A hierarchical MTD processing method based on inter-frame extraction

Abstract

The present invention relates to the detection of slow-moving targets on the ground, and more particularly to a signal processing method for detecting pedestrians on the ground using a pulse radar. The present invention comprises the steps of obtaining data by sliding a window between frames; filtering and extracting between frames; and fast Doppler filtering of the extracted data. In the application of ground security monitoring, the present invention can efficiently realize the detection of pedestrians moving at a close distance.

Description

Technical Field

[0001] The present invention relates to the detection of slow-moving targets on the ground, and in particular to a signal processing method for detecting pedestrians on the ground using a pulse radar. In the application of ground security monitoring, the method can efficiently realize the detection of pedestrians moving at a close distance. Background Art

[0002] In order to achieve large-scale real-time and automatic monitoring tasks, radar is used to monitor people and vehicles in an area of ​​1km to 5km, which is an advanced means of security monitoring. In order to achieve long-distance monitoring, pulse-based phased array radar or MIMO radar is generally used, which has a long detection distance and strong clutter suppression ability.

[0003] The signal processing of traditional pulse radar adopts MTI or MTD processing method. In security applications, since the pulse repetition period corresponding to the distance is relatively short, a large number of pulses can be obtained in one coherent processing cycle, which is more suitable for PD processing.

[0004] The core of PD signal processing is the Doppler filter group. The filter group can be implemented one by one using FIR filters. However, in order to reduce the amount of calculation, especially when there are a large number of pulses, FFT is generally used to implement the output of all filter groups at one time. The processing order of FFT is the same as the number of pulses. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention adopts a new filter group generation method for targets such as pedestrians and vehicles, which greatly reduces the processing order of FFT. Under the premise of ensuring PD processing performance, it further reduces the complexity of the processing circuit, greatly reduces the requirements for system resources, and reduces production costs.

[0006] The technical solution of the present invention is: a layered MTD processing method based on inter-frame extraction, characterized in that it includes the following steps:

[0007] Step 1: Slide the window between frames to obtain data;

[0008] The input data of the inter-frame sliding window is rearranged according to the distance unit before the inter-frame sliding window filtering process according to the inter-frame extraction order D and the inter-frame filter order L. The data matrix corresponding to the i-th distance unit and the k-th group of input data to be processed is:

[0009]

[0010] k takes values ​​of 1, 2, 3, ..., and the maximum value of k is M / D, where M represents the number of total pulse repetition cycles required for the current coherent processing cycle;

[0011] Step 2: inter-frame filtering and extraction;

[0012] Inter-frame filter coefficients:

[0013]

[0014] The inter-frame filter coefficient is matrix-multiplied with the input data Xb to obtain the filtered and extracted data Ya.

[0015] Ya=Wb*Xb;

[0016] Step 3: Fast Doppler filter group of extracted data

[0017] The Ya data extracted between the i-th range unit frames is subjected to a windowing process again to suppress the side lobes generated by FFT, and the extracted windowed data Yb is obtained. Then, an M / D-order FFT operation is performed on Yb to obtain the output data Yc of M / D Doppler filter groups.

[0018] According to the hierarchical MTD processing method based on inter-frame extraction as described above, it is characterized in that: the output data Yc=FFT(Yb,M / D), and the subsequent signal processing flow is performed according to M / D Doppler channels.

[0019] According to the hierarchical MTD processing method based on inter-frame extraction as described above, it is characterized in that: M takes a value of 8192 and D takes a value of 16.

[0020] According to the above-mentioned hierarchical MTD processing method based on inter-frame extraction, it is characterized in that: the inter-frame filter order L is 64

[0021] According to the above-mentioned hierarchical MTD processing method based on inter-frame extraction, it is characterized by being applied in pulse radar.

[0022] The beneficial effects of the present invention are as follows: first, in view of the characteristics of slow pedestrian targets, the Doppler velocity in the slow area is extracted in the complete Doppler velocity detection area, and finally only FFT processing is required in the slow area, thereby reducing the number of FFT processing orders. Second, for the short-range detection of pedestrians and vehicles, while maintaining the system performance indicators, the data transmission bandwidth and the MTD computing and storage resource consumption are greatly reduced, effectively reducing the difficulty and cost of engineering implementation. Third, the difficulty and production cost of engineering implementation of pulse radars used for detecting slow-moving pedestrians / vehicles are effectively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is the filter characteristic diagram.

[0024] Figure 2 Invention flow chart. DETAILED DESCRIPTION

[0025] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0026] In this embodiment, the radar is a C-band radar (wavelength nmd is 0.054m). When PRI is 30us (repetition frequency fr is 33.3kHz), the first blind speed is 899m / s (33300*0.054 / 2=899m / s). After D=16 times extraction, the speed range is 28m / s, and pedestrian and general vehicle detection (speed within 25m / s) is not affected.

[0027] The present invention comprises the following steps:

[0028] Step 1: Slide the window between frames to obtain data

[0029] The single-channel input signal is S(n,m), where n is the distance sampling unit number. When there are N distance sampling units in each repetition period, the values ​​are 1, 2, 3...N in sequence. m is the pulse repetition period number. When the total number of pulse repetition periods (PRI) required for the current coherent processing period (CPI) is M repetition periods in total, m takes values ​​1, 2, 3...M in sequence.

[0030] The data matrix of the input signal is:

[0031]

[0032] According to the actual application requirements, the input data of the inter-frame sliding window needs to be rearranged according to the distance unit before the inter-frame sliding window filtering process according to the inter-frame extraction order D and the inter-frame filter order L. The data matrix corresponding to the i-th distance unit and the k-th group of input data to be processed is:

[0033]

[0034] In the current coherent processing cycle, the k-th column data in the matrix corresponds to each group of input data of the inter-frame filter in sequence, and k takes the value of 1, 2, 3, ..., and the maximum value is M / D.

[0035] The extraction order D is determined by the system indicator requirements, which is mainly related to the radar's operating frequency, pulse repetition period and maximum measurement speed. For a system with an operating frequency of C band and a repetition period of about 30us, when the maximum measurement speed is required to be no less than 25m / s, the maximum extraction order can be about D=18. When only pedestrian target detection is required, the maximum measurement speed is generally required to be no less than 2m / s. At this time, the maximum extraction order can be about 225.

[0036] The inter-frame filter order L is related to the decimation order D. The higher the D value, the narrower the passband of the decimation filter. In order to maintain the side lobes and in-band flatness of the filter, the larger the filter order L is.

[0037] In this example, the system works in the C band, with a repetition period of 30us and a maximum measurement speed requirement of 25m / s. To facilitate engineering implementation, the system parameters are set to the total number of PRIs in the CPI M=8192, the extraction order D=16, and the inter-frame filter order L=64.

[0038] Step 2: Inter-frame filtering and decimation

[0039] The inter-frame filtering and decimation of the i-th range unit data is performed between PRIs to delete irrelevant areas and output only part of the data while maintaining valid information.

[0040] Inter-frame decimation filter (L = 64 levels), filter coefficients

[0041]

[0042] Filter characteristics Figure 1 As shown, the passband accounts for a proportion of 1 / D.

[0043] The inter-frame filter coefficient is matrix-multiplied with the input data Xb to obtain the filtered and extracted data Ya.

[0044] Ya=Wb*Xb

[0045] Each column of data in Xb is output as a data after completing the dot multiplication with the filter coefficient. Compared with the data volume of the original M PRI frames, D times of data is extracted. In the Ya data containing valid information, the number of PRI frames at this time is reduced to M / D. In this example, M / D = 8192 / 16 = 512.

[0046] Step 3: Fast Doppler filter group of extracted data

[0047] The Ya data extracted between the i-th range unit frames is subjected to a windowing process (M / D order Hamming window) to suppress the side lobes generated by FFT and obtain the extracted windowed data Yb. Then, an M / D order FFT operation is performed on Yb to obtain the output data Yc of M / D Doppler filter groups.

[0048] Yc=FFT(Yb,M / D)

[0049] The subsequent signal processing flow is performed according to M / D Doppler channels.

[0050] The FFT order of conventional processing is M, while the present invention is only M / D. The traditional MTD is divided into two levels through inter-frame filtering and extraction. The former part is used to reduce the MTD processing order, and the latter part performs low-order MTD processing.

[0051] According to the comparative statistics of the multiplication accumulation and addition calculation amount of the FFT algorithm, when M=8192 and D=16, the FFT calculation amount is saved by about 23 times.

[0052] The present invention ensures that the performance is not lost. By utilizing the characteristics of only detecting slow targets (people and cars / ships / drones, etc.) at short range (generally no more than 5 km), the data in the high-speed area is effectively deleted, which greatly reduces the amount of calculation. It has no effect on the detection performance of the system, such as detection distance / detection speed, etc.

Claims

1. A hierarchical MTD processing method based on inter-frame extraction for pulse radar, characterized by: The steps include: Step 1: Slide the window between frames to obtain data; The single-channel input signal is S(n,m), where n is the distance sampling unit number. When there are N distance sampling units in each repetition period, the values ​​are 1, 2, 3...N in sequence. m is the pulse repetition period number. When the total number of pulse repetition periods required for the current coherent processing period is M repetition periods, the values ​​of m are 1, 2, 3...M in sequence. The data matrix of the input signal is: According to the inter-frame extraction order D and the inter-frame filter order L, the input data Xa is rearranged according to the distance unit before the inter-frame sliding window filtering process. The data matrix corresponding to the i-th distance unit and the k-th group of input data to be processed is: k takes values ​​of 1, 2, 3, ..., and the maximum value of k is M / D, where M represents the number of total pulse repetition cycles required for the current coherent processing cycle; Step 2: inter-frame filtering and extraction; Inter-frame filter coefficients: Wb=〔w(1) w(2)……w(L)〕 The inter-frame filter coefficient is matrix-multiplied with the input data Xb to obtain the filtered and extracted data Ya. Ya=Wb*Xb; Step 3: Fast Doppler filter group of the extracted data; The Ya data extracted between the i-th range unit frames is subjected to a windowing process again to suppress the side lobes generated by FFT, and the extracted windowed data Yb is obtained. Then, an M / D-order FFT operation is performed on Yb to obtain the output data Yc of M / D Doppler filter groups.

2. The hierarchical MTD processing method based on inter-frame extraction for pulse radar according to claim 1 is characterized in that: The output data Yc=FFT(Yb,M / D), and the subsequent signal processing flow is performed according to M / D Doppler channels.

3. The hierarchical MTD processing method based on inter-frame extraction for pulse radar according to claim 1 is characterized in that: The value of M is 8192 and the value of D is 16.

4. The hierarchical MTD processing method based on inter-frame extraction for pulse radar according to claim 1 is characterized in that: The inter-frame filter order L is 64.

5. The hierarchical MTD processing method based on inter-frame extraction for pulse radar according to claim 1 is characterized in that: Used in pulse radar.

Images