CN104914422A - Adaptive TBD radar weak target detection method - Google Patents

Abstract

The present invention relates to an adaptive TBD radar weak target detection method which is mainly applied to the detection of a heavy clutter zone weak group target in a sea guard radar. For an original data, according to the needed detection probability and false alarm rate, a first threshold is set, points which are lower than the threshold are abandoned, and the coordinates and amplitudes of points which are higher than the threshold are recorded. After 6 frames of data are accumulated, all recorded points are subjected to Hough transform, in the process of carrying out target detection in a transform domain, the accumulated value is determined according to the signal to noise ratio of an echo, energy accumulation is employed for a low signal to noise ratio, binary accumulation is employed for a high signal to noise ratio, amplitude accumulation is employed for a middle signal to noise ratio, a second threshold is set, the target is announced when the threshold is exceeded, a related point is found through inverse transformation, at the same time the track of the target is announced, the binary accumulation is employed for the high signal to noise ratio, the energy accumulation is employed for the low signal to noise ratio, and the amplitude accumulation is employed for the middle signal to noise ratio.

Description

A kind of self-adaptation TBD radar Weak target detecting method

Technical field

The invention belongs to sea guard radar object detection field, particularly heavy clutter district multiple targets based on tracking technique before the detection of Hough transform.

Background technology

Dim targets detection under clutter environment is the significant challenge that modern sensor faces, traditional detection technique detects every frame signal, and by not having the data associated to lose, owing to not accumulating the echo information of interframe, be difficult to improve the detection perform to Small object.And based on following the tracks of (TBD) technology before the detection of Hough transform, achieving the interframe non-inherent accumulation of target, can detection perform be improved.

1994, Hough transform is applied in search radar target detection by the people such as the Carlson of the U.S. first, propose the TBD algorithm based on Hough transform, and analyze the detection and tracking performance of this algorithm, give the computing formula of detection probability and false-alarm probability.Compared with classic method, the method has lot of advantages, as improve detection probability, solving range walk problem, being easier to realize; By emulation, for dissimilar target, signal to noise ratio (S/N ratio) can improve 4dB, but there is the problem that Small object covered by general objective, improved afterwards based in the binary integration device of HT, although can address this problem, performance declines to some extent, have the SNR of 3dB to improve for non-fluctuating target, strong fluctuating target has the SNR of 1dB to improve.

On the basis of Carlson work, the scholars such as Kabakchiev discuss in different clutter background situation, the detection and tracking performance of the combined detector that CFAR detects and Hough transform is formed.But above-mentioned algorithm all hypothetical target does linear uniform motion along radar radial direction, and within multiple antenna cycle, target echo does not cross over localizer unit.For the situation that target leap localizer unit is walked about, Garvanov proposes a kind of polar coordinates Hough transform TBD algorithm, directly processes the radar measurement data under azimuth-range dimension polar coordinates, and this algorithm is more suitable for process radar measured data.2011, the Moyer of the U.S. proposed a kind of multidimensional HT-TBD algorithm for weak target under strong clutter background, and this algorithm can overcome the defect of traditional HT-TBD algorithm, strengthened the target detection probability under random noise background.

However, when adopting the TBD algorithm based on Hough transform to detect multiple targets, if the RCS difference between target is larger, large target can produce Small object covers, if reduction detection threshold, improve the detection probability to Small object, then can comprise more false-alarm in testing result.

Summary of the invention

On the basis of the present invention's track algorithm before heavy clutter district is based on the detection of the multiple targets of Hough transform, propose a kind of adaptive detection method.The amplitude range of target echo is larger, under different signal to noise ratio (S/N ratio) conditions, adopt different detection methods to improve detection probability, reduces false alarm rate.Traditional based on track algorithm before the detection of Hough transform, the echoed signal of time domain is transformed to parameter field, accumulates at parameter field, adopt the weights of single accumulation, be difficult in the clutter situation of change, obtain best detection perform.Herein by the adaptive mode selecting accumulation, effectively solve target occlusion and the low problem of detection probability.

Accompanying drawing explanation

Accompanying drawing 1 is trace flow before the self-adapting detecting based on Hough transform.

Embodiment

Tracking before the adaptive detection of clutter district multiple targets, concrete implementation step following (see accompanying drawing 1):

1. the echo data of continuous 6 frames of sequential storage surveyed area.

Hough transform will carry out multiframe data processing, and then target echo accumulation is inadequate very little for frame number, and target detection probability is low.If frame number is too many, then cause the problems such as memory space is large, calculated amount is large, false-alarm is high.Choosing comprehensively selects 6 frame data Combined Treatment.

2. calculate the signal to noise ratio (S/N ratio)/signal to noise ratio of echo.

The computing formula of the signal to noise ratio (S/N ratio) of echo is SNR=E/N ₀, wherein E is the power of target echo, N ₀it is Carrier To Noise Power Density.

3. filter raw radar data.

If all carry out Hough transform to every frame data of luv space, calculated amount will be very huge.In fact raw radar data contains a large amount of noises, adopts one-level thresholding to filter luv space data, only carries out Hough transform to the data crossing thresholding.Calculated amount increases with one-level thresholding and declines.Lower one-level thresholding, owing to having processed more data, improve detection perform, but calculated amount is large.

In distance to a declared goal-time domain, the initial false-alarm probability of each resolution element is p _f, then:

$p_f={\∫}_{\mathrm{\η}}^{\mathrm{\∞}}p\left(x\right)dx={\∫}_{\mathrm{\η}}^{\mathrm{\∞}}{e}^{-x}dx=e^{-\mathrm{\η}}$

Thus can to obtain one-level thresholding be η=-ln (p _f), each resolution element detection probability p _dfor:

$p_d={\∫}_{\mathrm{\η}}^{\mathrm{\∞}}p\left(x\right|S)dx={\∫}_{\mathrm{\η}}^{\mathrm{\∞}}\frac{e^{-x}/(1+S)}{1+S}dx=e^{-\mathrm{\η}/(1+S)}$

In formula, S represents signal to noise ratio (S/N ratio), as S=0, and can from p _dobtain p _f.

4. adaptive H ough converts

By the data Hough transform of η of one-level thresholding excessively all in distance verses time territory to parameter field, when parameter field is weighted accumulation, according to the signal to noise ratio (S/N ratio) that step 2 calculates, according to the selection weighting accumulative means in the interval at signal to noise ratio (S/N ratio) place.By the detection perform curve of Monte-Carlo Simulation three kinds of accumulative means along with signal to noise ratio (S/N ratio), signal to noise ratio, between the application area determining three kinds of accumulative means.Signal amplitude accumulates: using the echo amplitude of each point as accumulation item; Binary integration: having crossed one-level thresholding is then 1, but is then 0; Energy accumulation: accumulate again after the amplitude square of signal.

5. detection probability and false alarm rate calculate

Suppose that the data in distance verses time territory have N number of unit to parameter field contributive, wherein have the data (x of m unit _i, i=1 ..., probability m) having exceeded initial threshold η is and this m unit sum must be over the second thresholding ξ, namely in parameter field, a certain unit ζ more than the probability of the second thresholding ξ is:

$\mathrm{Pr}(\mathrm{\ζ}>\mathrm{\ξ})=\underset{m=1}{\overset{N}{\Σ}}{C}_N^m{\left(p_d\right)}^m{(1-p_d)}^{N-m}\×\mathrm{Pr}(y=\underset{i=1}{\overset{m}{\Σ}}{x}_i>\mathrm{\ξ})$

Carry out approximate processing to the Section 2 of above formula, finally obtaining detection probability is:

$P_D=1-{(1-p_d)}^N-\underset{m=1}{\overset{m^{*}-1}{\Σ}}{C}_N^m{\left(p_d\right)}^m{(1-p_d)}^{N-m}\×\[1-e^{-\frac{\mathrm{\ξ}-m\mathrm{\η}}{1+S}}\underset{k=0}{\overset{m-1}{\Σ}}{\left(\frac{\mathrm{\ξ}-m\mathrm{\η}}{1+S}\right)}^k/k!\]$

M in formula ^*=ceil (ξ/η) is for being not less than the smallest positive integral of ξ/η. and when S is 0, can obtain false-alarm probability is:

$P_F=1-{(1-p_f)}^N-\underset{m=1}{\overset{m*-1}{\Σ}}{C}_N^m{\left(p_f\right)}^m{(1-p_f)}^{N-m}\×\[1-e^{-(\mathrm{\ξ}-m\mathrm{\η})}\underset{k=0}{\overset{m-1}{\Σ}}{(\mathrm{\ξ}-m\mathrm{\η})}^k/k!\]$

6. the parameter corresponding according to check point, the track of backtracking target, announces testing result.

For the ease of data processing, by parameter space discretize, form grid (unit) one by one, owing to there being the coordinate points of Cartesian coordinates that the Curves of common intersection is corresponding must be point-blank in parameter space, second thresholding ξ is set in parameter field again, as final detection threshold.By inverse Hough transform, check point mapped time data space of parameter space, the targetpath of formation time-distance plane, can realize the detection to target.

Claims (3)

1. a self-adaptation TBD radar Weak target detecting method, it is characterized in that: testing result is not announced to single frames target, by Hough transform Combined Treatment multiframe data, in the mode of transform domain according to the adaptive selection weighting accumulation of the signal to noise ratio (S/N ratio) of echo: high s/n ratio adopts binary weighting accumulation, middle signal to noise ratio (S/N ratio) adopts amplitude weighting accumulation, and low signal-to-noise ratio adopts the weighting accumulation of energy.

2. self-adaptation TBD radar Weak target detecting method according to claim 1, is characterized in that: the determination methods in the height of signal to noise ratio (S/N ratio), for determining two threshold values, is greater than high threshold, is judged to be high s/n ratio situation; Be less than threshold ones and be judged to be low signal-to-noise ratio; Be in the middle of two threshold values is middle signal to noise ratio (S/N ratio).

3. self-adaptation TBD radar Weak target detecting method according to claim 1 and 2, it is characterized in that the defining method of described threshold value is: in the clutter meeting K distribution, add two batches of varying strength targets, detection probability under three kinds of weighting accumulative means is drawn by Monte-Carlo Simulation, false alarm rate, with the change curve of signal to noise ratio (S/N ratio), determines the threshold value of two signal to noise ratio (S/N ratio)s by the distribution of curve.