CN104808173B - Hough transformation-based false point elimination method for direction-finding cross location system - Google Patents

Abstract

The invention discloses a Hough transformation-based false point elimination method for a direction-finding cross location system. The Hough transformation-based false point elimination method is used for realizing the elimination of most false intersection points through Hough transformation and eliminating false intersection points having the same characteristics as those of target intersection points through adjacent period comparison. The method disclosed by the invention comprises the following steps: (1) analyzing the quantity and coincidence characteristics of cross location points; (2) calculating the maximum number of elements divided in a parameter space through binary hypothesis testing; (3) carrying out Hough transformation on the intersection points, voting in a corresponding accumulation unit and recording the numbers of the accumulation units and corresponding voting quantity; (4) establishing dual thresholds and gradually eliminating independent intersection points and intersection points having low coincidence number; (5) reducing an intersection point state corresponding to the accumulation unit sequence through Hough inverse transformation; (6) comparing the intersection points reserved in the adjacent periods, and further eliminating false points. The method disclosed by the invention is small in calculated amount, high in positioning accuracy and easy in engineering realization.

Description

Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform

Technical field

The invention belongs to sensing data process field is it is adaptable to passive passive sensor Passive Bearing-Crossing Location Systems are to many The removal in false cross point in object tracking process.

Background technology

Passive location system concealment is good, strong antijamming capability, operating distance are remote, a kind of the most frequently used passive location side Method is the positioning carrying out target using the angle measurement information of radiation source, and direction cross positioning is the main of Multi-Station passive location One of mode, can be effective against disturbing.Passive sensor obtains target according to the angle information recording by cross bearing Positional information, but, when this system is in target-rich environment, serious false cross point will be faced and eliminate problem.

False cross point is determined by the relative position of sensor and target, and has certain between different observation cycles Changing Pattern, the observation to real goal track causes very big upset and fascination, how effectively to remove these falsenesses and intersects Point, existing method mainly has：

(1) minimum distance method.Simply effective on this theoretical method, but when number of sensors is more, amount of calculation is in Index increases, and is unsuitable for number of sensors and is more than the practical application in the case of 3；

(2) azimuth slightly associate, thin correlation method.The method is identical with the principle of minimum distance method, presses sensor first Order positions to target two-by-two, calculates the distance between anchor point, retains distance and less mark respective sensor measures, then These sensors are measured and by all possible combination of two, target is positioned.The drawbacks of the method is, thick association process The effective data of discard portion is although greatly reduce amount of calculation, but is easy to navigate to false cross point.

Content of the invention

1. technical problem to be solved

The purpose of the present invention falls into identical accumulative element according to the cross point overlapping in parameter space, proposes one kind and is based on The Passive Bearing-Crossing Location Systems False Intersection Points removing method of Hough transform, can solve the problems, such as quantity of information in the elimination of false cross point Less, the poor problem of computationally intensive, locating effect.

2. technical scheme

Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform proposed by the present invention it is characterised in that Including following technical measures：

Step one, direction finding cross point quantity, coincidence situation analysis：Calculate the intersection of corresponding target by geometric triangulation relation The quantity of point, these points overlap, and point situation discusses the situation that false cross point overlaps, and concrete measure is：

M target followed the tracks of by 2 passive sensors of n ＞, and every sensor obtains multiple orientation angle measurements, and sensor is handed over two-by-two Fork positioning, at most obtainsIndividual cross point, the cross point of wherein corresponding target overlaps, and false cross point is big Mostly it is isolated, but there is also the situation of coincidence：

(1) n ＞ m, at False Intersection Points, maximum points of intersecting are

(2) n=m, at False Intersection Points, maximum points of intersecting areAnd the point of this kind of situation is only possible to there is one；

(3) n ＜ m, at False Intersection Points, maximum points of intersecting areAnd the point of this kind of situation there may be multiple；

Step 2, the initialization of Hough transform data processing parameters：According to sensing station and target location, every sensing Device obtains the angle measurement of multiple targets, and the angle measurement error according to sensor is split to the angle dimension of parameter space, Determine the unit number of segmentation, concrete measure is：

For the position coordinateses of i-th sensor,For i-th sensor error in measurement variance, D_ijFor i-th and Jth portion sensor base line length,Orientation angle measurements for i-th sensor；ρ-θ is that Hough transform parameter is empty Between, willIndividual point homologous thread is voted in accumulative element, and corresponding ballot matrix is A_l, R_maxFor single cross point in ginseng Quantity space respective distances maximum, R_maxFor single cross point in parameter space respective distances maximum；Δ θ, Δ ρ are ρ-θ parameter The cell length of side of space accumulative element, wherein, Δ θ=π/N_θ, Δ ρ=(R_max-R_min)/N_ρ, N_θFor the segmentation hop count of parameter θ, N_ρFor the segmentation hop count of parameter ρ, Thr is binary hypothesis test thresholding.

Step 3, cross bearing：Sensor is grouped two-by-two, according to the length of base between sensing station, sensor and many The angle measurement of individual target, obtains coordinate under cartesian coordinate system for the cross point, and concrete measure is：

(1) set up memory array Store (i, j)；

(2) sensor is grouped two-by-two, every group has two groups of dataWith

(3) two groups of data respectively take an element to combine D_ijTriangle polyester fibre, location dataP=m²It is stored in memorizer Array

Step 4, cross point project to parameter space, curve obtained C through Hough transform_lCarry out in corresponding accumulative element Ballot accumulation：Error in measurement according to sensor calculates accumulative element distance for the largest unit number dividing, and makes corresponding same mesh The cross point of punctuate can fall into identical accumulative element, is sequentially recorded the list of the distance dimension of the accumulative element that each point falls into Unit number, obtains sequence L_l, after homologous thread ballot in all cross points finishes, record each sequence L_lCorresponding accumulating value, obtains sequence Row H_l, concrete measure is：

(1) accumulative element unit number

Set up hypothesis testing：

H₀:The cross point coming from same target corresponds to identical accumulative element completely

H₁:The cross point coming from same target not exclusively corresponds to identical accumulative element

According to sensor error in measurement and geometry of position relation, during decision-making thresholding κ ＞ Thr, come from the intersection energy of same target Enough fall into identical accumulative element, and calculate the maximum division unit number of accumulative element accordingly

(2) parameter space divides

Parameter space ρ-θ is carried out being divided into accumulative element, ρ=[ρ₁,ρ₂,...,ρ_n], θ=[θ₁,θ₂,...,θ_n] each The central point of accumulative element is

Wherein,

(3) accumulative element accumulator and memorizer are set up

Set up accumulative element accumulator P (ρ, θ), putting each unit is 0, sets up memory array Memory (ρ, θ), puts and deposit Storage unit pointer is 1：Index (ρ, θ)=1；

(4) by cross bearing state { X_ijBe mapped in parameter space through Hough transform, that is, shape in cartesian space State { X_ijEach of status dataAccording to following Hough transform equations turned to parameter space

Defining corresponding parameter curve is C_l；

(5) ballot accumulation and status data storage

In curve C_lCorresponding accumulative element is voted, and the unit number of residing ρ accumulative element is recorded as sequence L_l, records series L_lEach accumulative element on accumulating value become sequence H_l, simultaneously by status dataIt is stored in corresponding Memory element Memory (ρ, θ) in；

(6) (3)～(5) are repeated until all positioning dotted state are mapped in parameter space.

Step 5, the judgement of double threshold False Intersection Points：According to accumulative element accumulating value, eliminate intersection independent or that coincidence number is few Point homologous thread, the curve of reservation, according to the similar degree of accumulative element number, determines the corresponding curve of target, and concrete measure is：

(1) set thresholding Thr₁, calculate votesT=1,2 ..., N_ρNumber, when number exceedes certain value During Q, retain the corresponding unit l that votes；

(2) reset thresholding Thr₂, the recording unit retaining is compared, if coincidence number Q^*＞ Thr₂, retain corresponding Ballot unit, otherwise, give up.

Step 6, Hough inverse transformation：After step 5, obtain effective parametric lineIn corresponding memory cellThe corresponding status data of middle readingIt is finally completed effective cross point from parameter space to Descartes The mapping in space；

Step 7 and target are intersected dot characteristics identical falseness cross point and are eliminated：

The false cross point that cannot distinguish is there may be with real goal cross point, from single observation week in cross bearing point Phase cannot remove, and can be able to dissipate to differentiate according to different observation cycles stubbornness cross point.

3. beneficial effect

(1) the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform that the present invention adopts is fixed by intersecting Site carries out batch processing it is not necessary to calculate the distance between the different cross points of different sensors group, substantially reduces amount of calculation, holds Easily Project Realization.

(2) using always most specific, in conjunction with cartesian space same coordinate position in the cross point of corresponding real goal Point falls into the characteristic of identical accumulative element after Hough transform, is filtered by thresholding and similarity-rough set retains real goal Corresponding cross point.

Brief description

Accompanying drawing 1 is that the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform proposed by the present invention is overall Flow chart；

Accompanying drawing 2 is the cross bearing point mark figure of signal period in the embodiment of the present invention (taking the cycle 3 as a example)；

Accompanying drawing 3 is signal period in the embodiment of the present invention (taking the cycle 3 as a example) accumulative element accumulation rectangular histogram；

Accompanying drawing 4 is point mark figure after signal period in the embodiment of the present invention (taking the cycle 3 as a example) Hough inverse transformation；

Accompanying drawing 5 is that in the embodiment of the present invention, Hough transform intersects the similar False Intersection Points judgement flow chart of dot characteristics to target；

Accompanying drawing 6 is 8 target cycle real motion trajectory diagrams in the embodiment of the present invention；

Accompanying drawing 7 is 8 period crossover anchor point mark figures of sensor in the embodiment of the present invention；

Accompanying drawing 8 is the Targets Dots adopting Hough transform False Intersection Points to eliminate rear 8 cycles positioning in the embodiment of the present invention；

Specific embodiment

Embodiment condition：Assume three targets in X-Y plane linear uniform motion, the initial position of target 1 (7km, 1km), x direction speed and y direction speed are respectively (300m/s, 100m/s), the initial position (4.5km, 6.5km) of target 2, x Direction speed and y direction speed are (350m/s, 0m/s), initial position (4km, 4km), x direction speed and the y direction of target 3 Speed is (300m/s, 20m/s).Follow the tracks of targets using four passive passive sensors, the position coordinateses of sensor 1 (0km, 0km), the position coordinateses (5km, 0km) of sensor 2, the position coordinateses (15km, 0km) of sensor 3, the position coordinateses of sensor 4 (20km, 0km), the error in measurement of four sensors is all set as 0.1 °, and the scan period is T=1s, altogether 8 scanning weeks of emulation Phase, in Hough parameter space, distance-azimuth discrimination unit is (Δ ρ, 1 °), and Δ ρ has value lower limit in each cycle.

Below in conjunction with Figure of description 1, the present invention is disappeared based on the Passive Bearing-Crossing Location Systems False Intersection Points of Hough transform Except method is described in detail.With reference to Figure of description 1, the handling process of the present invention divides following steps：

Step 1：Direction finding intersects point analysiss

3 targets followed the tracks of by 4 passive sensors, and every sensor obtains 3 orientation angle measurements；Sensor is grouped friendship two-by-two Fork positioning, every group at most can obtain intersection points is N_ij=9, the intersection points altogether obtaining are up to N_sum=54, each mesh Intersection at mark is counted and isHerein, number of sensors is more than number of targets, intersects points and be to the maximum at False Intersection Points

According to geometric triangulation relation, calculate the orientation angle measurements that 4 sensors obtain

Every sensor can obtain 3 azimuthal measuring values, and all the sensors position data, metric data are sent into radar Data handling machine；

Step 2：Initialized according to simulated conditions

D₁₂=5km, D₁₃=15km, D₁₄=20km, D₂₃=10km, D₂₄=15km, D₃₄=5km；

θ=[0:Δθ:π], ρ=[0:Δρ:R_max]；

K=8；

Thr is binary hypothesis test thresholding；

Step 3：Cross bearing

(1) set up memory array Store (i, j)；

(2) sensor is grouped two-by-two, every group has two groups of dataWith

(3) two groups of data respectively take an element to combine D_ijTriangle polyester fibre, location dataP=1,2 ..., 8 storages At memory array (Fig. 2)

Step 4：Hough transform

The feature of Hough transform is cartesian space can be in collinear point and be transformed into parameter space to intersect at A bit, the accumulating value highest of this accumulative element residing for point；It is empty that the point that cartesian space is coincided with same coordinate is transformed into parameter Between overlap into a curve, residing for this curve, the accumulating value of accumulative element is identical and higher.

(1) accumulative element unit number

Set up hypothesis testing：

H₀:The cross point coming from same target corresponds to identical accumulative element completely

H₁:The cross point coming from same target not exclusively corresponds to identical accumulative element

According to sensor error in measurement and geometry of position relation, sensor i, j cross bearing is horizontal, the position error of vertical coordinate For：

Formula is divided according to Hough transform formula and accumulative element, has

Wherein,p,q By (x_ij,y_ij) determine.

By upper,

Construction association statisticses

λ_ijObey the χ that degree of freedom is 1²Distribution.4 sensors, 6 association statisticses of construction are simultaneously sued for peace, note

To 3 targets, continue summation, obtain statistic of test：

Statistic of test obeys the χ that average is 18²Distribution.Given rejection probability η, according to

Pr κ ＞ Thr | H₀}=η

Decision-making thresholding Thr can be tried to achieve, come from during decision-making thresholding κ ＞ Thr same target intersection can fall into identical amass Tired unit, and calculate the maximum division unit number of accumulative element accordingly

(2) parameter space divides

Parameter space ρ-θ is carried out being divided into accumulative element, ρ=[ρ₁,ρ₂,...,ρ_n], θ=[θ₁,θ₂,...,θ_n] each The central point of accumulative element is：

Wherein,

(3) accumulative element accumulator and memorizer are set up

Set up accumulative element accumulator P (ρ, θ), putting each unit is 0, sets up memory array Memory (ρ, θ), puts and deposit Storage unit pointer is 1：Index (ρ, θ)=1；

(4) by cross bearing state { X_ijBe mapped in parameter space through Hough transform, that is, shape in cartesian space State { X_ijEach of status dataAccording to following Hough transform equations turned to parameter space：

Defining corresponding parameter curve is C_l；

(5) ballot accumulation and status data storage

In curve C_lCorresponding accumulative element is voted, and is designated as A_l, and will build up on unit number and be recorded as sequence L_l(only need The unit number of the residing ρ accumulative element of record), by sequence L_lEach accumulative element on accumulating value be recorded as sequence H_l, simultaneously By status dataIt is stored in corresponding memory element Memory (ρ, θ)：

A_l→L_l

P (ρ, θ)=P (ρ, θ)+A_l

P(ρ,θ)→H_l

(6) (3)～(5) are repeated until all positioning dotted state are mapped in parameter space；

Obtain Hough transform accumulation rectangular histogram (Fig. 3) through above-mentioned process；

Step 5：False Intersection Points are adjudicated

(1) set thresholding Thr₁, calculate votesT=1,2 ..., N_ρNumber, when number exceedes certain value During Q, retain the corresponding unit l that votes；

(2) reset thresholding Thr₂, the recording unit retaining is compared, if coincidence number Q^*＞ Thr₂, retain corresponding Ballot unit, otherwise, give up.

Step 6：Hough inverse transformation

After step 6, obtain effective parametric lineIn corresponding memory cellMiddle reading phase The status data answeredIt is finally completed mapping (Fig. 4) from parameter space to cartesian space for effective cross point；

Step 7：Repeat step 2～step 6

The false cross point that cannot distinguish is there may be with real goal cross point, from single observation week in cross bearing point Phase cannot remove, and can be able to dissipate to differentiate according to different observation cycles stubbornness cross point, its result can be used for instructing flight path The process (Fig. 5) such as initiate and follow the tracks of.

The real motion track of three targets and the space layout of four sensors in 8 observation cycles are as shown in fig. 6, four Portion sensor obtains the cross point mark after orientation angle measurements cross bearing as shown in fig. 7, true by data processing of the present invention after Fixed Targets Dots and the Targets Dots that azimuth is thick, thin associated data determines after processing are as shown in Figure 8.

In embodiment condition, number of sensors is more than 3, and the data processing time of the present invention is about 0.05 second, is suitable for In practical implementation；The accuracy that after data processing, data retains is about 83% it is easy to further track initiation and tracking Deng data processing.

Claims (6)

1. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform, its feature comprises the following steps：

Step one, direction finding cross point quantity, coincidence situation analysis：Calculate the cross point of corresponding target by geometric triangulation relation Quantity, these points overlap, and point situation discusses the situation that false cross point overlaps；

Step 2, the initialization of Hough transform data processing parameters：According to sensing station and target location, every sensor obtains Obtain the angle measurement of multiple targets, the angle measurement error according to sensor is split to the angle dimension of parameter space, determines The unit number of segmentation；

Step 3, cross bearing：Sensor is grouped two-by-two, according to the length of base between sensing station, sensor and multiple mesh Target angle measurement, obtains coordinate under cartesian coordinate system for the cross point；

Step 4, cross point project to parameter space, curve obtained C through Hough transform_lVoted in corresponding accumulative element Accumulation：Error in measurement according to sensor calculates accumulative element distance for the largest unit number dividing, and makes corresponding same impact point Cross point can fall into identical accumulative element, be sequentially recorded each point fall into accumulative element distance dimension unit Number, obtain sequence L_l, after homologous thread ballot in all cross points finishes, record each sequence L_lCorresponding accumulating value, obtains sequence H_l；

Step 5, the judgement of double threshold False Intersection Points：According to accumulative element accumulating value, eliminate cross point pair independent or that coincidence number is few Answer curve, the curve of reservation, according to the similar degree of accumulative element number, determines the corresponding curve of target；

Step 6, Hough inverse transformation：After step 5, obtain effective parametric lineIn corresponding memory cellThe corresponding status data of middle readingIt is finally completed effective cross point from parameter space to Descartes The mapping in space；

Step 7 and target are intersected dot characteristics identical falseness cross point and are eliminated.

2. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform according to claim 1, its feature Direction finding described in step one intersects point analysiss

M target followed the tracks of by 2 passive sensors of n ＞, and every sensor obtains multiple orientation angle measurements, and sensor intersects fixed two-by-two Position, at most obtainsIndividual cross point, the cross point of wherein corresponding target overlaps, and false cross point is lonely mostly Vertical, but there is also the situation of coincidence：

(1) n ＞ m, at False Intersection Points, maximum points of intersecting are

(2) n=m, at False Intersection Points, maximum points of intersecting areAnd the point of this kind of situation is only possible to there is one；

(3) n ＜ m, at False Intersection Points, maximum points of intersecting areAnd the point of this kind of situation there may be multiple.

3. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform according to claim 1, its feature Hough transform data processing initial method described in step 2 is

For the position coordinateses of i-th sensor,For i-th sensor error in measurement variance, D_ijFor i-th and jth portion Sensor base line length,For the orientation angle measurements of i-th sensor, ρ-θ is Hough transform parameter space, willIndividual point homologous thread is voted in accumulative element, and corresponding ballot matrix is A_l, R_maxFor single cross point in parameter space Respective distances maximum, Δ θ, Δ ρ are the cell length of side of ρ-θ parameter space accumulative element, wherein, Δ θ=π/N_θ, Δ ρ= (R_max-R_min)/N_ρ, N_θFor the segmentation hop count of parameter θ, N_ρFor the segmentation hop count of parameter ρ, Thr is binary hypothesis test thresholding.

4. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform according to claim 1, its feature Curve after Hough transform for the cross point described in step 4 in the accumulation method of accumulative element is

(1) accumulative element unit number

Set up hypothesis testing：

H₀:The cross point coming from same target corresponds to identical accumulative element completely

H₁:The cross point coming from same target not exclusively corresponds to identical accumulative element

According to sensor error in measurement and geometry of position relation, the intersection coming from same target during decision-making thresholding κ ＞ Thr can fall Enter identical accumulative element, and calculate the maximum division unit number of accumulative element accordingly

(2) parameter space divides

Parameter space ρ-θ is carried out being divided into accumulative element, ρ=[ρ₁, ρ₂,...,ρ_n], θ=[θ₁,θ₂,...,θ_n] each accumulation The central point of unit is：

$\begin{array}{cc}{\mathrm{\θ}}_n=(\stackrel{\‾}{n}-1/2)\mathrm{\Δ}\mathrm{\θ}& \stackrel{\‾}{n}=1,2,...,N_{\mathrm{\θ}}\end{array}$

$\begin{array}{cc}{\mathrm{\ρ}}_n=(\stackrel{\‾}{n}-1/2)\mathrm{\Δ}\mathrm{\ρ}& \stackrel{\‾}{n}=1,2,...,N_{\mathrm{\ρ}}\end{array}$

Wherein,

(3) accumulative element accumulator and memorizer are set up

Set up accumulative element accumulator P (ρ, θ), putting each unit is 0, sets up memory array Memory (ρ, θ), puts storage single First pointer is 1：Index (ρ, θ)=1；

(4) by cross bearing state { X_ijBe mapped in parameter space through Hough transform, that is, state in cartesian space {X_ijEach of status dataAccording to following Hough transform equations turned to parameter space

$\mathrm{\ρ}=x_{ij}^{p}cos\mathrm{\θ}+y_{ij}^{p}sin\mathrm{\θ},\mathrm{\θ}\∈\[0,\mathrm{\π}\]$

Defining corresponding parameter curve is C_l；

(5) ballot accumulation and status data storage

In curve C_lCorresponding accumulative element is voted, and the unit number of residing ρ accumulative element is recorded as sequence L_l, note Record sequence L_lEach accumulative element on accumulating value become sequence H_l, simultaneously by status dataIt is stored in corresponding storage In unit Memory (ρ, θ)；

(6) (3)～(5) are repeated until all positioning dotted state are mapped in parameter space.

5. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform according to claim 1, its feature Falseness described in step 5 intersects decision method

(1) set thresholding Thr₁, calculate votesT=1,2 ..., N_ρNumber, when number exceedes certain value Q, Retain the corresponding unit l that votes；

(2) reset thresholding Thr₂, the recording unit retaining is compared, if coincidence number Q^*＞ Thr₂, retain corresponding throwing Ticket unit, otherwise, gives up.

6. the Passive Bearing-Crossing Location Systems False Intersection Points removing method based on Hough transform according to claim 1, its feature The false cross point removing method having identical characteristics with target described in step 7 is

The false cross point that cannot distinguish is there may be with real goal cross point, from single observation cycle no in cross bearing point Method removes, and can be able to dissipate to differentiate according to different observation cycles stubbornness cross point.