CN108490410B - Two-coordinate radar sea target joint detection and tracking method - Google Patents

Abstract

The invention provides a method for jointly detecting and tracking a sea target by a two-coordinate radar, which comprises the steps of firstly carrying out background self-adaptive threshold detection on original video data, preliminarily removing clutter and noise, then extracting point track information according to the sequence of distance first and direction second, secondarily removing threshold passing echoes which do not meet a condensation criterion, calculating point track quality according to the two-dimensional distance direction echo characteristics condensed into the point track, then carrying out dynamic hypothesis path search, defining a search path cost function by taking the point track quality as a reference, forming a dynamic hypothesis track cluster, and finally carrying out detection judgment and track optimization merging through a multi-period cost function accumulation result to realize the joint detection tracking processing on the target.

Description

Two-coordinate radar sea target joint detection and tracking method

Technical Field

The invention belongs to the technical field of radar target detection and tracking.

Background

The detection and tracking of a radar target under a sea clutter background is one of the hot problems in the research of the radar detection field, and because sea clutter has strong echo amplitude in a time domain and presents the characteristic of non-uniform distribution, a large amount of clutter is always left after time domain detection; in a frequency domain, the motion speed of a sea target is low, the Doppler frequency of the target is located in the frequency spectrum range of sea clutter, effective separation cannot be carried out, and a large number of false targets are always generated along with automatic detection and extraction of the sea target. In a radar weak multi-target detection method based on a dynamic programming method (electronic journal, 2006, Vol.34, No.12, pp: 2142-; the clutter amplitude is strong under the sea clutter background, and the sea clutter can be accumulated in a non-coherent accumulation mode and cannot be separated from the target. In a document, namely a novel algorithm for tracking before radar detection based on dynamic programming (electronic bulletin, 2008, Vol.36, No.9, pp:1824 and 1828), a novel algorithm for tracking before detection is provided by taking a linearly moving target as an example, the maximum conversion state of a signal is determined by using the maximum motion information of the target, then a detection threshold is set by using the characteristics of clutter distribution to filter the accumulated clutter, and then target information is extracted; in practical application, when the sea situation is high in sea level, the rayleigh distribution cannot accurately describe the non-uniform and non-stable characteristics of the sea clutter, the motion speed of the sea target is low, and the motion characteristics of the target echo in a short time are not obvious due to the existence of radar detection errors.

Aiming at the problem existing in the tracking Processing Before detection only by using position information or amplitude information, the author proposes the idea of tracking Processing Before detection at the point Track level in the document A Novel Dynamic Programming Algorithm for Track-Before-Detect in radio Systems (IEEE Transactions on Signal Processing,2013, Vol.61, No.10, pp 2608-; however, the local signal-to-noise-and-noise ratio of the point trace is used as an index function, so that the target and the clutter cannot be effectively distinguished, and the performance is also rapidly deteriorated in a strong clutter environment.

Disclosure of Invention

The invention provides a method for jointly detecting and tracking a sea target by a two-coordinate radar, aiming at the defects in the background technology, firstly carrying out background self-adaptive threshold detection on original video data, different detection criteria are selected in a self-adaptive manner by counting the characteristics of echo data in a reference unit of a unit to be detected, the clutter and the noise are preliminarily removed, then the trace point information is extracted according to the sequence of distance first and direction second, secondary elimination is carried out on the threshold-passing echoes which do not meet the condensation criterion, the point trace quality is calculated according to the distance and direction two-dimensional echo characteristics condensed into the point trace, and then the dynamic hypothesis path search is carried out, and finally, carrying out detection judgment and track optimization merging through a multi-period cost function accumulation result to realize combined detection tracking processing on the target.

The method comprises the following steps: background self-adaptive threshold detection, namely judging that a local scene is a noise background, a clutter background or an adjacent strong target scene according to data in a reference window of a unit to be detected, self-adaptively selecting a detection criterion and a detection threshold, and performing initial threshold detection pretreatment on original video data;

step two: point trace extraction based on quality evaluation, two coordinate radars adopt M in the distance direction according to the sequence of distance first and orientation second_r/N_rCriterion sliding window detection of condensation, azimuth using M_a/N_aThe criterion sliding window detects the agglomeration, calculates the trace quality while the trace agglomeration process, and for each distance trace, the distance agglomeration quality is defined as Q_r＝N_{r_plot}-M_r，N_{r_plot}The number of distance units condensed into the current trace point; setting M to form the final trace_aA distance trace amplitude of A_{r_plot_i},i＝1,…,M_aThe quality of the dot trace after the azimuthal agglomeration is defined as

Wherein N is_{a_plot}Number of distance traces condensed as current trace, C_qIs a trace point quality normalization constant; counting the number of traces in each period of clutter region, and when the number of the traces exceeds the upper limit value N of the set number of the traces_{plot_max}Then, sorting and outputting the points according to the trace point quality from large to small;

step three: forming a search path based on the dynamic hypothesis cluster, and setting the maximum value of the motion speed of the sea target as V_maxFor all the traces of points formed in the k-th period, Plot_{k_i},i＝1,…,N_{plot_k}，N_{plot_k}The current point trace Plot is used as the number of the point traces in the k period_{k_i}Position-centered, with V_maxT is a radius to establish a circular wave gate, and T is an antenna scanning period; wave inThe search path formed by the first k-1 periods is correlated in the gate, and the correlation criterion is to select the path with the maximum cost function value in the gate for correlation, namely psi_k(x_k)＝argmax(I(x_k-1) Wherein I (x)_k-1)＝Q_{plot_k-1}+max(I(x_k-2) Is a search path cost function, x_kFor the kth period trace information, Q_{plot_k-1}Associating trace point quality for the k-1 th period; forming a starting point of a search path for all the point traces of the 1 st period;

step four: extracting a multi-period joint detection judgment target, setting the number of joint detection periods to be N_cycleSetting a multi-period accumulation cost function detection threshold T for a search path with the duration reaching a joint detection period_IFor the search path satisfying the detection threshold, according to psi_k(·),k＝1,…,N_cycleExtracting the path position to form a flight path; after the flight paths are extracted, position comparison is carried out between every two flight paths generated in the current period, if the position is N_cycleWithin one period there is M_cycleAnd if the paths of the points in the periods are the same, merging the paths according to the principle that the maximum cost function is the optimal function, and finishing the extraction of the multi-period combined judgment target.

The invention has the innovation points that sea target trajectory information is extracted through initial adaptive threshold detection and then based on quality evaluation, and targets under a strong clutter background are jointly detected through a path search cost function constructed based on the trajectory quality in a multi-period mode, so that the problem that sea targets cannot be effectively extracted through position information due to unobvious motion characteristics is solved, and the problem of unstable accumulation performance of amplitude accumulation or signal-to-noise ratio is solved.

Drawings

FIG. 1 is a processing flow chart of a two-coordinate radar sea target joint detection and tracking method of the present invention.

Fig. 2 is a set of measured collected video data of a bank-to-sea radar used in an embodiment of the present invention.

Fig. 3 is a result of adaptive threshold detection processing on measured data according to an embodiment of the present invention.

FIG. 4 shows the result of the trace aggregation process for 50 consecutive antenna cycles in accordance with one embodiment of the present invention.

FIG. 5 is a statistical analysis of the quality of 15 target and clutter traces in an exemplary embodiment of the invention.

FIG. 6 is a diagram illustrating traces left after removing some false traces according to trace quality in an embodiment of the present invention.

Fig. 7 shows the result of jointly detecting and tracking the sea target according to the method of the present invention in the embodiment of the present invention.

Detailed Description

The processing flow of the sea target joint detection and tracking method by the two-coordinate radar is shown in fig. 1, and the implementation of the method is specifically explained by combining the flow chart and the embodiment, wherein the process comprises the following steps:

the method comprises the following steps: background adaptive threshold detection.

Let the original video data processed by each main trigger signal be f_MTP(i),i＝1,…,N_rangeIn which N is_rangeAnd (4) performing self-adaptive threshold detection on distance direction sampling points along a distance direction sliding window, and preliminarily eliminating clutter and noise. Let the detection reference unit size be N_aveProtection unit size N_proFor each distance unit, counting the current unit f to be detected_MTP(i) The maximum value of data in the left reference unit and the right reference unit is Amp_{max_1}And Amp_{max_2}If the maximum value exceeds the set threshold

Then N is left and right to the distance unit where the maximum value is located_proDeleting the data in each reference unit, and if the maximum value does not exceed the threshold, not deleting; then, counting the current unit f to be detected_MTP(i) The mean values of the data in the left and right reference units are respectively

And

let the mean value of the radar noise floor be A_noiseIf | A_noise-μ₁< Δ A and | A_noise-μ₂If | < Δ A, Δ A is the set threshold, then the current unit f_MTP(i) The detection threshold is V_T＝β₁·(μ₁+μ₂) /2, otherwise, the detection threshold is V_T＝β₂·max(μ₁,μ₂)，β₁And beta₂For detecting the threshold coefficient, max () is a get big operation; if the detection threshold is satisfied, f is reserved_MTP(i) If not, the distance unit data is set to zero.

In the embodiment of the invention, a set of measured data is adopted for processing, the set of measured data is collected by a certain type of shore-to-sea radar at a port, as shown in fig. 2, most targets in a scene are fishing vessels catching fish, and the result after self-adaptive threshold detection is shown in fig. 3.

Step two: and (4) point trace extraction based on quality evaluation.

After self-adaptive threshold detection, trace point information is extracted, and M is adopted in the distance direction_r/N_rThe condensation is detected by a standard sliding window, and the range-direction resolution of two coordinate radars is set as R_resThe unit: meter, distance sampling unit size is Δ R, unit: the weight of the rice is reduced,

if N is present_rLess than or equal to 3, taking N_r＝3，

Which represents a rounding-up operation on the upper part,

indicating a rounding-down operation if consecutive N_rWithin each distance unit is more than or equal to M_rIf the data of each unit is more than 0, performing distance condensation treatment; direction of orientation adopts M_a/N_aThe method comprises detecting condensation by sliding window, and setting the 3dB main lobe width of azimuth antenna beam as phi_3dBThe unit: the degree of the magnetic field is measured,in a main lobe width phi_3dBThe number of the internal correlation pulses is N_a，

If N is consecutive_aWithin one pulse is more than or equal to M_aIf the pulses have distance traces with the same distance, the azimuth agglomeration treatment is carried out to extract trace information.

Calculating trace point quality while performing trace point agglomeration, wherein for each trace point distance agglomeration quality is defined as Q_r＝N_{r_plot}-M_r，N_{r_plot}The number of distance units condensed into the current trace point; setting M to form the final trace_aA distance trace amplitude of A_{r_plot_i},i＝1,…,M_aThe quality of the dot trace after the azimuthal agglomeration is defined as

Wherein N is_{a_plot}Number of distance traces condensed as current trace, C_qIs a trace point quality normalization constant; counting the number of traces in each period of clutter region, and when the number of the traces exceeds the upper limit value N of the set number of the traces_{plot_max}In time, the trace points are sorted from high to low in quality, and only N with high quality is output_{plot_max}And (6) dot trace.

The point trace aggregation processing is performed on the set of measured data in this embodiment, and a result of aggregation processing on data point traces of 50 consecutive antenna periods is shown in fig. 4, a result of averaging the multiple period point traces of 15 target point traces and an average statistical result of averaging the clutter point trace quality is shown in fig. 5, and a remaining point trace after partial false point trace elimination according to the point trace quality is shown in fig. 6.

Step three: a search path is formed based on the dynamically hypothesized cluster.

The sea object has slow moving speed, and the maximum value of the moving speed is set as V_maxFor all the traces of points formed in the k-th period, Plot_{k_i},i＝1,…,N_{plot_k}，N_{plot_k}The current point trace Plot is used as the number of the point traces in the k period_{k_i}Position-centered, with V_maxT is a radius to establish a circular wave gate, and T is an antenna scanning period; in thatCorrelating the search paths formed by the first k-1 periods in the wave gate, wherein the correlation criterion is to select the path with the maximum cost function value in the wave gate for correlation, namely psi_k(x_k)＝argmax(I(x_k-1) Wherein I (x)_k-1)＝Q_{plot_k-1}+max(I(x_k-2) Is a search path cost function, x_kFor the kth period trace information, Q_{plot_k-1}Associating trace point quality for the k-1 th period; the search path starting point is formed for all the traces of the 1 st period.

Step four: and extracting a multi-period joint detection judgment target.

Setting the number of joint detection cycles to N_cycleSetting a multi-period accumulation cost function detection threshold T for a search path with the duration reaching a joint detection period_IFor the search path satisfying the detection threshold, according to psi_k(·),k＝1,…,N_cycleExtracting the path position to form a flight path; after the flight paths are extracted, position comparison is carried out between every two flight paths generated in the current period, if the position is N_cycleWithin one period there is M_cycleAnd if the paths of the points in the periods are the same, merging the paths according to the principle that the maximum cost function is the optimal function, and finishing the extraction of the multi-period combined judgment target.

The path search and multi-cycle joint detection decision target extraction based on the dynamic hypothesis cluster are performed on the set of measured data in this embodiment, and the processing result is shown in fig. 7.

Claims (1)

1. A method for jointly detecting and tracking sea targets by using two-coordinate radar is characterized by comprising the following steps:

the method comprises the following steps: background self-adaptive threshold detection, namely judging that a local scene is a noise background, a clutter background or an adjacent strong target scene according to data in a reference window of a unit to be detected, self-adaptively selecting a detection criterion and a detection threshold, and performing initial threshold detection pretreatment on original video data;

step two: point trace extraction based on quality evaluation, two coordinate radars adopt M in the distance direction according to the sequence of distance first and orientation second_r/N_rCriterion sliding window detection of condensation, azimuth using M_a/N_aThe criterion sliding window detects the agglomeration, calculates the trace quality while the trace agglomeration process, and for each distance trace, the distance agglomeration quality is defined as Q_r＝N_{r_plot}-M_r，N_{r_plot}The number of distance units condensed into the current trace point; setting M to form the final trace_aA distance trace amplitude of A_{r_plot_i},i＝1,…,M_aThe quality of the dot trace after the azimuthal agglomeration is defined as

Wherein N is_{a_plot}Number of distance traces condensed as current trace, C_qIs a trace point quality normalization constant; counting the number of traces in each period of clutter region, and when the number of the traces exceeds the upper limit value N of the set number of the traces_{plot_max}Then, sorting and outputting the points according to the trace point quality from large to small;

step three: forming a search path based on the dynamic hypothesis cluster, and setting the maximum value of the motion speed of the sea target as V_maxFor all the traces of points formed in the k-th period, Plot_{k_i},i＝1,…,N_{plot_k}，N_{plot_k}The current point trace Plot is used as the number of the point traces in the k period_{k_i}Position-centered, with V_maxT is a radius to establish a circular wave gate, and T is an antenna scanning period; the search paths formed in k-1 periods before the correlation in the gate are associated, and the association criterion is to select the path with the maximum cost function value in the gate for association, namely psi_k(x_k)＝argmax(I(x_k-1) Wherein I (x)_k-1)＝Q_{plot_k-1}+max(I(x_k-2) Is a search path cost function, x_kFor the kth period trace information, Q_{plot_k-1}Associating trace point quality for the k-1 th period; forming a starting point of a search path for all the point traces of the 1 st period;

step four: extracting a multi-period joint detection judgment target, setting the number of joint detection periods to be N_cycleSetting a multi-period accumulation cost function detection threshold T for a search path with the duration reaching a joint detection period_IFor the search path satisfying the detection threshold, according to psi_k(·),k＝1,…,N_cycleExtracting the path position to form a flight path; after the flight paths are extracted, position comparison is carried out between every two flight paths generated in the current period, if the position is N_cycleWithin one period there is M_cycleAnd if the paths of the points in the periods are the same, merging the paths according to the principle that the maximum cost function is the optimal function, and finishing the extraction of the multi-period combined judgment target.

Images