CN105806315B - Noncooperative target relative measurement system and measuring method based on active coding information - Google Patents
Noncooperative target relative measurement system and measuring method based on active coding information Download PDFInfo
- Publication number
- CN105806315B CN105806315B CN201410844755.XA CN201410844755A CN105806315B CN 105806315 B CN105806315 B CN 105806315B CN 201410844755 A CN201410844755 A CN 201410844755A CN 105806315 B CN105806315 B CN 105806315B
- Authority
- CN
- China
- Prior art keywords
- noncooperative target
- dimensional
- coding information
- image
- target relative
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The present invention relates to noncooperative target relative measurement systems and its measuring method based on active coding information, measuring system includes structured light projector, third control unit control structure light projector projects the structure light by coding to noncooperative target, which is the light model with certain coding information;The structure light forms light model image on noncooperative target surface;First camera and second camera are separately positioned on the both sides of the structured light projector, are used to pickup light phantom images image;The light model imaged image that decoded first camera is shot is sent to data processing unit by the first control unit, the light model imaged image that decoded second camera is shot is sent to data processing unit by the second control unit, the three-dimensional reconstruction of image is realized by data processing unit, and then obtains noncooperative target relative position and relative attitude.The present invention relates to noncooperative target relative measurement system and its measuring method based on active coding information, high certainty of measurement, stability are strong.
Description
Technical field
The present invention relates to spacecraft relative measurement techniques, and in particular to a kind of noncooperative target based on active coding information
Relative measurement system and measuring method.
Background technology
It needs to obtain in Technique in Rendezvous and Docking task closely opposite between high-precision pursuit spacecraft and target satellite
The relative movement informations such as position, relative attitude angle, relative velocity and relative attitude angular speed, in the prior art using passive measurement
Mode, but passive measurement mode has the following disadvantages:Because the hypothesis such as similarity measurement, continuity and objective reality are there are deviation,
The dynamic of dynamic object image makes passive measurement mode be easier to lead to image masking and ambiguity due to visual angle change;Passively
Measurement method is illuminated by the light and the influence of veiling glare environment is big, high to target configuration requirement, difficulty of matching is big.These shortcomings cause passively to survey
Amount mode measurement accuracy is difficult to ensure that.
Invention content
The purpose of the present invention is to provide a kind of based on the noncooperative target relative measurement system of active coding information and survey
Amount method, high certainty of measurement, robustness are strong.
To achieve the above object, the present invention provides a kind of noncooperative target relative measurement based on active coding information
System, including sequentially connected first camera, the first imaging unit and the first control unit, sequentially connected second camera,
Two imaging units and the second control unit and data processing unit;It is characterized in that, the noncooperative target relative measurement system
System further includes structured light projector and third control unit, and the third control unit control structure light projector is to non-cooperative target
Light model of the mark projection with certain structured coding information;The light model forms light model image on noncooperative target surface;
The first camera and second camera are separately positioned on the both sides of the structured light projector, are used to pickup light phantom images figure
Picture;The light model imaged image of the decoded first camera shooting is sent at the data by first control unit
Unit is managed, the light model imaged image that the decoded second camera is shot is sent to the number by second control unit
According to processing unit, realize that image high-precision three-dimensional point cloud matching is detected and carried with reconstruction and feature by the data processing unit
It takes, and then obtains noncooperative target relative position and relative attitude.
The above-mentioned noncooperative target relative measurement system based on active coding information, wherein, the light model using it is pseudo- with
Machine series coding mode is encoded.
The above-mentioned noncooperative target relative measurement system based on active coding information, wherein, the light model coding uses
Single frames structure light coding.
The above-mentioned noncooperative target relative measurement system based on active coding information, wherein, the light model image includes
Noncooperative target surface each point depth information and coding information.
Another technical solution provided by the invention is a kind of noncooperative target relative motion letter based on active coding information
Cease measuring method, using the above-mentioned noncooperative target relative measurement system based on active coding information, the measuring method include with
Lower step:1)The light model imaged image and second camera that the data processing unit is shot based on decoded first camera are clapped
The light model imaged image taken the photograph carries out high-precision three-dimensional point cloud matching with rebuilding;2)To the noncooperative target 3-D view of reconstruction into
Row feature detects and extraction;The feature includes three-dimensional planar, three-dimensional straight line segment, three-dimensional circle and three-D profile;3)Establish non-conjunction
Make target body coordinate system, the change in coordinate axis direction according to the noncooperative target body coordinate system established decomposites noncooperative target
Three attitude angles just obtain position of the noncooperative target with respect to pursuit spacecraft and posture.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 1)
It is middle to realize noncooperative target high-precision three-dimensional point cloud matching with rebuilding using improved iteration closest approach method;Improved iteration is nearest
Point method is successively according to 4 selection with eka-element, the determining of registration strategies, the foundation of search strategy and the solution of error function moulds
Block is unfolded, the specific steps are:
A cloud is sampled using the method for average sample first, completes the selection with eka-element;
Then determining for registration strategies is completed by the selection of characteristic measure and the selection of search strategy;
Secondly non-fully corresponding point set is handled using multidimensional binary search tree dynamic fusion method, by being established to cloud
KD trees can improve the efficiency of search closest approach, promote the speed of ICP iterative algorithms, complete the foundation of search strategy;
Finally using the solution that error function is determined based on singular value decomposition method, noncooperative target high-precision three-dimensional point is completed
Cloud is matched and is rebuild.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 2)
In, the detection and extraction of three-dimensional planar include rough detection and extraction and essence detection and extraction;The rough detection and extraction use
The method that clustering methodology, curved surface growth method and plane act of union blend;The essence detection and extraction and application step 1)It obtains
Three-dimensional point cloud, plane fitting is carried out to points of some parts, is found using the obtained equation of fitting and is obtained positioned at local fit
Plane in point, fitting again obtains new plane equation, and such iteration carries out, until obtained plane equation stabilizes to
Only, the last Typical Planar feature from the obtained plane of area maximum of fitting as noncooperative target.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 2)
In, the detection and extraction of three-dimensional straight line segment have been merged two dimensional image edge detection and have been detected with three-dimensional edges;First by three-dimensional point cloud
It all projects in original image, obtains effective image-region, and edge detection is carried out using canny operators, using being fitted
To the edge of a part, according to the relationship of image and three-dimensional point cloud, in these edge transitions to three dimensions, will obtain preliminary
Three-dimensional edges testing result;Secondly, detect and extract using three-dimensional planar as a result, ask friendship two-by-two using plane, if its
The result of friendship is asked just to against the original point cloud of image, then obtains some three-dimensional edges directly detected by three dimensions;
This two classes edge is put together, and close to each other is merged, and according to amalgamation result, the edge being connected to each other
It connects together;By the analysis of residual error, whether each section of edge of verification is straightway, if it is, obtaining corresponding three-dimensional
Straightway;Respective center of gravity is calculated to corresponding three-dimensional straight line segment, center of gravity is a point on each three-dimensional straight line segment, is then counted
Calculate the covariance matrix of these focus points, carry out Eigenvalues Decomposition to covariance matrix, feature corresponding to maximum eigenvalue to
Amount is the direction of this straightway.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 2)
In, three-dimensional loop truss and extraction carry out in a manner that three-dimensional is converted into two dimension;Specially:It is carried out in the image space of two dimension
Then this testing result is directly delivered in three-dimensional edges by the ellipses detection of the three-dimensional edges detected again, so as to complete
The inspection of three-dimensional circle.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 2)
In, three-D profile detection and extraction are based primarily upon a cloud, i.e. Triangulation Network Model, first to profile side into line trace, tracking according to
According to being, if a line of Triangulation Network Model is only abutted with a triangle, then this edge for profile while one it is candidate while,
All candidate contours sides for meeting conditions above are found out, the tracking on profile side is then carried out, a longest closed curve is taken
Go out, here it is most possible profile sides;Contour reconstruction is carried out from the profile side of detection, it is specific as follows:Calculate longest envelope
The tangent line rector that each in closed curve is put;Calculate the variable quantity of tangent line rector at each point;To these variable quantities, institute is found out
Some maximum, and point of these maximum more than certain threshold value is selected, as the candidate point of contour line inflection point, these are candidate
Closed curve is divided into many different sections by point;The curved section different to these, is fitted using straight line, analyzes its residual error,
If maximum residul difference is more than some threshold value, this section of straightway is split into both ends at maximum residul difference, such recurrence carries out
Until it can not divide;To division complete straightway merge detection according to the order of connection, if two be connected with each other
Straightway is fitted as one, and maximum residul difference is less than given threshold value, then merges this two straightways, this
Process wants recurrence to carry out, until can not find the straightway that can merge.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 3)
In, noncooperative target body coordinate system is established the stronger feature of selection repeatability and is carried out.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, select three-dimensional circle
Coordinate origin of the center as noncooperative target body coordinate system, the normal vector of the direction of longer straightway or larger plane
Reference axis as noncooperative target body coordinate system.
The above-mentioned noncooperative target relative movement information measuring method based on active coding information, wherein, the step 3)
In, noncooperative target body coordinate system foundation is carried out using following strategy:
(a)In the starting stage, the direction of the longer straightway of use or the normal vector of larger plane carry out noncooperative target
The estimation of reference axis, and using be closer to noncooperative target center a point as noncooperative target body coordinate system original
Point;
(b)According to step(a)Method constantly establish new noncooperative target body coordinate system, the foundation of each coordinate system
After the completion, under the Feature Conversion to noncooperative target body coordinate system not utilized others under current coordinate system, and
It is preserved;
(c)When certain is primary not detected for establishing the feature of noncooperative target body coordinate system, then select
It is supplemented with the feature saved.
The noncooperative target relative measurement system and measuring method based on active coding information of the present invention is compiled by active
The mode of code structure light, can efficiently solving image existing for passive measurement mode, easily masking is matched with caused by ambiguity
The problem of difficulty is big, while can overcome that passive measurement mode is easily illuminated by the light and the factors such as veiling glare environment, target configuration influence
The shortcomings that.Can meet the needs of closely noncooperative target high-precision, high steady relative measurement.
Description of the drawings
The noncooperative target relative measurement system based on active coding information and measuring method of the present invention is by following reality
It applies example and attached drawing provides.
Fig. 1 is the structure diagram of the noncooperative target relative measurement system based on active coding information of the present invention.
Specific embodiment
Below with reference to Fig. 1 to the present invention based on the noncooperative target relative measurement system of active coding information and measurement
Method is described in further detail.
Referring to Fig. 1, the noncooperative target relative measurement system of the invention based on active coding information is arranged on tracking boat
On its device, which includes first camera 1, second camera 2, structured light projector 3, the first imaging unit 4, second
Imaging unit 5, the first control unit 6, the second control unit 7, third control unit 8 and data processing unit 9;
The structured light projector 3 is used for noncooperative target(That is target satellite)Projection has certain structured coding information
Light model;Since noncooperative target surface each point depth has differences, the light model projects to meeting on noncooperative target surface
It deforms upon, the light model image through noncooperative target surface modulation, the light model image is formed on noncooperative target surface
Comprising noncooperative target surface each point depth information and coding information, coding information and the knot that the light model image includes
The coding information that the light model that structure light projector 3 projects includes matches;
The first camera 1 and second camera 2 are separately positioned on the both sides of the structured light projector 3, are used to shoot
The light model imaged image on noncooperative target surface;
First imaging unit 4 is used for the light model that the first camera 1 is controlled to expose and acquires the shooting of first camera 1
Imaged image, second imaging unit 5 are used for the optical mode that the second camera 2 is controlled to expose and acquires the shooting of second camera 2
Type imaged image;
First control unit 6 is connect with first imaging unit 4, the optical mode that first imaging unit 4 acquires
Type imaged image is sent into the data processing unit 9 after first control unit 6 decoding;Second control unit 7 with
Second imaging unit 5 connects, and the light model imaged image that second imaging unit 5 acquires is through second control unit
The data processing unit 9 is sent into after 7 decodings;
The third control unit 8 is used for coded structured light, and the structured light projector 3 is controlled to work;
The data processing unit 9 completes high-precision noncooperative target relative position and relative attitude by image procossing
It resolves.
In present pre-ferred embodiments, low power consumption projecting apparatus can be used in the structured light projector 3.
In present pre-ferred embodiments, the light model that the structured light projector 3 projects uses pseudorandom series coding staff
Formula is encoded, and using single frames structure light coding, make to measure every time only needs to carry out an Image Acquisition coding, does not need to be continuous
Multi collect, you can avoid the problem that moving object different moments generate pose variation and lead to measurement error, can also shorten every
The secondary acquisition required time.
In present pre-ferred embodiments, ccd detector can be used in first imaging unit 4(Charge coupling device,
Charge Coupled Device)Or cmos detector(Complementary metal oxide semiconductor, Complementary
Metal-Oxide-Semiconductor Transistor), second imaging unit 5 can be used ccd detector or
Cmos detector, preferably, first imaging unit, 4 and second imaging unit 5 is same probe.
In the present invention, the data processing unit 9 obtains noncooperative target relative position and relative attitude is based on including tool
There are the light model of certain structured coding information and the image of noncooperative target surface each point depth information, it is passive with the prior art
Measurement method(Noncooperative target image only is shot with two cameras, without structured light projector)It compares, image procossing and resolving
Difficulty substantially reduces, and precision is high, but the present invention based on the noncooperative target relative measurement system of active coding information to institute
It states data processing unit and obtains noncooperative target relative position and the image processing method involved in relative attitude process and resolving
Method is not construed as limiting.
Another technical solution of the present invention is to provide a kind of preferred acquisition noncooperative target for measuring system of the present invention
The method of relative position and relative attitude, but this is not that measuring system of the present invention obtains noncooperative target relative position and opposite appearance
The unique method of state.
In the preferred embodiment, the light model shadow that the data processing unit 9 is shot based on decoded first camera 1
As the light model imaged image that image and second camera 2 are shot carries out high-precision three-dimensional point cloud matching with rebuilding, and to reconstruction
Noncooperative target 3-D view carries out feature detection and extraction, then establishes noncooperative target body coordinate system, according to what is established
The change in coordinate axis direction of noncooperative target body coordinate system decomposites three attitude angles of noncooperative target, just obtains noncooperative target
Position and posture with respect to pursuit spacecraft.
The high-precision three-dimensional point cloud matching uses improved iteration closest approach method with rebuilding(Iterative Closest
Point, ICP).
Improved iteration closest approach method successively according to the selection with eka-element, the determining of registration strategies, search strategy is built
Vertical and error function 4 module expansion of solution, the specific steps are:
A cloud is sampled using the method for average sample, completes the selection with eka-element;
Determining for registration strategies is completed by the selection of characteristic measure and the selection of search strategy;It specifically includes and utilizes scene
The normal of middle centrostigma determines corresponding points with model points intersection of sets point, using the distance of point to face (point-to-plane)
Determine the distance of the tangent plane by corresponding points in the point to model data set that object function contextual data is concentrated;
Using multidimensional binary search tree(KD-Tree)The non-fully corresponding point set of dynamic fusion method processing, by point
Cloud, which establishes KD trees, can improve the efficiency of search closest approach, promote the speed of ICP iterative algorithms, complete the foundation of search strategy;It is main
Want step as follows:
(1)The KD trees of first frame point cloud are established using general KD trees method for building up;
(2)After subsequent point cloud completes registration, judge whether some clouds are correctly completed registration according to the registration error of the cloud, such as
Fruit is by then being merged;
(3)Degree of overlapping analysis is carried out to qualified cloud and first model point cloud, filters out nonoverlapping conduct
New point;
(4)New point is brought into KD trees using improved KD tree methods, improved method here essentially consists in increase newly
KD trees branch, existing KD leaves child node is updated;
Using the solution that error function is determined based on singular value decomposition method, noncooperative target high-precision three-dimensional point cloud is completed
With with reconstruction.
The feature includes three-dimensional planar, three-dimensional straight line segment, three-dimensional circle and three-D profile, now introduces the inspection of each feature one by one
Survey and extracting method.
The detection and extraction of three-dimensional planar include rough detection and extraction and essence detection and extraction.
The method that rough detection and extraction are blended using clustering methodology, curved surface growth method and plane act of union, specially:
Using K mean cluster method, cluster analysis is carried out as feature using the three-dimensional coordinate of cloud and normal vector, by three
Dimension point cloud is divided into possible multiple plane classes;
Each plane class is fitted, and is verified whether by regression criterion analysis as effective plane class, it will be effective
Plane class pick out;
It is identical using approximately the same plane normal vector based on effective plane class, using the method that region increases to this
A little planes are extended, if a flat blocks can effectively rise to another plane it is fast in if carry out the conjunctions of flat blocks
And;
These planes are fitted again, then using the parameter of each plane carry out it is further merge detection and
Merge, until planes all in certain threshold range are merged into same plane, can so obtain preferable plane monitoring-network effect
Fruit.
Essence detection and extraction carry out on the basis of rough detection and extraction, specially:Utilize the three-dimensional in the plane of acquisition
Point cloud carries out plane fitting to the point of some parts, and the plane equation obtained using fitting is found to be obtained positioned at the local fit
Plane in point(It is less than certain threshold value to fit Plane distance), it is fitted again and obtains new plane equation, such iteration
It carries out, until obtained plane equation is stablized(Front and rear fit Plane center of gravity is smaller than certain threshold value), finally from fitting
Typical Planar feature of the obtained plane of area maximum as noncooperative target.Typical Planar feature can be used for non-cooperative target
Mark the estimation of two attitude angles rotated around X and Y-axis, overcome based on noncooperative target contour edge it is irregular caused by angle
The problem of evaluated error is big.Essence detection described in the present embodiment and extraction are using three dimensions plane fitting, and basic principle is such as
Under:
Plane general equation is as follows:
(1)
Wherein,Normal vector for plane.
It is known to be located in planeA three-dimensional point(), using Eigenvalues Decomposition
Method is fitted plane.It willA known three-dimensional point carries out center of gravity, then:
(2)
By the three-dimensional point coordinate in plane(It is assumed that the random error with normal distribution), plane equation is substituted into, is obtained
To following error equation:
(,)(3)
It is write above error equation as matrix form, is obtained:
()(4)
Wherein,,。
Using least square principle, need to findValue so that following object function reaches minimum
Value:
(5)
In order to solve to obtainMinimum value, following Lagrangian is constructed using lagrange's method of multipliers:
(6)
First-order partial derivative is solved, and it is zero to enable it, obtains its extreme value necessary condition to Lagrangian:
(7)
It willAbove formula is substituted into, and abbreviation obtains:
(8)
It follows thatIt must be matrixCharacteristic value, andIt must be then corresponding feature vector.At this point, target letter
Several taken values are:
(9)
As it can be seen that the minimum value of object function must be matrixMinimal eigenvalue(BecauseFor positive definite or
Person's positive semidefinite matrix), being fitted obtained plane normal vector accordingly must be thenCorresponding feature vector.
According to above derivation, it is known that the step of carrying out plane fitting by Eigenvalues Decomposition mode is:
(1)Covariance matrix is calculated according to known three-dimensional point coordinate;
(2)Eigenvalues Decomposition is carried out to covariance matrix, obtains minimum characteristic valueAnd corresponding feature vector;
(3)Corresponding is exactly the normal vector of plane equation;
(4)Given that it is known that the barycentric coodinates of point are, the coefficient of Calculation Plane equation;
(5)To each known point, its residual error is calculated(Point arrives the distance of plane)In and
Error。
Three-dimensional straight Line segment detection and extraction have been merged two dimensional image edge detection and have been detected with three-dimensional edges, first by three-dimensional point
Cloud all projects to original image(The image of imaging system acquisition)In, obtain effective image-region, and using canny operators into
Row edge detection, using the edge that a part is obtained the methods of fitting, according to the relationship of image and three-dimensional point cloud, by these edges
It is transformed into three dimensions, has obtained preliminary three-dimensional edges testing result;Secondly, it detects and extracts using above-mentioned three-dimensional planar
As a result, ask friendship two-by-two using plane, if it asks the result of friendship just to against the original point cloud of image, can obtain by three
Some three-dimensional edges that dimension space directly detects;This two classes edge is put together, edge close to each other is merged,
And according to amalgamation result, the edge being connected to each other is connected together;It is every after verification connection by the analysis of residual error
Whether one section of edge is straightway, if it is, obtaining corresponding three-dimensional straight line segment;Corresponding three-dimensional straight line segment is calculated respective
Center of gravity, center of gravity are a points on each three-dimensional straight line segment, the covariance matrix of these focus points are then calculated, to covariance square
Battle array carries out Eigenvalues Decomposition, and the feature vector corresponding to maximum eigenvalue is the direction of this straightway.
The operation basic with extraction of three-dimensional straight Line segment detection is the fitting of straightway, and principle is as follows:
(a)It is assumed that certain one edge is straightway, and equation is
(10)
(b)If by marginal point whole centralization, the equation of straightway can be expressed as
(11)
(c)To obtained each point of one edge of detection, equation more than substitution can obtain a series of error
Equation, it is as follows
(12)
(d)Write a series of above error equations as matrix form, and considerThis linear equation constrains
Condition can obtain
(13)
(14)
Wherein。
(e)It is derived according to error equation form derived above, and with reference to least square method principle, X can be obtained
Optimal solution be matrixFeature vector corresponding to minimum singular value.So it only needs using singular value decomposition
Obtain the equation of straightway fitting.
Three-dimensional loop truss and extraction carry out in a manner that three-dimensional is converted into two dimension.Three-dimensional edges have been carried out in front
Detect and track, these three-dimensional edges can easily project to imaging using the projection relation between three-dimensional point and image
In unit the image collected.In order to rapidly detect required three-dimensional circle feature, the present embodiment will be in two dimension
Image space in the ellipses detection of three-dimensional edges that is detected, this testing result is then directly delivered to three-dimensional again
In edge, so as to complete the inspection of three-dimensional circle.
It is pointed out that it is transformed into image plane(The image space of two dimension)Interior three-dimensional edges are provided with the feature of two dimension, institute
The detection of three-dimensional circle can be carried out in a manner of using two-dimensional elliptic fitting.The principle of two-dimensional elliptic fitting is as follows:
(a)To each edge section, ellipse fitting is carried out.The elliptic equation that the present embodiment uses for:
(15)
For a point on edge, without loss of generality, it is assumed that a perseverances are 1, then the side that can be listed below
Journey:
(16)
For each point on edge, list more than equation, and uniformly write as the form of matrix, it is as follows:
(17)
According to least square method principle, the least square solution of above-mentioned equation can be obtained:
(18)
Above-mentioned least square solution is the coefficient of elliptic equation.
(b)Although Yi Shang least square obtains least squares sense elliptic equation coefficient solution, in order to more preferable
Ground considers the influence of a small amount of rough error marginal point, and the form of a weighting is employed herein, is further optimized.Namely exist
On the basis of above solution, it is weighted processing.For error equation, according to warp
It tests, the weight that may be used is
(19)
Wherein
(20)
(21)
It is weighed surely according to above mode, to each marginal point, error equation, and given phase is listed all in accordance with upper type
The weight answered can then obtain the error equation of following matrix form
(22)
Its least square solution is then accordingly
(23)
(c)According to the weights of above mode assigned error equation, resolve, then using calculation result, weigh surely again, so
It lists error equation again afterwards, and resolves, until iterating to result stabilization.It can be obtained by the oval algebraic equation of fitting in this way
.
It after obtaining elliptical algebraic equation, needs algebraic equation being converted into elliptical geometric parameter, so as to below
Processing and application.Elliptic equationThe practical quadratic form for standard of first three items, in order to
Oval incorporating parametric can be decomposited, needs to pass through orthogonal transformation(Namely spin matrix)Normal equation is converted into, is given thus
Go out matrix
(24)
Eigenvalues Decomposition is carried out to above matrix, as a result
(25)
Wherein U is spin matrix, it is assumed that
(26)
So after arranging, above-mentioned oval algebraic equation can turn to
(27)
It is if naturally, aboveWithIf jack per line, it may be elliptic equation to show above equation, otherwise will not be
Elliptic equation.WithUnder the premise of jack per line, by formula, standard ellipse equation can be converted into(It if cannot be by matching
Side is converted to elliptic equation, then it is not oval segmental arc to show the edge).According to obtained oval normal equation, can be easy to
Find out elliptical center, major semiaxis and section semiaxis, and can be then readily available according to the spin matrix U of front elliptical
Direction.
The detection and extraction of three-D profile are based primarily upon a cloud(Triangulation Network Model).First to profile side into line trace, with
The foundation of track is, if a line of Triangulation Network Model is only abutted with a triangle, then this edge is one of profile side
Candidate side.All candidate contours sides for meeting conditions above are found out first, the tracking on profile side are then carried out, by longest one
Closed curve takes out, and here it is most possible profile sides.In order to carry out further calculating below, need from detection
Contour reconstruction is carried out in profile side.The contour reconstruction method that the present embodiment uses is as follows:
(a)Calculate the tangent line rector that each in longest closed curve is put.The computational methods of tangent line are, to every on curve
One point finds out all the points in a certain size neighborhood on closed curve around the point, and calculates the covariance square of these points
Battle array carries out Eigenvalues Decomposition to the covariance matrix, and the feature vector corresponding to maximum eigenvalue is exactly the tangent vector of the point
Estimated value.
(b)Calculate the variable quantity of tangent line rector at each point.Method is each point found out on the vertex neighborhood inner curve
Tangent line rector, and calculate the angle between these tangent line rectors two-by-two, the maximum value side of these angles is considered to cut at this
The variable quantity of line vector.
(c)To these variable quantities, all maximum is found out, and point of these maximum more than certain threshold value is selected,
Candidate point as contour line inflection point.Closed curve is divided into many different sections by these candidate points.
(d)The curved section different to these, is fitted using straight line, analyzes its residual error, if the big Mr. Yu of maximum residul difference
This section of straightway is then split into both ends by one threshold value at maximum residul difference, until such recurrence proceeds to and can not divide.
(e)The straightway completed to division merges detection according to the order of connection, if the straight line of two interconnections
Section, is fitted as one, and maximum residul difference is less than given threshold value, then merges this two straightways.This process
Recurrence is wanted to carry out, until can not find the straightway that can merge.
Step more than can all detected the point of profile, and profile is reduced to a polygon expression.
Noncooperative target body coordinate system is established the stronger feature of selection repeatability and is carried out:Noncooperative target body coordinate system
Coordinate origin can select typical characteristic point, the center of such as three-dimensional circle, and noncooperative target body coordinate system reference axis is really
It is fixed, then preferably with some specific direction characters, such as the normal vector of the direction of longer straightway or larger plane.Due to
The calculating of noncooperative target relative position and posture must be based on fixed noncooperative target body coordinate system, so building in principle
The feature of vertical noncooperative target body coordinate system must extremely begin to the appearance that can be repeated eventually.However, as noncooperative target without
The known shape information of priori determines which feature will appear there is no method before noncooperative target measurement, also can not
Know which feature needs to detect, and data processing is also and not offline and post-processing, but one acquired according to data it is suitable
Sequence gradually incremental quasi real time processing procedure, so during the treatment and can not determine which feature can weigh without missing
It appears again existing.Therefore must be ensured to can not be detected for establishing the feature of noncooperative target body coordinate system with certain strategy
When remain able to steadily carry out the foundation of noncooperative target body coordinate system, and error accumulation effect is small.
The foundation of noncooperative target body coordinate system and position and attitude algorithm are carried out using following strategy:
(a)In the starting stage, the direction of the longer straightway of use or the normal vector of larger plane carry out noncooperative target
The estimation of reference axis, and using be typically closer to noncooperative target center a point as noncooperative target ontology coordinate
The origin of system, the change in coordinate axis direction according to the noncooperative target body coordinate system decomposite three attitude angles of noncooperative target,
The relative position and posture of noncooperative target under the noncooperative target body coordinate system can be obtained.
(b)According to step(a)Method constantly establish new noncooperative target body coordinate system(That is different moments, difference
The noncooperative target body coordinate system of motion state), will be other under current coordinate system after the completion of the foundation of each coordinate system
It under the Feature Conversion not utilized to noncooperative target body coordinate system, and is preserved, in this process, due to preserving
The number of repetition of feature gradually increases, and the precision of these features saved also can be gradually improved.
(c)When the feature that certain is once used to establish noncooperative target body coordinate system(The direction of i.e. longer straightway or
The normal vector of larger plane and the center of noncooperative target)When not detected, then select the feature that saves into
Row supplement.The complementary features of selection currently also must be capable of detecting when to come.This front and rear relationship of the same name can be utilized in this way
The current noncooperative target body coordinate system of direct estimation, and obtain current noncooperative target relative position and posture.
On the one hand above process carries out the individual noncooperative target body coordinate system of single frames using repeated characteristic and estimates,
And the relative position and posture of noncooperative target are calculated based on this, noncooperative target relative position and appearance has been effectively ensured in this
The independence of state parameter Estimation eliminates accumulated error effect.And the spare feature that noncooperative target preserves, in weight each time
In multiple, precision can step up, and these features are only in the direct feature estimated for noncooperative target body coordinate system
It can just be used in a complementary manner under conditions of detection failure, so above strategy, which meets, is ensureing that noncooperative target is opposite
Position and posture steadily and surely and under the premise of being carried out continuously estimation, can reduce the principle of error accumulation effect as far as possible.
Claims (13)
1. the noncooperative target relative measurement system based on active coding information, including sequentially connected first camera, the first one-tenth
As unit and the first control unit, at sequentially connected second camera, the second imaging unit and the second control unit and data
Manage unit;It is characterized in that, the noncooperative target relative measurement system further includes structured light projector and third control unit,
The third control unit control structure light projector has the light model of certain structured coding information to noncooperative target projection;
The light model forms light model image on noncooperative target surface;The first camera and second camera are separately positioned on described
The both sides of structured light projector are used to pickup light phantom images image;First control unit is by decoded described
The light model imaged image of one camera shooting is sent to the data processing unit, and second control unit is by decoded institute
The light model imaged image for stating second camera shooting is sent to the data processing unit, is realized and schemed by the data processing unit
Image height precision three-dimensional point cloud, which is matched, to be detected and extracts with reconstruction and feature, and then obtain noncooperative target relative position and opposite appearance
State.
2. the noncooperative target relative measurement system based on active coding information as described in claim 1, which is characterized in that institute
Light model is stated to be encoded using pseudorandom series coding mode.
3. the noncooperative target relative measurement system based on active coding information as described in claim 1, which is characterized in that institute
Light model coding is stated using single frames structure light coding.
4. the noncooperative target relative measurement system based on active coding information as described in claim 1, which is characterized in that institute
It states light model image and includes noncooperative target surface each point depth information and coding information.
5. the noncooperative target relative movement information measuring method based on active coding information, which is characterized in that using such as right
It is required that the noncooperative target relative measurement system based on active coding information in 1 to 4 described in any claim, the measurement side
Method includes the following steps:
1)Light model imaged image and the second camera shooting that the data processing unit is shot based on decoded first camera
Light model imaged image carry out high-precision three-dimensional point cloud matching with rebuild;
2)Feature detection and extraction are carried out to the noncooperative target 3-D view of reconstruction;
The feature includes three-dimensional planar, three-dimensional straight line segment, three-dimensional circle and three-D profile;
3)Noncooperative target body coordinate system is established, the change in coordinate axis direction according to the noncooperative target body coordinate system established decomposes
Go out three attitude angles of noncooperative target, just obtain position of the noncooperative target with respect to pursuit spacecraft and posture.
6. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 1)It is middle using improved iteration closest approach method realize noncooperative target high-precision three-dimensional point cloud matching with again
It builds;Improved iteration closest approach method is successively according to the selection with eka-element, the determining of registration strategies, the foundation of search strategy and mistake
4 module expansion of solution of difference function, the specific steps are:
A cloud is sampled using the method for average sample first, completes the selection with eka-element;
Then determining for registration strategies is completed by the selection of characteristic measure and the selection of search strategy;
Secondly non-fully corresponding point set is handled using multidimensional binary search tree dynamic fusion method, by establishing KD trees to a cloud
The efficiency of search closest approach can be improved, the speed of ICP iterative algorithms is promoted, completes the foundation of search strategy;
Finally using the solution that error function is determined based on singular value decomposition method, noncooperative target high-precision three-dimensional point cloud is completed
With with reconstruction.
7. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 2)In, the detection and extraction of three-dimensional planar include rough detection and extraction and essence detection and extraction;Institute
It states rough detection and extracts the method blended using clustering methodology, curved surface growth method and plane act of union;It is described essence detection and
Extraction and application step 1)The three-dimensional point cloud of acquisition carries out plane fitting to the point of some parts, and the equation obtained using fitting is looked for
To the point in the plane that local fit obtains, fitting again obtains new plane equation, and such iteration carries out, until obtaining
Plane equation stablize until, the last Typical Planar spy from the obtained plane of area maximum of fitting as noncooperative target
Sign.
8. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 2)In, two dimensional image edge detection and three-dimensional edges inspection have been merged in the detection and extraction of three-dimensional straight line segment
It surveys;Three-dimensional point cloud is all projected in original image first, obtains effective image-region, and side is carried out using canny operators
Edge detects, and the edge of a part is obtained using fitting, according to the relationship of image and three-dimensional point cloud, by these edge transitions to three-dimensional
In space, preliminary three-dimensional edges testing result has been obtained;Secondly, that is detected and extracted using three-dimensional planar puts down as a result, utilizing
Friendship is asked in face two-by-two, if it asks the result of friendship just to against the original point cloud of image, obtains being directly detected by three dimensions
Some three-dimensional edges;This two classes edge is put together, close to each other is merged, and according to amalgamation result, phase
The edge being connected to together is interconnected to connect together;By the analysis of residual error, whether each section of edge of verification is straightway, if
It is then to obtain corresponding three-dimensional straight line segment;Respective center of gravity is calculated to corresponding three-dimensional straight line segment, center of gravity is each 3 d-line
A point in section then calculates the covariance matrix of these focus points, and Eigenvalues Decomposition is carried out to covariance matrix, maximum special
Feature vector corresponding to value indicative is the direction of this straightway.
9. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 2)In, three-dimensional loop truss and extraction carry out in a manner that three-dimensional is converted into two dimension;Specially:Two
Then this testing result is directly delivered to three by the ellipses detection for the three-dimensional edges being detected in the image space of dimension again
It ties up in edge, so as to complete the inspection of three-dimensional circle.
10. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 2)In, three-D profile detection and extraction are based primarily upon a cloud, i.e. Triangulation Network Model, first to profile side
Into line trace, the foundation of tracking is, if a line of Triangulation Network Model is only abutted with a triangle, then this edge is wheel
It is wide while one it is candidate while, find out all candidate contours sides for meeting conditions above, the tracking on profile side then carried out, by longest
A closed curve take out, here it is most possible profile sides;Contour reconstruction is carried out from the profile side of detection, specifically
It is as follows:Calculate the tangent line rector that each in longest closed curve is put;Calculate the variable quantity of tangent line rector at each point;To this
A little variable quantities find out all maximum, and point of these maximum more than certain threshold value are selected, as contour line inflection point
Closed curve is divided into many different sections by candidate point, these candidate points;The curved section different to these, is intended using straight line
It closes, analyzes its residual error, if maximum residul difference is more than some threshold value, this section of straightway is split into two at maximum residul difference
End, until such recurrence proceeds to and can not divide;The straightway completed to division merges detection according to the order of connection, if
The straightway of two interconnections, is fitted as one, and maximum residul difference is less than given threshold value, then by this two straightways
It merges, this process wants recurrence to carry out, until can not find the straightway that can merge.
11. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 3)In, noncooperative target body coordinate system is established the stronger feature of selection repeatability and is carried out.
12. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 11,
It is characterized in that, selects coordinate origin of the center of three-dimensional circle as noncooperative target body coordinate system, the direction of longer straightway
Or reference axis of the normal vector of larger plane as noncooperative target body coordinate system.
13. the noncooperative target relative movement information measuring method based on active coding information as claimed in claim 5, special
Sign is, the step 3)In, noncooperative target body coordinate system foundation is carried out using following strategy:
(a)In the starting stage, the direction of the longer straightway of use or the normal vector of larger plane carry out noncooperative target coordinate
The estimation of axis, and using be closer to noncooperative target center a point as noncooperative target body coordinate system origin;
(b)According to step(a)Method constantly establish new noncooperative target body coordinate system, the foundation of each coordinate system is completed
Afterwards, it under the Feature Conversion to noncooperative target body coordinate system not utilized others under current coordinate system, and carries out
It preserves;
(c)When certain is primary not detected for establishing the feature of noncooperative target body coordinate system, then guarantor is selected
The feature stored away is supplemented.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410844755.XA CN105806315B (en) | 2014-12-31 | 2014-12-31 | Noncooperative target relative measurement system and measuring method based on active coding information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410844755.XA CN105806315B (en) | 2014-12-31 | 2014-12-31 | Noncooperative target relative measurement system and measuring method based on active coding information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105806315A CN105806315A (en) | 2016-07-27 |
CN105806315B true CN105806315B (en) | 2018-06-26 |
Family
ID=56421044
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410844755.XA Active CN105806315B (en) | 2014-12-31 | 2014-12-31 | Noncooperative target relative measurement system and measuring method based on active coding information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105806315B (en) |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106251376B (en) * | 2016-08-12 | 2019-08-06 | 南京航空航天大学 | One kind is towards colored structures pumped FIR laser and edge extracting method |
CN106327486B (en) * | 2016-08-16 | 2018-12-28 | 广州视源电子科技股份有限公司 | Track the method and device thereof of the finger web position |
CN106340040B (en) * | 2016-08-16 | 2019-06-14 | 广州视源电子科技股份有限公司 | Tracking refers to the method and device thereof of contouring |
CN106340039B (en) * | 2016-08-16 | 2019-06-21 | 广州视源电子科技股份有限公司 | Tracking refers to the method and device thereof of contouring |
CN108228931B (en) * | 2016-12-15 | 2022-09-06 | 北京金风科创风电设备有限公司 | Method and device for evaluating terrain for wind turbine sample |
CN106643701B (en) * | 2017-01-16 | 2019-05-14 | 深圳优地科技有限公司 | A kind of mutual detection method and device of robot |
CN107990846B (en) * | 2017-11-03 | 2020-01-31 | 西安电子科技大学 | Active and passive combination depth information acquisition method based on single-frame structured light |
CN108122254B (en) * | 2017-12-15 | 2021-06-22 | 中国科学院深圳先进技术研究院 | Three-dimensional image reconstruction method and device based on structured light and storage medium |
CN108122280A (en) * | 2017-12-20 | 2018-06-05 | 北京搜狐新媒体信息技术有限公司 | The method for reconstructing and device of a kind of three-dimensional point cloud |
CN108562274B (en) * | 2018-04-20 | 2020-10-27 | 南京邮电大学 | Marker-based non-cooperative target pose measurement method |
CN108876862B (en) * | 2018-07-13 | 2021-12-07 | 北京控制工程研究所 | Non-cooperative target point cloud position posture calculation method |
CN109035338B (en) * | 2018-07-16 | 2020-11-10 | 深圳辰视智能科技有限公司 | Point cloud and picture fusion method, device and equipment based on single-scale features |
CN109785349A (en) * | 2019-01-07 | 2019-05-21 | 哈尔滨理工大学 | A kind of polar coordinates edge coding method based on circle model |
CN111739111B (en) * | 2019-03-20 | 2023-05-30 | 上海交通大学 | Point cloud projection coding intra-block offset optimization method and system |
CN110308459B (en) * | 2019-06-30 | 2023-05-09 | 南京理工大学 | Model-independent non-cooperative satellite relative pose measurement method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101419070A (en) * | 2008-12-03 | 2009-04-29 | 南京航空航天大学 | Relative position and pose determining method based on laser ranging formatter |
EP2416113A2 (en) * | 2010-08-06 | 2012-02-08 | Canon Kabushiki Kaisha | Position and orientation measurement apparatus and position and orientation measurement method |
CN102589526A (en) * | 2011-12-23 | 2012-07-18 | 北京控制工程研究所 | Single baseline binocular measuring system for measuring non-cooperative targets |
CN102589530A (en) * | 2012-02-24 | 2012-07-18 | 合肥工业大学 | Method for measuring position and gesture of non-cooperative target based on fusion of two dimension camera and three dimension camera |
CN102788559A (en) * | 2012-07-19 | 2012-11-21 | 北京航空航天大学 | Optical vision measuring system with wide-field structure and measuring method thereof |
-
2014
- 2014-12-31 CN CN201410844755.XA patent/CN105806315B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101419070A (en) * | 2008-12-03 | 2009-04-29 | 南京航空航天大学 | Relative position and pose determining method based on laser ranging formatter |
EP2416113A2 (en) * | 2010-08-06 | 2012-02-08 | Canon Kabushiki Kaisha | Position and orientation measurement apparatus and position and orientation measurement method |
CN102589526A (en) * | 2011-12-23 | 2012-07-18 | 北京控制工程研究所 | Single baseline binocular measuring system for measuring non-cooperative targets |
CN102589530A (en) * | 2012-02-24 | 2012-07-18 | 合肥工业大学 | Method for measuring position and gesture of non-cooperative target based on fusion of two dimension camera and three dimension camera |
CN102788559A (en) * | 2012-07-19 | 2012-11-21 | 北京航空航天大学 | Optical vision measuring system with wide-field structure and measuring method thereof |
Non-Patent Citations (2)
Title |
---|
单站发射多站接收的空间目标激光测距技术;吴志波;《飞行器测控学报》;20120414;第33卷(第2期);第134-138页 * |
无合作目标测量中目标表面后向散射特性的实验研究;陶会荣;《红外与激光工程》;20141225;第43卷;第95-100页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105806315A (en) | 2016-07-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105806315B (en) | Noncooperative target relative measurement system and measuring method based on active coding information | |
Pizzoli et al. | REMODE: Probabilistic, monocular dense reconstruction in real time | |
CN107063228B (en) | Target attitude calculation method based on binocular vision | |
CN106023298B (en) | Point cloud Rigid Registration method based on local Poisson curve reestablishing | |
CN108615244B (en) | A kind of image depth estimation method and system based on CNN and depth filter | |
CN109558879A (en) | A kind of vision SLAM method and apparatus based on dotted line feature | |
US10288425B2 (en) | Generation of map data | |
WO2015096508A1 (en) | Attitude estimation method and system for on-orbit three-dimensional space object under model constraint | |
WO2019011958A1 (en) | System and method for pose-invariant face alignment | |
CN106485690A (en) | Cloud data based on a feature and the autoregistration fusion method of optical image | |
CN102697508A (en) | Method for performing gait recognition by adopting three-dimensional reconstruction of monocular vision | |
US10755433B2 (en) | Method and system for scanning an object using an RGB-D sensor | |
CN109425348A (en) | A kind of while positioning and the method and apparatus for building figure | |
CN104268880A (en) | Depth information obtaining method based on combination of features and region matching | |
CN113393524B (en) | Target pose estimation method combining deep learning and contour point cloud reconstruction | |
Shin et al. | A multi-camera calibration method using a 3-axis frame and wand | |
Rhodes et al. | A concise guide to feature histograms with applications to LIDAR-based spacecraft relative navigation | |
Kostavelis et al. | Visual odometry for autonomous robot navigation through efficient outlier rejection | |
CN114742894A (en) | Multi-camera calibration method in large scene, information processing terminal and storage medium | |
CN116468786B (en) | Semantic SLAM method based on point-line combination and oriented to dynamic environment | |
Phan et al. | Towards 3D human posture estimation using multiple kinects despite self-contacts | |
Kupervasser et al. | Robust positioning of drones for land use monitoring in strong terrain relief using vision-based navigation | |
Zhang et al. | Object depth measurement from monocular images based on feature segments | |
Xu et al. | A real-time ranging method based on parallel binocular vision | |
Wang et al. | A dynamic target visual positioning method based on ROI |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |