CN110361699A - A method of the ice radar data suitable for South Pole aviation measurement scene is handled - Google Patents

Abstract

The invention relates to a method for ice radar data processing suitable for Antarctic aerial observation sites, belonging to the field of computer parallel computing and radar signal processing. The radar profile, using static pre-allocated memory, parallel computing and data block processing methods in the data processing process, satisfies the high resolution, high signal-to-noise ratio and processing of the generated ice radar profile at the Antarctic aerial observation site short-term needs.

Description

A method of ice radar data processing suitable for the Antarctic aerial observation site

technical field

The invention relates to the field of radar signal processing and computer parallel computing, in particular to a method for ice radar data processing and its optimization acceleration suitable for Antarctic aerial observation sites.

Background technique

In the context of global warming, monitoring and assessing changes in polar ice caps and their impact on global climate and sea level is critical. Because the aviation platform has the advantages of high data acquisition efficiency, wide reachable area and large coverage, the aeronautical geophysical survey of synchronous observation of multiple scientific research equipment has become the mainstream method in the world. Among them, aviation ice radar has been used since the 1960s. Widely used in polar ice cap surveys. The electromagnetic waves in the very high frequency band emitted by the ice radar have strong penetrability to the ice body. It is currently the most important technical means to detect the geometric characteristics of polar ice sheets, ice shelves and glaciers, such as the ice-rock interface, the internal structure of the ice body, and the thickness of the ice. The polar ice sheet thickness and subglacial topography data sets obtained from ice radar observations are important parameters and boundary conditions for ice sheet and global climate modeling research.

Therefore, the Antarctic aerial observation site urgently needs a high-resolution, high-signal-to-noise ratio ice radar data processing method, so that the field personnel can conduct preliminary analysis of the observation data in time to grasp the valuable information, and mine new scientific clues and scientific question. At the same time, for the data collected by different scientific research equipment under extreme conditions, it is necessary to timely understand the preliminary results of the observation data of different scientific research equipment through on-site data quality control and analysis, judge its correctness and rationality, and then determine the airborne scientific research The actual performance and operating conditions of the equipment provide a basis for possible subsequent equipment maintenance and adjustment of flight parameters and equipment parameters. The ice radar data can accurately and intuitively reflect the subglacial terrain of the Antarctic, making the ice radar profile generated on the spot an important reference for on-site data quality control and analysis.

SUMMARY OF THE INVENTION

In view of this, in order to solve the problem of rapid acquisition of high-resolution, high-signal-to-noise ratio ice radar profiles at the Antarctic aerial observation site. The present invention proposes a method flow for processing ice radar data in Antarctic aerial observation site. Through down-conversion, DC component removal, pulse compression, coherent superposition and incoherent superposition modules respectively, the ice radar profile obtained through the processing algorithm, It has the characteristics of high signal-to-noise ratio and high resolution. In view of the above proposed algorithm, the methods of static pre-allocated memory, parallel computing and data block processing are used in the data processing process, which improves the processing efficiency of the algorithm by about 3.8 times and greatly improves the ice radar data of the aerial observation site. processing efficiency.

Down-conversion processing module: The collected ice radar raw data needs to undergo quadrature detection to extract the amplitude and phase information of the signal, and then complete the down-conversion from the baseband signal to the video signal (DC). However, the traditional quadrature detection method is completed by a phase detector, which requires the receiver to have an analog mixer and a low-pass filter, and a total of four detectors are required for the two gain channels, which increases the complexity of the receiver and the system cost. . At the same time, because the analog circuit is greatly affected by the difference of the device, and the IO channel amplitude and phase imbalance caused by the temperature offset often produces an image component that is difficult to eliminate, the processing performance of the ice radar raw data is limited. Therefore, the application of intermediate frequency sampling technology to this ice radar system can solve the above problems well.

Removal of DC component processing module: When the ice radar transmitter is turned off, the RF signal that is coupled to the antenna and radiated through the radio frequency (RF) switch, power line, etc., is called DC offset. The resulting RF signal is superimposed on the desired signal in the mixer. The method of removing the DC component first needs to estimate the noise electric average value for each channel of data, and then subtract the estimated value from the echo signal. S represents a complete echo data, nS represents the number of sampling points that can be regarded as noise at the end of the data, and N represents the noise level, then the average noise level N _Ave is (as shown in formula 1):

The echo signal after removing the DC component is expressed as (as shown in Equation 2)

Pulse compression processing module: Generally, the method of increasing the length of the transmitted signal is used to increase the signal-to-noise ratio of the received signal to obtain sufficiently accurate target parameters, but the length of the extended signal will be greater than the signal length required by the resolution (the signal bandwidth is too large. ), so that the normal frequency points cannot appear in the image. The pulse compression technique is to convolve the echo signal with the matched filter to obtain short pulses that can meet the resolution requirements, thereby improving the display of the interface between the ice surface and the bedrock in the ice radar profile. Pulse compression effectively utilizes the convolution feature in Fourier transform, and improves the range resolution of ice radar data by convolving each returned pulse signal with a matched filter (as shown in Equation 3).

In formula (3), y(t) is the output result of pulse compression; s(t) is the input signal; h(t) is the impulse response of the windowed matched filter.

The mathematical expression of the pulse compression process is as follows, set the transmitted signal as s(t), and perform Fourier transform on it as

where θ(t) is the modulation phase of the signal, and S(ω) is the signal after inverse Fourier transform.

The output signal after the matched filter is convoluted with the pulse signal is

where S(ω) is the signal after inverse Fourier transform, H(ω) is the impulse response of the windowed matched filter, and y(t) is the signal after pulse compression.

Coherent superposition processing module: After pulse compression processing, the range resolution of the ice radar profile can be effectively improved, but the noise of the ice radar image after pulse compression processing cannot be improved (the signal-to-noise ratio is too low), resulting in the intermediate layer Some of them cannot be effectively identified, and the bedrock interface is relatively vague, which cannot meet the requirements of subsequent quality control. Therefore, the ice radar data can be further denoised using the coherent superposition processing method.

Incoherent superposition processing module: The coherent superposition operation processing effectively improves the signal-to-noise ratio of ice radar data, but it will generate speckle noise in the ice radar profile, which greatly interferes with the extraction and analysis of the effective information of the profile in quality control. Therefore, it is necessary to perform incoherent superposition processing on the data to remove speckle noise. By incoherent stacking, some peaks (too bright) and nulls (too dark) can be removed from the data. After the incoherent superposition processing is completed, the complete ice radar profile can be obtained. The subsequent processing module accelerates and optimizes the proposed ice radar processing program according to the on-site requirements, so as to improve the speed of on-site ice radar data processing.

Static pre-allocated memory module: There are a lot of matrix scientific calculations in the ice radar field data processing process, which requires the CPU to perform a huge amount of read and write operations. The CPU reads the data mechanism. Each time the CPU finds data in the memory, it caches the continuous physical address data in the memory block where the data is located into the CPU cache. The read efficiency between the CPU cache and the CPU is the highest, and the data to be processed should be kept in the cache as much as possible in programming. There are two types of memory allocation methods: dynamic and static. Dynamic allocation of memory is likely to cause discontinuous physical addresses, which greatly increases the difficulty of the CPU to find the ice radar data in the cache to be processed in sequence, or causes the search to fail. Fall back to the next priority, that is, re-seek and read from memory with low read efficiency; while statically allocated memory can allocate sequential ice radar data to consecutive physical addresses, which can significantly improve the CPU read from cache. The hit rate of the next sequential pending ice radar data. Therefore, in order to improve the efficiency of CPU reading and writing ice radar data, the method of statically allocating memory is adopted.

Parallel computing module: Since the scientific calculation of the matrix accounts for more than 90% of the overall calculation, the ice radar data processing algorithm is a computationally intensive task. Computation-intensive tasks mainly rely on the computing power of the CPU. Therefore, according to the characteristics of computationally-intensive tasks, in the process of parallel transformation of the serial ice radar data processing algorithm, in order to ensure efficient parallel improvement, it is necessary to divide the tasks into as much as possible. Reduce the number of threads executing in parallel and reduce the scheduling time for the system to switch between tasks, so the number of threads that exceed the number of CPU cores is not considered. At the same time, in order to ensure the highest utilization efficiency of the CPU, each CPU processing core needs to be allocated to parallel threads, so the 4-thread parallel ice radar data on-site processing algorithm is finally adopted.

Data block processing module: Too large or too small amount of ice radar data will lead to a decrease in the efficiency of the parallelized ice radar data field processing program. The parallelized ice radar data processing program is divided into two parts: system scheduling assignment tasks and parallelized data processing. Therefore, the total time of the parallelized ice radar data processing program consists of two parts: the system assignment task time and the parallelized data processing time. The system scheduling and assigning task time is relatively fixed, mainly the time it takes for the main thread to open up 4 sub-threads. Therefore, in order to improve the efficiency of the parallelized program, the amount of data processed each time should be increased as much as possible; however, the excessively large amount of data processed in parallel is usually accompanied by high memory usage, which will prompt the CPU to read through the least efficient hard disk. Ice radar data, resulting in a mismatch between CPU reading data efficiency and CPU processing efficiency. So, to balance the constraints, we need to chunk the data and find the optimal number of chunks.

The invention proposes an algorithm flow for processing ice radar data in Antarctic aerial observation site. Ice radar data can be quickly processed on the Antarctic site to obtain ice radar profiles with high resolution and high signal-to-noise ratio. It provides a new idea for ice radar data processing in China's Antarctic aerial observation site.

Description of drawings

Fig. 1 is the overall flow chart of the implementation of the present invention

FIG. 2 is a schematic flowchart of the down-conversion module of the present invention.

FIG. 3 is a schematic flowchart of the present invention for removing a DC component module.

FIG. 4 is a schematic flow chart of the pulse compression module of the present invention.

FIG. 5 is a schematic flowchart of the coherent superposition module of the present invention.

FIG. 6 is a schematic flowchart of the incoherent superposition module of the present invention.

FIG. 7 is a cross-sectional view of the ice radar obtained by the processing of the present invention.

FIG. 8 is a schematic flow chart of static pre-allocation of memory modules.

FIG. 9 is a schematic diagram of the data processing flow of the parallelized ice radar.

FIG. 10 is a schematic diagram of a data block processing flow.

Detailed ways

The present invention will be described in detail below with reference to the specific embodiments shown in the accompanying drawings.

Fig. 1 is an overall flow chart of an algorithm for ice radar data processing at Antarctic aerial observation sites proposed by the present invention, as shown in Fig. 1, including:

Down-conversion module, DC component removal module, pulse compression module, coherent superposition module and incoherent superposition module.

Down-conversion module: The IF sampling technology adopts the low-pass filtering method, which directly mixes the original data of the ice radar with the 70MHz local oscillator signal, then removes the carrier frequency, and then passes the low-pass filter (LPF) to filter out the frequency generated during the mixing. Upper sideband signal and higher harmonic signal. The low-pass filter is of order 128, and the cutoff frequency is The FIR filter is used to complete quadrature detection and down-conversion, and the down-converted ice radar data is obtained.

Removal of DC component module: Estimate the noise electric average value of the ice radar data after down-conversion processing, and then subtract the estimated value from the down-converted signal (using the detrend function in matlab) to obtain the processed flow component. Ice radar data.

Pulse compression module: Perform fast Fourier transform on the ice radar data after removing the DC component to obtain the frequency domain signal of the ice radar data (using the fft function in matlab), then convolve with the matched filter, and finally perform the inverse fast Fourier Leaf transform (using the ifft function in matlab) converts ice radar data into time-domain signals to obtain pulse-compressed ice radar data.

Coherent superposition module: directly superimpose 10 data of ice radar data after pulse compression to obtain ice radar data after coherent superposition.

Incoherent superposition module: First, extract the amplitude data of the ice radar data after coherent superposition (using the abs function in matlab), and then superimpose 5 channels of data to obtain the ice radar data after incoherent superposition, that is, after the processing is completed The ice radar data can be generated with high signal-to-noise ratio and high-resolution ice radar profile. The subsequent processing module is based on the field requirements, and the proposed ice radar processing program is accelerated and optimized to improve the speed of on-site ice radar data processing.

Static pre-allocated memory: We adopt the method of using static pre-allocated memory for intermediate variables to improve the fragmentation of fragmented memory, that is, pre-allocate a fixed continuous memory for each intermediate variable in data processing in advance before ice radar data processing To ensure that all intermediate quantities generated by each link in each data processing process have a corresponding and unchanged memory address for storage, so as to avoid a large number of memory fragments caused by repeated memory allocation and release operations. By reducing the generation of memory fragments, the processing efficiency of the parallelized ice radar processing program can be effectively improved.

Parallel computing module: The program first reads the total amount of ice radar data n, and then performs multi-threaded static pre-allocation of memory according to the designed number of threads 4, divides the total amount of ice radar data evenly according to the number of threads, and then performs 4 serializations at the same time. The data of the row is read and processed until all the tasks assigned by each thread are completed. The tasks in different core processors are parallel, and the tasks within each thread are still carried out in the order of the serial processing flow. Finally, the parallel results are sequentially stored in the pre-allocated memory addresses to complete data processing.

Data block processing module: The ice radar data to be processed will be processed for 1100 of them each time. After the data processing of 1100 data is completed and stored, the data processing of the following 1100 data will be carried out, and the above process will be repeated until all the data are processed. After all the processed data is completed, the block-by-block processing module is completed. After the optimization of the above three modules is completed, the processing speed of the proposed on-site ice radar processing algorithm can meet the on-site requirements, and the ice with high resolution and high signal-to-noise ratio can be quickly generated. Radar profile.

It should be understood that although this specification is described according to embodiments, not every embodiment only includes an independent technical solution, and this description in the specification is only for the sake of clarity, and those skilled in the art should take the specification as a whole, The technical solutions in each embodiment can also be appropriately combined and implemented according to the understanding of those skilled in the art.

The series of detailed descriptions listed above are only specific descriptions for the feasible embodiments of the present invention, and they are not used to limit the protection scope of the present invention. Any equivalent embodiments or changes made without departing from the technical spirit of the invention are should be included within the protection scope of the present invention.

Claims (5)

1. a kind of method that the ice radar data suitable for South Pole aviation measurement scene is handled, it is characterised in that: including down coversion, DC component, pulse compression, coherent superposition and non-coherent addition module are removed, ice radar profile is obtained, in data processing The method of static predistribution memory, parallel computation and deblocking processing has been used in journey.

Wherein, down-converted module: the Intermediate Frequency Sampling use low pass filtering method, directly by original ice radar data with The local oscillation signal mixing of 70MHz moves back carrier frequency, then by low-pass filter LPF filter out the upper side band signal generated when mixing with Higher hamonic wave signal；

It removes DC component treatment module: noise level Estimation of Mean being carried out to the ice radar data after down coversion, then from lower change The estimated value is subtracted in frequency treated ice radar data, the ice radar data after obtaining removal component；

Process of pulse-compression module: Fast Fourier Transform (FFT) is carried out to ice radar data after removal DC component, obtains ice radar The frequency-region signal of data, then convolution is carried out with matched filter, inverse fast fourier transform is finally carried out, ice radar data is turned It is changed to time-domain signal, obtains the compressed ice radar data of pulse；

Coherent superposition processing module: the superposition of 10 track datas is directly carried out to the compressed ice radar data of pulse, is concerned with Superimposed ice radar data；

Non-coherent addition module: the processing of amplitude data is extracted to the ice radar data after coherent superposition first, then carries out 5 The superposition of road amplitude data is to get to the ice radar data that can be used for generating high s/n ratio, high-resolution ice radar profile.

2. the method that a kind of ice radar data suitable for South Pole aviation measurement scene according to claim 1 is handled, Be characterized in that: the low-pass filter use 128 ranks, cutoff frequency forFIR filter complete orthogonal detection And down coversion, the ice radar data after obtaining down coversion, wherein B indicates bandwidth, f_sIndicate sample frequency.

3. the method that a kind of ice radar data suitable for South Pole aviation measurement scene according to claim 1 is handled, Be characterized in that: this method uses centering using the processing mode for the static predistribution memory that can effectively improve treatment effeciency The area of a room improves the case where fragment RAM fragmentation using the method for static predistribution memory, i.e., handles it in ice radar data Before, a fixed contiguous memory physical address is pre-allocated to each intermediate variable of Data processing in advance, to guarantee per pass number One piece of memory address that is corresponding and remaining unchanged is owned by according to all intermediate quantities that link each in treatment process generates to carry out Storage.

4. the method that a kind of ice radar data suitable for South Pole aviation measurement scene according to claim 1 is handled, Be characterized in that: this method uses parallel computation mode, i.e. program reads ice radar data total amount n first, further according to the line of design Number of passes 4 carry out multithreading static predistribution memory, by ice radar data total amount according to Thread Count even partition, then simultaneously into Row 4 serial reading data are handled, until the task of per thread distribution is fully completed；In different core processor Task be it is parallel, task in per thread is carried out still according to serial process process sequence；It is each after finally will be parallel As a result it is stored in the memory address allocated in advance in order, completes data processing.

5. the method that a kind of ice radar data suitable for South Pole aviation measurement scene according to claim 1 is handled, Be characterized in that: this method uses deblocking processing mode: ice radar data to be processed will do it therein 1100 every time Data processing is carried out, after waiting the processing of 1100 track datas to complete and store, then carries out the data processing of following 1100 track data, The circulation above process is fully completed until whole pending datas.