CN105488852A - Three-dimensional image splicing method based on geography coding and multidimensional calibration - Google Patents

Abstract

The invention discloses a three-dimensional image splicing method based on geography coding and multidimensional calibration, comprising steps of performing three-dimensional space position geography coding of the sonar image body data through towed body position calculation and sonar parameter correction, choosing an cruising-depth dimension slice from a multi-measurement-line superposition area image, obtaining the translation amount of the depth dimension through matching of the image characteristics, choosing a cruising-orientation dimension slice, calculating the translation amount and an angle rotation amount through an object long shaft image characteristic, and realizing the multi-dimension-calibration splicing of the three-dimensional image through the multi-dimension calibration of the geography coding of the three-dimension space. The invention performs splicing processing on the three-dimension image which is generated by the sonar platform and contain the four-dimension information of cruising position, the orientation position, the depth position and the reflection intensity value, and creatively promotes the three-dimensional image splicing method based on geography coding and multidimensional calibration.

Description

A kind of 3-D view joining method based on geocoding and multidimensional calibrating

Technical field

The invention belongs to the category of sonar image Data Post technology, mainly a kind of 3-D view joining method based on geocoding and multidimensional calibrating.

Background technology

Image mosaic technology refers to several have the image of lap to be combined into the technology of a large-scale seamless high-definition picture, is extensively present in the image processing field such as remote sensing, radar, medical science, optics, digital video.This technology mainly comprises image registration and image co-registration two parts, wherein image registration is the core of image mosaic, refer in particular to and adopt certain matching strategy, find out template in image to be spliced or the corresponding in a reference image position of unique point, and then determine the transformation relation between two width images.Registration Algorithm is broadly divided into the method for the method based on model, the method based on transform domain, the method for being correlated with based on gray scale and feature based.Because image mosaic technology does not have the test pattern of standard, also ununified evaluation criterion, the good and bad main subjective vision perception relying on people of measure algorithm.

Along with the fast development of signal processing technology and image processing techniques, the view data scale that image sonar produces develops into three-dimensional by two dimension: a class is multibeam echosounding equipment, be used as submarine topography imaging specially, generate pseudo-three-dimensional data (walk to navigate position X, position of orientation Y, height value H) type, still can adopt conventional two-dimensional connecting method, after superposition elevation information, form dimensional topography splicing effect; An other class is multi-beam synthetic aperture three-dimensional imaging sonar, it uses the acoustic method of advanced imitative Medical CT technology to realize water body, seabed, stratum three-dimensional imaging, creates the three-dimensional image volume data type containing four-dimensional information (walk to navigate position X, position of orientation Y, depth location Z, reflection intensity values P).When processing this type of and containing the 3-D view volume data of four-dimensional information, face following two class problems:

(1) joining method of comparatively ripe pseudo-three-dimensional image data (i.e. two 2-D datas) has been had at present both at home and abroad, Successful utilization is in the process of multibeam echosounder image mosaic, but rarely has the technique study that the 3-D view volume data containing four-dimensional information is spliced.In view of acoustic picture is compared with optics, electromagnetic image, due to formation mechenism and environment for use constrained, there is contrast weak, reverberation noise is many, image detail is deficient, the features such as resolving power is lower, the effect that the stitching algorithm of introducing traditional optical, electromagnetics image processing field is directly applied in acoustic picture splicing is unsatisfactory.

(2) sonar platforms is under the different course line of reality, is subject to the impact that the factor such as non-homogeneous skew in boat walked by wave, ocean current, tide and towboat, exists up and down, freely the changing of trim, roll attitude.Most attitude error can be compensated after motion compensation signal process and multiparameter correction, but under the factor such as the skew in course line and the impact of platform kinematic error can cause different survey line, the image bit of same target is equipped with relatively large deviation.Superposition is inlayed in direct employing locus, there will be the entanglement of multidimensional image locus, causes splicing erroneous results, and visual effect is inferior.

Summary of the invention

Object of the present invention is exactly to overcome above-mentioned problems of the prior art, and provides a kind of 3-D view joining method based on geocoding and multidimensional calibrating.

The object of the invention is to have come by following technical solution.This 3-D view joining method based on geocoding and multidimensional calibrating, first the three-dimensional space position geocoding of sonar image volume data is carried out by towed body dead reckoning, sonar parameter correction, then boat-depth dimension section is chosen away to many surveys line overlapping region image, by Image Feature Matching, obtain the translational movement of depth dimension; Choose away boat-azimuth dimension section, by target long axial images feature calculation translational movement and angle rotation amount, finally by the multidimensional calibrating of three-dimensional space position geocoding, realize the 3-D view splicing of multidimensional calibrating.

Content of the present invention is divided into three-dimensional space position geocoding, multidimensional to mate and geocoding is calibrated, 3-D view splices three parts, and each several part is described as follows:

(1) three-dimensional space position geocoding

Calculate sonar towed body position, revise the track course angle of towed body, fixing roll angle, sound velocity error, adjustment sonar image two dimension yardstick, obtain the geocoding of 3-D view volume data under cartesian space rectangular coordinate system under different survey line.

1. sonar towed body position is calculated: lash ship configures GPS navigation equipment, according to position relationship between lash ship, towed body, extrapolate sonar towed body longitude and latitude position data, then filtering process is carried out to towed body position data, reduce the locus positioning error because of the introducing of only a few sonar towed body position distortion.

2. towed body track course angle is calculated: with reference under the earth plane coordinate system, select towed body in the latitude and longitude value of different track points positions, space, ask for the arc tangent triangular transformation value of warp, difference of latitude in time span, obtain towed body track course heading, weaken because of towed body in the process of moving bow to the locus positioning error that disturbance is introduced.

3. compensate towed body and fix roll angle: the sensor roll output valve obtaining synchronous recording, is fixed roll angle angle value, compensates the locus positioning error introduced because towed body fixes roll angle after carrying out statistical average.

4. revise sound velocity error: the sonic velocity change value obtaining operation marine site after utilizing sound velocimeter to measure, revise the acoustic velocity value of respective depth, eliminate the locus positioning error introduced because sound field sound velocity error is comparatively large.

5. convergent-divergent two dimension yardstick: 3-D view volume data walk boat and azimuth dimension carry out two-dimentional scaling process, make away boat consistent with azimuth dimension graphical rule value, dispel the locus positioning error because of the inconsistent introducing of two-dimentional yardstick.

6. seabed plane latitude and longitude value: the geometric model fixing roll angle, sound velocity error, towed body height, beam angle, seabed involuting wave oblique distance according to sonar towed body position, towed body track course angle, towed body, obtains the latitude and longitude value of seabed plane in 3-D view volume data.

7. cartesian space right angle geocoding: set up rectangular coordinate system in space, the latitude and longitude value of the seabed plane of 3-D view volume data is configured to corresponding flat coordinate axis, and depth information is configured in vertical coordinate axle, make 3-D view volume data possess cartesian space right angle geocoding.

(2) multidimensional coupling and geocoding calibration

Choose away boat-depth dimension section for many surveys line overlapping region image, by Image Feature Matching, obtain the translational movement of depth dimension; Choose away boat-azimuth dimension section, by target long axial images feature calculation translational movement and angle rotation amount, obtain away the calibration value of boat-orientation-depth dimension, complete the multidimensional calibrating of three-dimensional space position geocoding (cartesian space right angle geocoding).

(3) 3-D view splicing

According to three-dimensional space position geocoding, complete the 3-D view splicing under different survey line.

Beneficial effect of the present invention is: the present invention is directed to that sonar platforms generates, containing the 3-D view volume data splicing of four-dimensional information (walk to navigate position, position of orientation, depth location, reflection intensity values), novelty proposes a kind of 3-D view joining method based on geocoding and multidimensional calibrating.

Accompanying drawing explanation

Fig. 1 is the 3-D view joining method FB(flow block) based on geocoding and multidimensional calibrating.

Fig. 2 is towed body position data filtering principle schematic diagram.

Under the different survey line of Fig. 3 same area 3 d image data overlook result (in water, water-bed target).

Bead target depth dimension coupling figure in water under the two survey line of Fig. 4.

Fig. 5 two survey line 3-D view splicing result (with water-bed target in water).

Fig. 6 without multidimensional coupling calibration 3-D view splicing result (target is buried on stratum).

Based on the rear result of two dimension coupling calibration walking boat/orientation of burying target under the two survey line of Fig. 7.

Fig. 8 splices result (target is buried on stratum) through the 3 d image data of geocoding and the calibration of multidimensional coupling.

Embodiment

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

This 3-D view joining method based on geocoding and multidimensional calibrating of the present invention, concrete steps are as follows:

101) three-dimensional space position geocoding

The three-dimensional space position geocoding of sonar image volume data is carried out by towed body dead reckoning, sonar parameter correction.

1 calculates sonar towed body position

Towed body position data filtering: choose adjacent three track points i successively, i+1, i+2 form triangle, with i, i+2 line is base, the vertical line equation that calculated i+1 point is crossing with base, try to achieve the coordinate of cut-point P with the ration of division of 1:2 and be assigned to track points i+1, can at the distance proportion of level and smooth flight path coordinate simultaneously preferably between maintenance track points, filtering signal is as shown in Figure 2.By longitude and latitude (WD, JD) the coordinate transformation planimetric rectangular coordinates value of track points i, i+1, i+2, conversion formula is:

Y＝WD·1852，X＝JD·cos(WD/60/180·π)·1852

The rectangular coordinate value of known i, i+2 point, can obtain the expression formula of straight line L1 slope intercept form y=kx+b and general expression A1x+B1y+C1=0, slope is mutually vertical with straight line L1 owing to crossing i+1, P point straight line L2, its slope is straight line L2 expression formula A2x+B2y+C2=0 can be obtained equally, calculate the intersection point Q coordinate figure of L1 and L2 and then according to an i+1 and the coordinate putting Q, adopt the fixed coordinate figure obtaining a P than formula of sectional type here select λ=0.5, the P point coordinate value of acquisition is assigned to track points i+1, becomes new coordinate figure.

2 calculate towed body track course angle: definition (WD, JD) be expressed as the latitude/longitude of towed body track points, the difference (dy, dx) of track points longitude and latitude in certain hour span done arc tangent conversion, obtain track course heading θ, the derivation of equation is as follows:

dy＝ΔWD·1852＝(WD _new-WD _old)·1852

dx＝ΔJD·1852·cos(WD _old/60/180*π)＝(JD _new-JD _old)·cos(WD _old/60/180*π)·1852

Track course angle: θ=arctan2 (abs (dx/dy))

3 compensate towed body fixes roll angle: the sensor output value obtaining synchronous recording, is fixed roll angle angle value γ after carrying out statistical average.

4 sound velocity errors: adopt sound velocimeter to measure the sonic velocity change value obtaining operation marine site, revise.

5 zoomed images two dimension yardsticks: 3-D view volume data walk boat and azimuth dimension carry out two-dimentional scaling process, select linear interpolation processing, the 3-D view volume data after process walk navigate consistent with azimuth dimension graphical rule value.

6 seabed plane longitudes and latitudes: the combination by three-dimensional image volume Data Placement being the two-dimensional image data under several beam angles.Utilize sonar towed body position, different beams angle [alpha], fixing roll angle γ, towed body track course angle θ and seabed interface echo oblique distance sr _α(after revising sound velocity error), the reference being obtained this wave cover sea-bed area by beam amplitude projection is flat apart from R (α)=sr _αcos (α-γ), obtains horizontal scanning line angle by the relation that horizontal scanning line is vertical with track course

Definition towed body is in discrete track points place, space latitude and longitude coordinates value then latitude and longitude value (the WD of seabed interface position under different beams angle ⁱ _α, JD ⁱ _α), wherein sign represents left side or the right side that seabed interface is positioned at towed body.

$\{\begin{array}{c}{{\mathrm{WD}}^i}_{\mathrm{\α}}={{\mathrm{WD}}^i}_{tow}-\frac{R\left(\mathrm{\α}\right)\·\cos \left({\mathrm{\θ}}_i\±\mathrm{\π}/2\right)}{1852}\\ {{\mathrm{JD}}^i}_{\mathrm{\α}}={{\mathrm{JD}}^i}_{tow}-\frac{R\left(\mathrm{\α}\right)\·\sin \left({\mathrm{\θ}}_i\±\mathrm{\π}/2\right)}{1852\·\cos \left({{\mathrm{WD}}^i}_{\mathrm{\α}}/60/180\·\mathrm{\π}\right)}\end{array},i=1,2,\mathrm{...}n$

7 cartesian space rectangular coordinate system codings: by three-dimensional image volume data transformation in cartesian space rectangular coordinate system, coordinate origin unification is walk to navigate the intersection location at initial cross section and beam positional angle, the left side, be x-axis forward along direction of walking to navigate, beam positional angular direction, the right is y-axis forward, and depth direction is z-axis forward.The latitude and longitude value of the seabed plane calculated in step 6 is corresponded in XOY plane coordinate figure, and set up the Z axis coordinate figure of view data in XOZ plane, finally all 3-D view volume data point all possess three dimensional space coordinate position and gray value information, i.e. M (x, y, z)=ff.

102) multidimensional coupling and geocoding calibration

Choose overlapping region image under many surveys line, obtain image notable feature, mainly comprise two classes: a class includes in water in image or the cable, gully, pipeline etc. in seabed have continuous distribution type target.Another kind of be include in water, seabed, strong reflection target in stratum or long strip type regular targets;

Obtain image notable feature and walk boat-degree of depth two dimension slicing figure in different images volume data, according to characteristics of image, obtain the degree of depth matched position z1 of target in different slice map, z2, and calculate depth dimension calibration value Δ z=z1-z2.

Obtain image notable feature in different images volume data, walk boat-orientation two dimension slicing figure, complete target and walk boat-azimuth dimension location matches, image volumetric data projects along Z-direction, obtain flat square XOY coordinate system, obtain the coordinate figure of target major axis at plane right-angle coordinate, set up straight line model y=kx+b, calculate straight line expression formula y=k respectively ₁x+b ₁and y=k ₂x+b ₂, convert out angle rotation amount Δ θ=tan ^-1k ₂-tan ^-1k ₁.And choose target major axis centre position point coordinate (x1, y1) in different slice maps, (x2, y2), calculates distance between two points:

$\mathrm{\Δ}d=\sqrt{{(x2-x1)}^2+{(y2-y1)}^2},$ The angle of line and X-axis: $\mathrm{\σ}={\tan }^{-1}\left(\frac{y2-y1}{x2-x1}\right),$ Obtain the projection components value g=Δ dcos σ of translational movement on X/Y axle, h=Δ dsin σ, according to angle rotation amount and translational component, obtains away boat-azimuth dimension calibration value:

Δx＝g+X·cos(Δθ)-Y·sin(Δθ)

Δy＝h+X·sin(Δθ)+Y·cos(Δθ)

Under completing corresponding survey line, 3-D view volume data geocoding in cartesian space rectangular coordinate system is calibrated:

M(x+Δx,y+Δy,z+Δz)＝ff

103) image volumetric data splicing is merged

3 d image data under different survey line splices according to geocoding, and the view data of overlapping region have employed gray-scale value and gets the fusion of large method.

Choose the 3 d image data of two surveys line, there is bead target in water, the bottom has heavy piece, 3 d image data overlook result as shown in Figure 3.

Every bar survey line data, after three-dimensional space position geocoding, choose the target that bead in water is aimed at as coupling, and by walking the analysis of boat-depth dimension image slice to the bead under different survey line, carry out depth calibration, result as shown in Figure 4.

To bead target depth dimension coupling in water under different survey line, after carrying out the calibration of three-dimensional space position geocoding, splicing as shown in Figure 5, blue portion is seawater, in water, yellow target is bead, the yellow target in seabed is counterweight stone, in water in bead enlarged drawing without diplopia, bottom target geometric configuration is regular, target walking boat, orientation and depth dimension position overlap substantially.

At sea lay and bury chlorine cylinder target, aimed dia 0.5 meter, length 2.0 meters, burying depth about 1.0 meters, choose two surveys line to pass through with different course and bury target area, the 3 d image data of generation after three-dimensional space position geocoding direct splicing result as shown in Figure 6, target all cannot overlap in the degree of depth, orientation, objective contour feature Fuzzy, is unfavorable for target interpretation.

Boat-the degree of depth slice map of walking of burying target choosing different survey line carries out degree of depth coupling, chooses away boat-orientation slice map and carries out walking the registration of boat-azimuth dimension as shown in Figure 7, corresponding angle θ ₁=-37.65 ° and θ ₂=-67.55 °, angle rotation amount Δ θ=θ ₁-θ ₂=30.1 °.

As shown in Figure 8, target substantially overlaps the splicing result of the 3 d image data after geocoding and the calibration of multidimensional coupling in the degree of depth and azimuth dimension, effectively interpretation can go out the dimensional profile features of target.

In addition to the implementation, the present invention can also have other embodiments, and all employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop on the protection domain of application claims.

Claims (2)

1. the 3-D view joining method based on geocoding and multidimensional calibrating, it is characterized in that: the three-dimensional space position geocoding of first being carried out sonar image volume data by towed body dead reckoning, sonar parameter correction, then boat-depth dimension section is chosen away to many surveys line overlapping region image, by Image Feature Matching, obtain the translational movement of depth dimension; Choose away boat-azimuth dimension section, by target long axial images feature calculation translational movement and angle rotation amount, finally by the multidimensional calibrating of three-dimensional space position geocoding, realize the 3-D view splicing of multidimensional calibrating.

2. the 3-D view joining method based on geocoding and multidimensional calibrating according to claim 1, it is characterized in that: in three-dimensional space position geocoding, calculate sonar towed body position, revise the track course angle of towed body, fixing roll angle, sound velocity error, adjustment sonar image two dimension yardstick, obtains the geocoding of 3-D view volume data under cartesian space rectangular coordinate system under different survey line.