A kind of point cloud acquisition platform of Oriented Green plant temporal model
Technical field
The utility model belongs to plant three-dimensional shape phenotype field of measuring technique, specifically, being related to a kind of Oriented Green
The point cloud acquisition platform of plant temporal model.
Background technique
Plant phenotype refers to that a series of physical, physiology and the biochemistry of reflection plant structure, composition, growth course and result are special
Property and character;Research plant phenotype can instruct plant genotype to study, development, grasp plant development to Research of Plant Genomics
Rule has great importance.
By measuring the phenotypic parameter of plant, analysis parameters change with time rule, can construct the dynamic of plant
State growth model, the i.e. temporal model of plant, for intuitively showing that plant deduces the appearance of root, stem, leaf, flower and fruit at any time
The variation of form, topological structure, and can be due to the plant growing way under prediction varying environment condition and cultivation step, while can also tie
Close computer graphics techniques, the parameter source as digital virtual plant growth.
The phenotypic parameter of plant is various, wherein being the most intuitively exactly the geometry topological structure of plant growth, i.e. plant
Form phenotypic parameter.Due to the relative complex spatial form and structure of plant, the image analysis of simple two dimension level is difficult to obtain more
Accurate measured value, it is therefore desirable to obtain the ginseng of the three-dimensional configuration phenotype in growing process by some three-dimensional measurement modes
Number;Before obtaining three-dimensional configuration phenotypic parameter, the threedimensional model of plant is significantly reconstructed.
The threedimensional model of reconstruct plant is necessarily required in plant from sowing, take root, germinate, come into leaves, bloom, the full life of result
During long period, continuous, multiple, repeated acquisition plant form phenotypic parameter, each collection result is not due to environmental change
It is impacted, and collection result precision is high, reliability is good, rapid convenience.
The method of the threedimensional model of reconstruction of objects can totally be divided into two kinds, one kind be it is active, i.e., projected to object
Special light source or energy obtains three-dimensional configuration information by detecting the energy of transmission or reflection, such as utilizes 3-D scanning
Equipment is scanned real-world object using technologies such as structure light, encoded light, laser, directly obtains the information of object space point,
And then reconstruct three-dimensional configuration;Another kind is passive type, that is, does not need to project certain light source or energy to object, but detection pair
As certain energy of reflection environment, for example, using video camera, digital camera from two width of the real-world objects of each viewing angles or
Person's multiple image is reconstructed.
The method that plant phenotype detection field obtains plant threedimensional model at present is mainly the laser method in active, is passed through
To plant pair as emitting laser signal, calculates from the time for receiving reflection signal or phase change is transmitted signals to, estimate
The range ability of signal, to calculate the spatial positional information of object.The advantages of laser method, is that reconstruction accuracy is good, but right
The required precision of instrument is very high, thus cost is also very high.
In active to plant emitting structural light, encoded light, the light for shooting plant subject surface reflects distribution, according to
It is also often used to obtain the threedimensional model of plant according to the location information that geometry calculates surface.The advantages of this method, is to count
Calculation is simple, scanning speed is fast;Certain fluctuation is had, but whole according to the structure light of transmitting, the state of encoded light in precision aspect
Precision wants poor compared with laser method for body.
Other than the active acquisition plant model method of above-mentioned two kinds, monocular or binocular vision in passive type
Method be also commonly used for the three-dimensional model reconfiguration of plant.Its principle be pre-placed industrial camera, and to industrial camera into
Rower is fixed, calculates the inside and outside ginseng matrix of industrial camera;Monocular vision method uses a camera, and plant object is enabled successively to revolve
Turn certain angle shot two-dimension picture, then calculates on two-dimension picture pixel in space with respect to the position of camera;Binocular
Visible sensation method then respectively places a camera in plant two sides, and plant object is enabled successively to rotate a certain angle, and two cameras are clapped simultaneously
Two-dimension picture is taken the photograph, then calculates on two-dimension picture pixel in space with respect to the position of two cameras.
The above common plant model method that obtains can access accurate plant threedimensional model, but it is all
There is a general character, in order to obtain the three-dimensional information of multi-angle, all has chosen the motion mode that plant rotates and sensor is fixed.
And when reconstructed object is that stem is elongated and the plant of blade small, flexible, the motion mode of this plant rotation is easy to cause
The shake of the organs such as plant stem, blade, seriously affects reconstruction quality.It, on the one hand can be in order to avoid the shake of plant organ
Optimize the control system rotated with animals and plants, but can inevitably reduce the speed of rotation, is unfavorable for the high efficiency rebuild;Another party
Face also can choose and plant enabled to fix and the scheme of sensor movement.But due to Laser emission receiver precise structure itself and
Complexity, structure light, encoded light and monocular, binocular vision method require in advance to demarcate system so that it is determined that camera exists
Parameter in system, if fixing plant sensor move, therefore, to assure that a set of precision and fluency it is high driving sensing
The hardware system of device system motion necessitates condition, but still also to solve the problems, such as that other influences are rebuild.
To sum up, the device is complicated, required precision is high, at high cost for the method for active reconstruct plant threedimensional model, in passive type
Monocular or binocular vision for plant three-dimensional model reconfiguration equipment is simple, required precision is low, relative inexpensiveness, if but
It is low that complexity, reconstruction quality are demarcated in such a way that plant rotation and monocular or binocular camera are fixed.Therefore camera reconstruction is being used
When plant threedimensional model, fixed using plant and becoming by the way of camera motion ensures that reconstruction model precision is high, equipment cost is low
Approach.
It is a kind of construction threedimensional model in passive type from exercise recovery shape (Structure from Motion, SfM)
Algorithm, basic thought are to enable object and camera that relative motion occurs and shoot the two-dimensional image sequence of the multiple angles of object, are led to
It crosses and mathematical analysis is carried out to the target in the image sequence, calculate the three-dimensional motion parameter of object or camera, thus obtain object
The three-dimensional space point cloud of body.Common three-dimensional reconstruction method is to be demarcated in advance to system before acquisition, and camera is in system
In relative position determine in advance, but the use of the coordinate system of SfM algorithm threedimensional model generated is to estimate after acquisition
Come, i.e., the most image of Automatic-searching characteristic matching number to as initial pictures pair, start to calculate in outer ginseng and space three
Dimension point;Calculated result based on this initial pictures pair, the new image of addition that or else breaks are calculated.When random selected camera site
And shooting angle, and when shooting environmental (such as light source, intensity of illumination) is unstable, to same its image sequence of object multi collect
For three-dimensionalreconstruction, the most image of characteristic matching number often will cause in every wheel image sequence to inconsistent (i.e. initial every time
Position when image is to shooting is not necessarily), cause the coordinate system of the threedimensional model generated every time inconsistent, as shown in Figure 1,
Minute of four groups of different initial pictures to four groups of point clouds generated in the same coordinate system is chosen to same a collection of image sequence
Cloth situation, it can be found that the coordinate disunity of object plant four times threedimensional models, mutual position and scale all deviations compared with
Greatly, this will cause to need when to these threedimensional model extracting parameters additionally relevant for the coordinate system adjustment generated every time
Parameter, so that the repeatability of algorithm is unreliable.
And the temporal model research of plant needs the plant threedimensional model based on multiple periods, such as certain plant growing cycle
It is six weeks, to study its temporal model, needs its threedimensional model of daily timing acquiring;If using SfM algorithm, the three-dimensional mould of generation
Type coordinate system will be inconsistent to bring inconvenience for subsequent parameter extraction, such as: comprising in addition to planting in the three-dimensional point cloud that SfM algorithm generates
Other environment point clouds (because these environment point clouds are also taken into image sequence and are identified as feature) outside strain object-point cloud,
If the coordinate system generated every time is unified, it may be determined that each coordinate value range directly filters out plant object-point cloud;If generating every time
Coordinate system disunity is just difficult to determine each coordinate value range.Meanwhile coordinate system is inconsistent inconsistent comprising scale, if right
Same plant obtains two threedimensional model in a short time, and its scale is inconsistent, will lead to same in two threedimensional models
The length of one blade is inconsistent, leads to the large error of measurement result.In addition, being needed in some common point cloud segmentation algorithms
Segmentation threshold parameter (such as European cluster segmentation) is preset, causes scale inconsistent if coordinate system is inconsistent, such point
It cuts threshold parameter and is also difficult to standardization.
It should also be emphasized that although SfM algorithm is not strict with camera motion track, to protect as far as possible
The result for holding repeated acquisition is stablized, still should be to the movement of camera from the engineering angle of plant phenotype parameter acquisition
Track is set;Meanwhile the feature that in view of environment various factors will affect SfM algorithm identifies, such as different light source, illumination are strong
The quality of colour that degree will lead to plant surface reflected light is different, to be difficult to determine when using color threshold segmentation plant point cloud
Color threshold parameter, therefore also should be ensured that the stabilization of environmental factor in shooting process.So build one it is reliable, stablize and
It is particularly important for being suitble to the acquisition hardware platform of SfM algorithm.
Utility model content
1, it to solve the problems, such as
For existing SfM algorithm plant three-dimensional shape phenotype fields of measurement there are coordinate systems inconsistent, poor repeatability,
The problem of reliability low defect, the utility model provides a kind of point cloud acquisition platform of Oriented Green plant temporal model;This
Utility model adapts to the Image Acquisition mode that plant is fixed, sensor moves, and can effectively avoid leading since sensor moves
The disadvantages of sensor position parameter of cause lacks, precision can not ensure, image noise is too big.
2, technical solution
To solve the above problems, the utility model adopts the following technical scheme.
A kind of point cloud acquisition platform of Oriented Green plant temporal model, including light source, box frame, box baseplate and plant
Object carrying platform, the inside top of the box frame are equipped with light source, further include upright guide rail, horizontal guide rail, directional wheel, electronic
Turntable, rotating platform and colored target;Top in the box frame is equipped with LED light source, the box baseplate installation electricity
Worm gear is arranged in dynamic turntable, the electric rotary table Pivot axle and cabinet center overlapping of axles, the electric rotary table inside
Rotating platform is arranged in worm screw speed changer, the worm gear and endless screw speed changer input terminal setting motor, output end;
Described horizontal guide rail one end and rotating platform are fixed, directional wheel is arranged in the other end, are provided on the horizontal guide rail
Upright guide rail, the upright guide rail can be moved along the length direction of horizontal guide rail, and camera mounting bracket is arranged on the upright guide rail,
The camera mounting bracket can be moved along the length direction of upright guide rail;
The plant carrying platform passes through the central through hole of electric rotary table, is fixed in box baseplate, top planes
The center of the colored target of upper placement, the colour target is overlapped with the center of top planes, is placed and is planted on colored target
Object object, the basin bottom center of the plant pair elephant are located at the center of colored target.
Further, the box frame setting black Flocked fabric is as background in frame, to ensure Image Acquisition ring
The stability of border light source and from interference.
Further, the box frame is built using aluminum profile, and the top in box frame is equipped with LED light source, institute
It states and camera is set in camera mounting bracket.
Further, red, blue two kinds of targets are distributed on the colored target, using black as background color, wherein red
In color mark plate, red line segment mark plate one, red line segment mark plate two are symmetrically distributed, red square target one, red square target
Two are also symmetrically distributed, and share a symmetric points, i.e., the center of colored target;There is a rectangle indigo plant in one end of colored target
Color mark plate.
Further, red, blue two kinds of targets are distributed in the colored target, using black as background color, wherein red
Target is divided into red sub- target one, red sub- target two, is symmetric with the symmetry axis of colored target entirety;Red sub- target
One, red sub- target two is made of several red sub- squares, and number is at least 5, and meets central symmetry distribution simultaneously;?
Equally there is the blue target an of rectangle in one end of colored target.
3, beneficial effect
Compared with the prior art, the utility model has the following beneficial effects:
(1) construct Oriented Green plant temporal model point cloud acquisition platform, by artificially control light source, intensity of illumination,
Background, shooting distance, angle provide stable image capture environment, solve when because of acquisition caused by the variation of external environment
The problem that the three-dimensional point cloud model precision of acquisition is low, effect is unstable can satisfy the full growth week of plant temporal model research
Phase is continuous, repeatedly, repeated acquisition need so that collection result is not impacted because of environmental change every time;
(2) design plant fix, the platform structure that sensor is mobile, solve traditional sensor fix, plant it is mobile
In mode, stalk is elongated, blade small, flexible plant is easy to happen shake in shooting process to influence three-dimensional reconstruction effect
The problem of fruit, it is ensured that obtain static plant threedimensional model, it is accurate, reliable, easily extract plant phenotype parameter;
(3) the coordinate system standardized method for proposing the Oriented Green plant temporal model based on colored target, can be to back
Scape, green plants object, colored target three point cloud independently operate, from the angle of hardware, meet multi collect
The coordinate system of threedimensional model remains unified requirement, realizes with same set of tool and algorithm and does not need repeatedly to restructure
Or in the case where algorithm parameter, parameter extraction can be carried out to plant object in each unit time;
(4) one kind is designed for the standardized colored target of point cloud coordinate system, on the one hand solves traditional camera calibration
On the other hand the cumbersome and problem not flexible, algorithm is complicated of equal operations solves existing SfM algorithm and generates coordinate system at random and makes
At the threedimensional model coordinate system inconsistence problems generated every time, it can realize that coordinate system standardizes from the angle of hardware configuration, one
Denier colour target is placed in platform, can be achieved with camera motion multi collect, and no longer needs to manually adjust, and improves acquisition
Efficiency realizes the high efficiency of Image Acquisition;
(5) a kind of one end connection rotating platform is designed, it is camera that the other end, which carries mode by the camera of directional wheel support,
Around fixed plant powered rotation, and motion positions are realized in the case where not needing to be arranged additional Precise Orbit, simplify knot
Structure reduces cost.
Detailed description of the invention
Fig. 1 is to choose four groups of different initial pictures to four groups of point clouds generated in the same seat with a collection of image sequence
Distribution situation figure in mark system;
Fig. 2 is point cloud acquisition platform structure simplified schematic diagram;
Fig. 3 is the definition schematic diagram of normalized coordinates system;
Fig. 4 (a) is the structural schematic diagram of line segment type colour target of the invention;Fig. 4 (b) is that diamond type of the invention is colored
The structural schematic diagram of target;Fig. 4 (c) is the structural schematic diagram of comprehensive colored target of the invention;
Fig. 5 is that schematic diagram is arranged in space fixed point;
Fig. 6 is the small-sized side of blue that one line segment type colour target of scheme and two diamond type colour target schematic diagram of scheme identify
Block barycentric coodinates distribution map.
Fig. 7 is the small-sized square barycentric coodinates distribution of blue of the comprehensive colored target identification of scheme three;
Fig. 8 is colored target schematic diagram after optimization;
Fig. 9 is the point cloud coordinate system generation method algorithm flow chart of Oriented Green plant temporal model.
In figure: 1, LED light source;101, red line segment mark plate one;102, red line segment mark plate two;2, box frame;201,
Red square target one;202, red square target two;3, upright guide rail;4, horizontal guide rail;4101, red sub- target one;
4102, red sub- target two;5, directional wheel;6, the bottom of box;7, plant carrying platform;8, motor;9, electric rotary table;10,
Rotating platform;11, the central point of colored target;12, the small-sized square of blue.
Specific embodiment
It is practical new below in conjunction with this to keep the objectives, technical solutions, and advantages of the embodiments of the present invention clearer
Attached drawing in type embodiment, the technical scheme in the utility model embodiment is clearly and completely described.Wherein, it is retouched
The embodiment stated is the utility model a part of the embodiment, instead of all the embodiments.Therefore, below to providing in the accompanying drawings
The detailed description of the embodiments of the present invention be not intended to limit the range of claimed invention, but only
Indicate the selected embodiment of the utility model.
Embodiment 1
With reference to Fig. 2, a kind of point cloud acquisition platform of Oriented Green plant temporal model, including LED are present embodiments provided
Light source 1, box frame 2, box baseplate 6 and plant carrying platform 7, wherein black Flocked fabric is arranged as frame in box frame 2
Interior background, and its internal setting black Flocked fabric is as background in frame, with ensure image capture environment light source stability and
From interference, while the top in box frame 2 is equipped with LED light source 1, and box baseplate 6 installs electric rotary table 9, electronic rotation
Worm gear and endless screw speed changer, worm and gear speed change is arranged inside electric rotary table 9 in 9 Pivot axle of platform and cabinet center overlapping of axles
Device input terminal is provided with motor 8, output end is provided with rotating platform 10.
Wherein, rotating platform 10 and one end of horizontal guide rail 4 are fixed, and 4 other end of horizontal guide rail is provided with directional wheel 5, together
When horizontal guide rail 4 on be provided with upright guide rail 3, upright guide rail 3 can be moved along the length direction of horizontal guide rail 4, upright guide rail 3
On be additionally provided with camera mounting bracket, camera mounting bracket can be moved along the length direction of upright guide rail 3, wherein be set in camera mounting bracket
It is equipped with camera.
Plant carrying platform 7 passes through the central through hole of electric rotary table 9, is fixed in box baseplate 6, in top planes
Chromatic colour target is placed, the center of colored target is overlapped with the center of top planes, plant object is placed on colored target,
The basin bottom center of plant pair elephant is located at the center of colored target.
Plant carrying platform 7 passes through the central through hole of electric rotary table 9, is fixed in box baseplate 6, in top planes
Chromatic colour target is placed, the center of colored target is overlapped with the center of top planes, plant object is placed on colored target,
The basin bottom center of plant pair elephant is located at the center of colored target.
With reference to Fig. 3, red, blue two kinds of targets are distributed on colored target, using black as background color, wherein red target
In, red line segment mark plate 1, red line segment mark plate 2 102 are symmetrically distributed, red square target 1, red square mark
Plate 2 202 is also symmetrically distributed, and shares a symmetric points, that is, the central point 11 of colored target;In one end of colored target
There is a rectangle blue target.
A kind of point cloud acquisition method of Oriented Green plant temporal model, step are specific as follows:
(1) green plants object is placed in plant carrying platform 7, opens top light source 1, the phase in all box frames 2
8 power supply of machine and motor, confirmation green plants is located in the middle part of camera shooting visual angle, and can take plant bottom, be placed on
The colored target at 7 top of plant carrying platform, closes background cloth window shade.
(2) camera shoots samples pictures sequence, makes around electric rotary table 9 Pivot axle, 360 ° of movements, constant duration
This scene original point cloud P is generated with SfM algorithmInit。
(3) it presets color of object threshold range: according to RGB color model, need to determine tri- Color Channels of R, G, B
Threshold range;Guarantee that G, channel B take gamut, and the optimal threshold range in the channel R is determined using dichotomy, that is, is partitioned into first
Point cloud quality is best;Secondly true using dichotomy using the fixed channel R optimal threshold range and the channel B of gamut
Determine the optimal threshold range in the channel G;The fixed channel R, G optimal threshold range is finally used, determines channel B using dichotomy
Optimal threshold range.
In the hardware test environment of embodiment, the optimal threshold range of green plants object-point cloud is R (0,150), G
(50,255), B (0,150), the optimal threshold range of red target point cloud are R (60,255), and G (0,50), B (0,100) are blue
The optimal threshold range of color mark plate point cloud is R (0,150), G (0,150), B (100,255)), using color filter, from original
Point cloud PInitIn be partitioned into green plants object-point cloud PG, colored target point cloud P, and red is partitioned into from colored target point cloud
Target point cloud PR, blue target point cloud PB。
(4) the normalized coordinates system C of colored target point cloud P is obtained1, and combine the initial coordinate system of colored target point cloud P
C is calculated and is converted normalized coordinates system C for initial coordinate system C1Spin matrix, translation matrix, scaled matrix, will be green
Plant object-point cloud PGAll the points coordinate successively with above-mentioned matrix dot product, obtain normalized coordinates system C1Under green point cloud
PGnew.Wherein obtain the normalized coordinates system C of colored target point cloud P1, especially by the point cloud of Oriented Green plant temporal model
Coordinate system generation method is obtained, and with reference to Fig. 9, step is specific as follows:
(4.1) design of colored target: as shown in figure 3, red, blue two kinds of targets are distributed on colored target, with black
Color is background color, wherein red line segment mark plate 1, red line segment mark plate 2 102 are symmetrically distributed, red side in red target
Block target 1, red square target 2 202 are also symmetrically distributed, and share a symmetric points, that is, the center of colored target
Point 11, has a rectangle blue target 301 in one end of colored target.
The design of colored target directly affects the stability of normalized coordinates system generated, colored target of the invention
Then red line segment mark plate and red square target are combined together, as shown in attached drawing 4 (c).
(4.2) normalized coordinates system: the foundation for using colour target point cloud P to generate as coordinate system is defined;Colored target
Point cloud P initial coordinate system C has x-axis, y-axis, z-axis, is marked with normalized coordinates system C1With X-axis, Y-axis, Z axis, each axis mutually hangs down
Directly, intersection point is set as normalized coordinates system C1Origin;As shown in Fig. 3, by normalized coordinates system C1Unified definition is as follows:
(4.2.1) coordinate origin O: normalized coordinates system C1Origin be defined as the symmetric points of red line segment mark plate, i.e.,
The central point 11 of colored target.
(4.2.2) X-axis forward direction unit vector: by normalized coordinates system C1X-axis position be defined on red line segment mark plate one
101, on the line of the respective center of gravity of red line segment mark plate 2 102, direction is from center of gravity (the red line segment mark plate far from blue target
2 102 center of gravity) it is directed toward the direction of center of gravity (center of gravity of red line segment mark plate 1) close to blue target, calculate this vector
And it is unitization, as normalized coordinates system C1X-axis forward direction unit vector.
(4.2.3) Z axis forward direction unit vector: by normalized coordinates system C1Z axis forward direction unit vector be defined as vertical coloured silk
Color mark plate plane, and it is directed toward the unit normal vector of plant;
(4.2.4) Y-axis forward direction unit vector: being based on right-hand rule, takes and makees both perpendicular to the unit vector of X-axis and Z axis
For normalized coordinates system C1Y-axis unit vector.
(4.3) normalized coordinate system C1
(4.3.1) determines C1Origin: red target point cloud P is calculatedRBarycentric coodinates gC(xC,yC,zC) sat as standardization
Mark system C1Origin.
(4.3.2) determines X-axis forward direction unit vector: European clustering procedure is used, by red target point cloud PRIt is divided into two
Point, respectively PR1、PR2, wherein PR1、PR2Respectively comprising a red line segment point cloud and red square point cloud;Further, using Europe
Formula clustering procedure, by PR1It is divided into red line segment point cloud PR1LineWith red square point cloud PR1Squ, by PR1It is divided into red line segment point
Cloud PR2LineWith red square point cloud PR2Squ。
Calculate separately PR1Line、PR2LineAnd blue target point cloud PBBarycentric coodinates gR1L(xR1L,yR1L,zR1L)、gR2L
(xR2L,yR2L,zR2L)、gB(xB,yB,zB), calculate separately gR1L、gR2LOpposite gBDistance Dis1, Dis2, gR1L、 gR2LSpacing
Dis。
The unit vector of primary Calculation X-axis:
If Dis1 < Dis2,If Dis1 > Dis2,
(4.3.3) determines Z axis forward direction unit vector: calculating red target point cloud PRUnit normal vectorThe plant of green
Object-point cloud PGBarycentric coodinates gG(xG,yG,zG) and by gCIt is directed toward gGVectorSince hardware platform Green plant is put
Be put in above colored target, thus Z axis forward direction unit vector withAngle certainly less than 90 °;IfWithAngle be θ,
It calculatesIf result is greater than 0,WithAngle theta meets preset coordinate direction, Z axis is just less than 90 °
To unit vectorIf result less than 0,WithAngle theta is greater than 90 °, and contrary with preset coordinate, Z axis is positive single
Bit vector
(4.3.4) determines Y-axis forward direction unit vector: calculate both perpendicular toWithUnit vectorAs
Y-axis forward direction unit vector.
In actual application test, the point cloud that discovery SfM algorithm is rebuild has the feature that larger for color change
Boundary Recognition rate it is higher, irregular cavity is easy to produce to the reconstruction of continuously distributed solid color partial interior.This will be right
The center of gravity calculation of feature brings uncertainty: if being unevenly distributed inside colored target point cloud, then its whole center of gravity can be with
The colored actual center of gravity of target shifts, and generates certain error;Again due to it is such be unevenly distributed be it is uncontrollable, because
This this error is also random uncontrollable.
Bring adverse effect is standardized to solve above situation to coordinate system, is made without using continuously distributed solid color
For the feature of colored target, and line segment should be used as essential characteristic;But in the Z axis of normalized coordinate system, need
It will be by calculating the whole normal vectors of all red target point clouds as the unit vector of Z axis, if line segment conduct coloured silk is used only
The essential characteristic of color mark plate may then will affect the selection of Z axis.
For this purpose, design battery of tests, illustrates the stability effect of three kinds of colored target scheme normalized coordinates systems generated
Fruit.
As shown in figure 4, figure (a), (b), (c) are respectively line segment type, diamond type, comprehensive colored target, each piece of colour
There is identical blue rectangle target in the left end of target, and in addition to this, (a) colour target contains only that there are two symmetrical red line segments
Target, (b) colored target contains only that there are two symmetrical red square targets, and (c) colored target then marks both the above red
It is hardened to be combined.The one small-sized square 12 (as shown in Figure 5) of blue of fixed placement in platform, is distinguished using three kinds of colored targets
Three groups of point clouds are acquired from three different angles to 10 basin green plants, and generate normalized coordinates system.Calculate every group of point cloud standard
Change the barycentric coodinates of small-sized 12 clouds of square of blue in coordinate system, each scheme barycentric coodinates distribution such as attached drawing 6, attached drawing generated
Shown in 7.
The barycentric coodinates entirety center of gravity of all blues 12 clouds of small-sized square identified in every kind of scheme is calculated, and every
The Euclidean distance of a whole center of gravity of barycentric coodinates distance, by comparing the variance of Euclidean distance in every kind of scheme, variance is smaller, says
Bright barycentric coodinates distribution is more concentrated, and the stability for generating normalized coordinates system is better.As can be known from the results of Table 1, with regard to variance yields and
Speech, the normalized coordinates system that scheme three generates when being less than scheme one less than scheme two, i.e. three colour target of operational version are most stable;
And line segment feature is better than square feature.
The small-sized square point cloud barycentric coodinates stability of blue under the different colored target schemes of table 1
Based on this, the present invention can also expand out new prioritization scheme on the basis of former scheme, and the prioritization scheme is in scheme two
On the basis of, red square is all transformed into several dice square, specifically: the colored target after the optimization is distributed with red
Color, blue two kinds of targets, using black as background color, wherein red target is divided into red sub- target 1, red sub- target two
4102, it is symmetric with the symmetry axis of colored target entirety;Red sub- target 1, red sub- target 2 4102 are by several
A sub- square composition of red (number is at least 5, and meets central symmetry distribution simultaneously);Equally have in one end of colored target
The blue target of one rectangle, specific distribution are as shown in Figure 8.For the prioritization scheme relative plan two, reduce solid color
It is continuously distributed, increase boundary characteristic, can make generate red target point cloud feature identification it is more acurrate;It is five small simultaneously red
Color square is centrosymmetric distribution with respect to center, even if there is certain cavity inside each small red square point cloud, to whole
Offset caused by the position of centre of gravity of body is as this is distributed and maintains in lesser range, and then promotes normalized coordinates system
The stability of generation.
Schematically the utility model and embodiments thereof are described above, description is not limiting, attached drawing
Shown in be also one of the embodiments of the present invention, actual structure is not limited to this.So if this field
Those of ordinary skill enlightened by it, without deviating from the purpose of the present invention, not inventively design
Frame mode similar with the technical solution and embodiment, all should belong to the protection range of the utility model.