-block focusing;
[0040] Step S6: geometrically correct and quantize the single-precision real power image to obtain a SAR grayscale image.
[0041] This application is based on the powerful computing performance of FT-M6678 in OpenMP multi-core parallel processing, and implements the azimuth processing of airborne SAR imaging with half-precision access in parallel. The core is to use OpenMP parallelism to extract the half-precision data in the external memory to the internal memory, and then convert it into single-precision data for processing. After the processing is completed, the single-precision data is converted back into half-precision data and stored in the external memory. The processing in the memory mainly includes the azimuth interpolation processing of step S3, the distance IFFT processing of step S4, and the whole block focusing processing of step S5.
[0042] In some optional implementations, in step S3 and step S5, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-core, and in step S4, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-core.
[0043] In some optional implementations, in steps S3 to S5, single-precision data and processing of single-precision data use the exclusive L2 space within the high-bandwidth core of FT-M6678 and the multi-core shared space SL2 as intermediate cache.
[0044] In step S1, the distance FFT processing is performed. At the same time, the processor FT-M6678 generates a reference function according to the control parameters sent by the front end: number of sampling points (NrIn), pulse width, sampling rate, and bandwidth. According to the instantaneous slant distance and imaging slant distance, a motion compensation factor is generated, and then the distance IFFT processing can be performed.
[0045] In step S2, the processing result of the previous step is cut in the center, the cut length is NrOut, and the distance FFT and distance interpolation are performed. The processing result is converted into a half-precision format and stored in the external memory.
[0046] In this step, since the azimuth accumulation length is NfIn, a total of NfIn echo data needs to be sent down, and step S2 is repeated NfIn times, and the processing results are stored in the external memory in sequence, wherein the reference function only needs to be generated once. After the accumulation is completed, the external memory stores a total of NfIn*NrOut*4B of half-precision data.
[0047] In step S3, according to the range gate loop and OpenMP multi-core parallelism, the half-precision data of the azimuth line length corresponding to the current range gate is obtained segment by segment from the half-precision data of the SAR range processing of the entire large frame, the half-precision data is converted into single-precision data, and the azimuth interpolation processing is performed segment by segment. The single-precision data output by the azimuth interpolation processing is converted into half-precision data segment by segment and stored in the external memory.
[0048] In step S4, OpenMP multi-core parallelism is used to loop according to the azimuth line number, and the half-precision data of the range gate length corresponding to the current azimuth line number is obtained from the result of the azimuth interpolation processing, the half-precision data is converted to single-precision data, and the corresponding range IFFT processing is performed. The single-precision data output by the range IFFT processing is converted to half-precision data and stored in the external memory.
[0049] In step S5, OpenMP multi-core parallel half-precision whole block focusing processing is performed. In some optional implementations, step S5 further includes:
[0050] Step S51, obtaining half-precision data of the azimuth line length from the external memory in sections, converting it into single-precision data and performing contrast processing;
[0051] Step S52, performing focus error iterative processing on the single-precision data after contrast processing to obtain error data;
[0052] Step S53, re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
[0053] In step S51, the half-precision data of the azimuth line length corresponding to the current range gate is obtained from the result of the range IFFT processing through OpenMP multi-core parallel processing according to the range gate loop, and the half-precision data is converted into single-precision data for contrast processing. In step S52, the self-focusing error is iteratively processed through the OpenMP multi-core parallel mode. In step S53, the half-precision data of the azimuth line length corresponding to the current range gate is obtained from the result of the range IFFT processing according to the range gate loop through OpenMP multi-core parallel processing, and the half-precision data is converted into single-precision data for error compensation processing. The single-precision real power image after the whole block is focused is output.
[0054] Finally, in step S6, geometric correction and image quantization processing are performed. The geometric correction processing refers to establishing the coordinates of the corrected image pixel points and remapping the image pixel points according to the coordinate points; the image quantization processing refers to converting the image into a grayscale image and then sending it to the avionics integrated equipment to display the image.
[0055] Taking a certain project as a specific example, in a certain SAR imaging mode, the echo data length is 32770 (NrIn), the length after range processing is 16380 (NrOut), and the accumulated length in azimuth is 16380 (NfIn). The imaging width is 16380 (NrOut) in the range output length, and 8190 (NfOut) in the azimuth output length. When the azimuth processing time is very sufficient, the half-precision data accumulated for the entire large frame of SAR range processing is 16380 (NrOut) * 16380 (NfIn) * 4 (B), which is about 1GB. One FT-M6678 uses 1.6GB of external memory, so only one FT-M6678 is needed, while two pieces are needed if single-precision storage is used. After the azimuth interpolation processing is completed, the half-precision data is output, 16380 (NrOut) * 8190 (NfOut) * 4 (B), which is about 0.5GB. Since one FT-M6678 uses 1.6GB of external memory, two areas can be opened, one area is 1GB, starting at AddrA, and the other area is 0.5GB, starting at AddrB. In this way, the azimuth interpolation processing, the range IFFT and the whole block focusing can use AddrA and AddrB for input and output alternating processing, which improves the convenience of the software. If single-precision storage is used, the above work cannot be completed, and the hardware resources of 2 FT-M6678s are halved to 1 FT-M6678. Through the playback verification and comparison of the test flight data, the SAR imaging effect of half-precision access and single-precision access is the same, the processing time is almost the same, and the processing performance is not affected.
[0056] The second aspect of the present application provides an airborne SAR imaging device with FT-M6678 half-precision access corresponding to the above method, mainly comprising:
[0057] The distance processing module is used to perform distance FFT processing on the received echo data, multiply it by the given reference function and motion compensation factor, and then perform distance IFFT processing; intercept the data of the set length at the echo data center, perform distance FFT processing and distance interpolation, convert the processing results into a semi-precision format to form SAR semi-precision data, and store it in the external memory;
[0058] The post-accumulation azimuth processing module is used to obtain the half-precision data of the azimuth line length in segments for the entire large-frame SAR half-precision data accumulated in the external memory, convert it into single-precision data, and then perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0059] The post-accumulation distance processing module is used to obtain the half-precision data of the range gate length from the external memory, convert it into single-precision data, perform distance IFFT processing, and convert the processed single-precision data into half-precision data and store it in the external memory;
[0060] The whole block focusing processing module after accumulation is used to obtain the half-precision data of the azimuth line length from the external memory, convert it into single-precision data, and then perform whole block focusing processing, and calculate the single-precision real power image after the whole block is focused;
[0061] The geometric correction and image quantization processing module is used to perform geometric correction and image quantization processing on the single-precision real power image to obtain a SAR grayscale image.
[0062] In some optional implementations, the post-accumulation whole-block focusing processing module includes:
[0063] A contrast processing unit, used for acquiring half-precision data of the azimuth line length from the external memory in segments, converting it into single-precision data and performing contrast processing thereafter;
[0064] An iterative processing unit, used for performing focus error iterative processing on the single-precision data after contrast processing to obtain error data;
[0065] The error compensation unit is used to re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
[0066] In some optional embodiments, in the post-accumulation azimuth processing module and the post-accumulation whole-block focusing processing module, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-cores; in the post-accumulation distance processing module, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-cores.
[0067] In some optional implementations, single-precision data and processing of single-precision data use the exclusive L2 space within the high-bandwidth core of the FT-M6678 and the multi-core shared space SL2 as intermediate cache.
[0068] A third aspect of the present application provides an FT-M6678 digital signal processor, comprising a storage unit and a processing unit, wherein the processing unit is configured to execute the FT-M6678 half-precision access airborne SAR imaging method as described above.
[0069] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. An airborne SAR imaging method with FT-M6678 half-precision access, characterized in that: include: Step S1, performing range FFT processing on the received echo data, multiplying it by a given reference function and a motion compensation factor, and then performing range IFFT processing; Step S2, intercepting data of a set length at the echo data center, performing range FFT processing and range interpolation, converting the processing results into a half-precision format to form SAR half-precision data, and storing them in an external memory; Step S3, for the SAR half-precision data of the entire large frame accumulated in the external memory, obtain the half-precision data of the azimuth line length in sections, convert it into single-precision data, perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory; Step S4, obtaining the half-precision data of the range gate length from the external memory, converting it into single-precision data, and then performing range IFFT processing, converting the processed single-precision data into half-precision data and storing it in the external memory; Step S5, obtaining half-precision data of the azimuth line length from the external memory, converting it into single-precision data, and then performing whole-block focusing processing, and calculating the single-precision real power image after the whole-block focusing; Step S6: geometrically correct and quantize the single-precision real power image to obtain a SAR grayscale image.
2. The airborne SAR imaging method of FT-M6678 half-precision access as claimed in claim 1, characterized in that: Step S5 further comprises: Step S51, obtaining half-precision data of the azimuth line length from the external memory in sections, converting it into single-precision data and performing contrast processing; Step S52, performing focus error iterative processing on the single-precision data after contrast processing to obtain error data; Step S53, re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
3. The airborne SAR imaging method of FT-M6678 half-precision access as claimed in claim 1, characterized in that: In step S3 and step S5, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-core. In step S4, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-core.
4. The airborne SAR imaging method of FT-M6678 half-precision access as claimed in claim 1, characterized in that: In step S3 to step S5, single-precision data and the processing of single-precision data use the exclusive L2 space in the high-bandwidth core of FT-M6678 and the multi-core shared space SL2 as intermediate cache.
5. An airborne SAR imaging device with FT-M6678 half-precision access, characterized in that: include: The distance processing module is used to perform distance FFT processing on the received echo data, and after multiplying it by a given reference function and motion compensation factor, perform distance IFFT processing; At the echo data center, data of set length is intercepted, and distance FFT processing and distance interpolation are performed. The processing results are converted into a semi-precision format to form SAR semi-precision data, and then stored in the external memory; The post-accumulation azimuth processing module is used to obtain the half-precision data of the azimuth line length in segments for the entire large-frame SAR half-precision data accumulated in the external memory, convert it into single-precision data, and then perform azimuth interpolation processing, and convert the processed single-precision data into half-precision data and store it in the external memory; The post-accumulation distance processing module is used to obtain the half-precision data of the range gate length from the external memory, convert it into single-precision data, perform distance IFFT processing, and convert the processed single-precision data into half-precision data and store it in the external memory; The whole block focusing processing module after accumulation is used to obtain the half-precision data of the azimuth line length from the external memory, convert it into single-precision data, and then perform whole block focusing processing, and calculate the single-precision real power image after the whole block is focused; The geometric correction and image quantization processing module is used to perform geometric correction and image quantization processing on the single-precision real power image to obtain a SAR grayscale image.
6. The FT-M6678 half-precision access airborne SAR imaging device as claimed in claim 5, characterized in that: The post-accumulation whole-block focusing processing module comprises: A contrast processing unit, used for acquiring half-precision data of the azimuth line length from the external memory in segments, converting it into single-precision data and performing contrast processing thereafter; An iterative processing unit, used for performing focus error iterative processing on the single-precision data after contrast processing to obtain error data; The error compensation unit is used to re-acquire the half-precision data of the azimuth line length in segments from the external memory, convert it into single-precision data, and perform error compensation on the single-precision data based on the error data.
7. The FT-M6678 half-precision access airborne SAR imaging device as claimed in claim 5, characterized in that: In the post-accumulation azimuth processing module and the post-accumulation whole-block focusing processing module, the data of the azimuth line length of each range gate is processed in parallel by OpenMP multi-cores. In the post-accumulation distance processing module, the data of the range gate length of each azimuth line number is processed in parallel by OpenMP multi-cores.
8. The FT-M6678 half-precision access airborne SAR imaging device as claimed in claim 5, characterized in that: Single-precision data and its processing use the FT-M6678's high-bandwidth core exclusive L2 space and multi-core shared space SL2 as intermediate cache.
9. An FT-M6678 digital signal processor, characterized in that: The method comprises a storage unit and a processing unit, wherein the processing unit is configured to execute the FT-M6678 half-precision access airborne SAR imaging method according to any one of claims 1 to 4.