CN103558597B - Based on weak target detection method in the sea clutter of spectrum kurtosis - Google Patents

Abstract

The invention provides a kind of based on weak target detection method in the sea clutter of spectrum kurtosis, the steps include: the echo sequence of (1) each distance element m-doppler spectral when Short Time Fourier Transform obtains; Pair (2) time, on m-doppler spectral, each frequency calculates its kurtosis value, obtains composing kurtosis; (3) CFAR detection method is utilized to obtain adaptive threshold; (4) spectrum kurtosis is compared at value and the adaptive threshold of each frequency, adjudicate target and whether exist.The present invention utilizes in single distance element, the short-term stationarity of sea clutter and the instantaneity of target echo signal, reduce strong clutter region to the impact of target detection, can effectively detect moving-target signal, be applicable to the moving target detect under low signal to noise ratio condition.

Description

Based on weak target detection method in the sea clutter of spectrum kurtosis

Technical field

The invention belongs to radar signal processing field, relate to the Detection of Small and dim targets under sea clutter background.In particular to a kind of based on weak target detection method in the sea clutter of spectrum kurtosis.The present invention is applicable to the bank base microwave radar systems of various relevant mechanism.

Background technology

Surface, sea usually seriously limits radar to the back scattering of radar emission signal and to coexist to naval vessel, aircraft, guided missile, navigation buoy and other and surface, sea the detectability of the target in a radar resolution element.These undesired signals are commonly referred to as sea clutter or extra large surface echo.Sea superficial objects detection technique is all extremely important in military and civilian field.Therefore domestic and international many researchists are devoted to the research of this technology always.Early stage radar system is directly sent to display the video information obtained, and clutter and noise is shown by amplitude form, completes the detection to target by the human eye observation of operator to display.Modern radar system can realize automatic detection and tracking target, only need provide a detection threshold, then makes according to decision rule the judgement whether target exist.But under strong clutter background (particularly sea echo), detection and tracking automatically reliably can not be carried out to weak signal target.Although the detection probability to Small object can be improved by reduction detection threshold, false-alarm probability must be increased like this.

Kurtosis (Kurtosis) is a kind of physical quantity of characterization graph steep, can reflect compared with normal distribution curve, the sharpness of stochastic variable distribution curve or flatness.When curve becomes sharp-pointed, kurtosis value becomes large, and curve becomes smooth, and kurtosis value diminishes, and thus has and gets low value to Gaussian noise, impact signal is got to the characteristic of high level.Kurtosis, as a kind of statistical tool, in the status monitoring that noise is less, can utilize it to the exception response of the sensitivity Detection system of singular signal.But it cannot reflect the situation of change of particular signal component as a general indices, be not suitable for the status monitoring problem under strong noise environment.For overcoming the deficiency of kurtosis in engineer applied, Dwyer(1984) first proposed spectrum kurtosis (Spectral Kurtosis, SK) theory, and define the normalization Fourth-order moment that SK is the real part of Instant Fourier Transform.Its basic ideas calculate its kurtosis after carrying out bandpass filtering to signal different frequency range, not only can indicate the non-gauss component in signal, and can provide their frequency domain position.Antoni J.(2006) to this has been further investigation, setting forth the theoretical background of spectrum kurtosis diagnosis mechanical fault, having described the existing achievement in this field, give the formal definition of spectrum kurtosis, and successfully spectrum of use kurtosis method has diagnosed actual machine fault.

At present, spectrum kurtosis method is successfully applied to the signal transacting fields such as mechanical fault diagnosis, rarely has its application in target detection.The present invention utilizes the superiority of spectrum kurtosis method in nonstationary random response, applies it in the small target deteection under sea clutter background, effectively can improve the detection perform of radar to sea weak signal target.

Summary of the invention

The object of the invention is to: based on the bank base microwave radar systems of reality, under being provided in sea clutter jamming pattern, be applicable to the sea surface small target self-adapting detecting method based on spectrum kurtosis of Weak target.

For achieving the above object, the present invention adopts following technical scheme:

Step 1, by the echo sequence of each distance element m-doppler spectral when Short Time Fourier Transform obtains;

Step 2, pair time m-doppler spectral on each frequency calculate its kurtosis value, obtain compose kurtosis;

Step 3, using the kurtosis value of each frequency as detection statistic, and utilizes CFAR Methods to obtain adaptive threshold;

Step 4, compares spectrum kurtosis at value and the adaptive threshold of each frequency, and whether judgement target exists, if kurtosis value is greater than adaptive threshold, then judge that target exists, otherwise then judgement target does not exist.

In described step 1, the coherent accumulation time of Short Time Fourier Transform is 0.5s, sliding window process zero lap region.

In described step 2, calculate spectrum kurtosis SK (f _i) formula as follows,

$\mathrm{SK}\left(f_i\right)=\frac{\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_i\right)\right|}^4}{{\left[\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_i\right)\right|}^2\right]}^2}-2,i=\mathrm{1,2},\·\·\·,N$

Wherein, N represents frequency number, and M represents time quantum number, S _j() represents the doppler spectral on a jth time quantum, f _irepresent Doppler's frequency.

In described step 3, adaptive threshold defining method is CA-CFAR detection method, and detailed process is:

For frequency i, about it, define protected location, reference unit respectively, ask for the average μ of reference unit _i, for given false-alarm probability p _fa, thresholding weighted value α has with it following relation:

${\stackrel{-}{P}}_{\mathrm{fa}}={(1+\frac{\mathrm{\α}}{L})}^{-L}$

Wherein, L represents number of reference, then adaptive threshold T _ibe expressed as T _i=α × μ _i.

Compared with prior art, tool of the present invention has the following advantages and beneficial effect:

1. the present invention effectively can suppress the impact of sea clutter, reduces radar in target detection process and, to the demand of signal to noise ratio, thus has the ability detecting weak target in strong sea clutter background.

2. detection statistic of the present invention calculates easy, is easy to realize real-time process.

3. the present invention can the accurate frequency position of localizing objects signal in doppler spectral, thus can the radial velocity of effective estimating target in subsequent treatment, for target following provides useful information.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention.

Fig. 2 is the energy profile of experimental data.

Fig. 3 is the check point mark figure that the present invention is applied to experimental data gained, and its orbicular spot represents truly puts mark, and little fork represents False Intersection Points mark.

Fig. 4 is under identical false-alarm probability condition, the CA-CFAR detecting device of frequency domain and detection perform comparison diagram of the present invention.

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described.

The sea pulse echo signal that radar receives can be modeled as following form:

s(t)＝c(t)+x(t)+n(t)

Wherein, c (t) represents sea clutter, can regard steady non-Gaussian signal as at short notice; X (t) represents echo signal, and in probe unit, the residence time is short, can regard transient state non-stationary signal as; N (t) represents additive noise, is caused, meet Gaussian distribution by receiver internal noise etc.

Radar detects sea a period of time under resident mode of operation, receives and stores I/Q two-way echo data as radar experimental data.Echo-pulse obtains A type spectrum through first time Fast Fourier Transform (FFT), therefrom decomposites the echo data of each distance element, to treat further process below.

Step 1, sets the suitable coherent accumulation time, single distance element echoed signal is divided into M part, makes Short Time Fourier Transform, obtains one group of doppler spectral:

S _i(f)＝C _i(f)+X _i(f)+N _i(f)，i＝1,2,…M

Step 2, according to following formulae discovery spectrum kurtosis:

$\mathrm{SK}\left(f_i\right)=\frac{\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_i\right)\right|}^4}{{\left[\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_i\right)\right|}^2\right]}^2}-2,i=\mathrm{1,2},\·\·\·,N$

Wherein, N represents frequency number, and M represents time quantum number, S _j() represents the doppler spectral on a jth time quantum, f _irepresent Doppler's frequency;

Step 3, for spectrum kurtosis sequence SK (f _i), using the kurtosis value of each frequency as detection statistic, and utilize CFAR Methods to obtain adaptive threshold;

The value γ of detection statistic _ifor

γ _i＝|SK(f _i)|

Adaptive threshold defining method is CA-CFAR (CA-CFAR) detection method; For frequency i, about it, define protected location, reference unit respectively, ask for the average μ of reference unit _i, for given false-alarm probability p _fa, thresholding weighted value α has with it following relation:

${\stackrel{-}{P}}_{\mathrm{fa}}={(1+\frac{\mathrm{\α}}{L})}^{-L}$

Wherein, L represents number of reference.

The false-alarm probability that setting needs, calculates weighted value according to the relation between false-alarm probability and thresholding weighted value, thus obtains adaptive threshold:

T _i＝α×μ _i

Step 4, based on the detection statistic obtained and adaptive threshold, for the radar return data of actual measurement, its judgement is as follows:

Then, jump to next distance element, repeat above-mentioned steps 1-4, until process all distance elements to be detected, gather gained testing result and estimate the relevant parameters of target motion.

Effect of the present invention can be verified further by experiment below.Testing radar return data used is that the S-band ocean telegauge of independent development will obtain in the end of the year 2012, radar major parameter is as follows: transmission frequency is 2.85GHz, range resolution is 7.5m, pulse repetition rate is 256Hz, scanning impulse number is 25600, range unit number is 40, wherein, echo signal in detection time altogether through 35 distance elements.The false-alarm probability of the present invention's setting is p _fa=10 ^-3.

Fig. 2 be experiment echo data energy time m-distance dimension on distribution plan.Can see obvious sea echo striped, strong sea clutter makes the fuzzy difficulty of target echo signal distinguish, seriously limits the detection perform of sea-surface target.Fig. 3 is that application the present invention detects the target echo point mark and radial velocity estimation thereof that obtain, and its orbicular spot represents truly puts mark, and little fork represents False Intersection Points mark.Show thus, the present invention effectively can detect target echo, and velocity estimation is reliable and stable.Fig. 4 is the detection perform comparison diagram utilizing real pure sea clutter data under similar sea state conditions and simulation objectives signal acquisition, and image gives under different signal to noise ratio condition, and CA-CFAR method and the detection perform of the present invention of frequency domain compare.As can be seen here, detection perform of the present invention is better than the detection perform of frequency domain CA-CFAR detecting device.This is mainly because the present invention effectively make use of the short-term stationarity of sea clutter data, and the kurtosis value making the strong sea clutter region in doppler spectral obtain is significantly less than the kurtosis value of fast-opening target echoed signal place frequency domain, thus effectively distinguishes echo signal.

Claims (2)

1., based on a weak target detection method in the sea clutter of spectrum kurtosis, it is characterized in that: comprise,

Step 1, by the echo sequence of each distance element m-doppler spectral when Short Time Fourier Transform obtains;

Step 2, pair time m-doppler spectral on each frequency calculate its kurtosis value, obtain compose kurtosis;

Step 3, using the kurtosis value of each frequency as detection statistic, and utilizes CFAR Methods to obtain adaptive threshold;

Whether step 4, compares spectrum kurtosis at value and the adaptive threshold of each frequency, adjudicate target and exist; If kurtosis value is greater than adaptive threshold, then judge that target exists, otherwise, then judge that target does not exist;

In described step 1, the coherent accumulation time of Short Time Fourier Transform is 0.5s, sliding window process zero lap region;

In described step 2, calculate spectrum kurtosis SK (f _i) formula as follows,

$\left(f_i\right)=\frac{\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_j\right)\right|}^4}{{\left[\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\left|S_j\left(f_i\right)\right|}^2\right]}^2}-2,i=\mathrm{1,2},...,N$

Wherein, N represents frequency number, and M represents time quantum number, S _j() represents the doppler spectral on a jth time quantum, f _irepresent Doppler's frequency.

It is 2. according to claim 1 that based on weak target detection method in the sea clutter of spectrum kurtosis, it is characterized in that: in described step 3, adaptive threshold defining method is CA-CFAR detection method, and detailed process is as follows,

For frequency i, about it, define protected location, reference unit respectively, ask for the average μ of reference unit _i, for given false-alarm probability p _fa, thresholding weighted value α has with it following relation:

${\stackrel{\‾}{P}}_{\mathrm{fa}}={(1+\frac{\mathrm{\α}}{L})}^{-L}$

Wherein, L represents number of reference, then adaptive threshold T _ibe expressed as: T _i=α × μ _i.