CN204924273U - Three -dimensional laser sensor data calibrating device of panorama - Google Patents

Abstract

The utility model relates to a three -dimensional laser sensor data calibrating device of panorama. The device is including three -dimensional laser sensor and calibration device. Three -dimensional laser sensor includes rotatory cloud platform and sets up in the two -dimentional laser sensor of rotatory cloud bench, calibration device is located the settlement distance in three -dimensional laser sensor the place ahead. The utility model provides a manual human error who selects the matched pair to introduce uncontrollable with the big problem of time expense, consider fully that the influence number is according to the possible factor of precision, improvement transducer calibration precision to calibration method's commonality has been promoted.

Description

A kind of panorama three-dimensional laser sensing data calibrating installation

Technical field

The utility model belongs to technical field of environmental perception, and a kind of device improving three-dimensional laser range measurement system collection panoramic view data precision, particularly relates to a kind of panorama three-dimensional laser sensing data calibrating installation specifically.

Background technology

Laser sensor belongs to active sensor, antijamming capability is strong, depth information accurately can be provided, environment geometric properties can be described in detail, the three-dimensional model of direct quick obtaining environment, therefore dominate in three-dimensional environment perception, has larger using value in fields such as comprising the navigation of intricately anthropomorphic robot, engineering survey.At present, have many manufacturers of selling three-dimensional laser detection systems in the world, as Austrian Riegl, Switzerland Leica, Japanese Topcon, U.S. Faro etc., these sensors are expensive, and often complex structure, be unfavorable for safeguarding and maintenance.

In order to reduce costs, most three-dimensional laser sensors on the basis of two-dimensional laser, adds one dimension rotate, thus obtain three-dimensional laser data (Jes ú sMorales, JorgeL.Mart í nez, AnthonyMandow, Alejandro andAlfonsoGarc í a-Cerezo, " DesignandDevelopmentofaFastandPreciseLow-Cost3DLaserRang efinder ", inProceedingsofthe2011IEEEInternationalConferenceonMecha tronics, 2011).Due to when installation two-dimensional laser sensor, the deviation on angle and position cannot being avoided, so in order to obtain three-dimensional laser data accurately, before this three-dimensional laser sensor of use, needing to demarcate sensor inherent parameters.

Document (Yu Zujun, Yang Yanan, Zhu Liqiang, " research of three-dimensional laser scanning measurement system scaling method ", electronic surveying and instrument journal, 21 (6), 31-35,2007) the demarcation object in is made up of a LMS200 two-dimensional laser sensor and a translating device, utilize two kinds of calibrating templates to realize rotation parameter between three-dimensional laser sensor coordinate system and extraneous measuring basis coordinate and translation parameters, this parameter can not correct the data error because installation question produces.

Document (Jes ú sMorales, JorgeL.Mart í nez, AnthonyMandow, AntonioJ.Reina, Alejandro andAlfonsoGarc í a-Cerezo, " BoresightCalibrationofConstructionMisalignmentsfor3DScan nersBuiltwitha2DLaserRangefinderRotatingonItsOpticalCent er ", Sensors, 14 (11), 20025-20040, 2014.) Nelder-Mead method is utilized to carry out flatness and the process of area largest optimization to panel data, demarcate two anglecs of rotation of two-dimensional laser sensor relative to three-dimensional data coordinate system, although three-dimensional data more accurately can be obtained, but owing to have ignored partial 3-D laser internal parameter, in time there is alignment error, still certain data error can be introduced.Equally, document (Gu Xiaojie, Bu Chunguang, Chen Cheng, week dredges wise man, " three-dimensional laser Ranging System and scaling method are studied ", Shenyang Univ. of Science and Engineering's journal, 33 (5), 10-14,47,2014) only considered the partial interior parameter of three-dimensional laser sensor, can not demarcate various three-dimensional laser sensor as general method.

Utility model content

For the above-mentioned problems in the prior art, problem to be solved in the utility model proposes a kind of panorama three-dimensional laser sensing data calibrating installation.

The technical scheme that the utility model is adopted for achieving the above object is: a kind of panorama three-dimensional laser sensing data calibrating installation comprises three-dimensional laser sensor and caliberating device; Described three-dimensional laser sensor comprises rotary head and is arranged at the two-dimensional laser sensor on rotary head; Described caliberating device is positioned at the setpoint distance in three-dimensional laser sensor front.

Described caliberating device is printing paper, square black square and white square that printing paper surface air brushing has n × m mutually to replace; There is hollow out the center of each black square and white square.

The shape of described hollow out is circular.

Described iteration optimization algorithms is least square method or Newton tangential method.

The utility model has following beneficial effect and advantage:

1. the utility model is only using three-dimensional laser sensor as measurement mechanism, with feature selected from environment for basis, realizes the calibration to panorama three-dimensional laser data.

2. the utility model solves panorama three-dimensional laser sensor internal parameter calibration problem, just can realize multiple unique point easily automatically extract by means of only a data acquisition.

3. the utility model solves manually to click and mates to introduced personal error the uncontrollable and large problem of time overhead, takes into full account the possible factor affecting data precision, improves transducer calibration precision, and improve the versatility of scaling method.

4. the utility model is that the transducer calibration of panorama three-dimensional laser lays the foundation, and can be used on the fields such as the navigation of intricately anthropomorphic robot, engineering survey.

5. the calibration steps that the utility model proposes considers position between the two-dimensional laser coordinate system that may exist and three-dimensional laser coordinate system and rotation error comprehensively, improves the versatility of method.

6. the caliberating device that the utility model adopts is convenient to make, easy to carry, reusable.

Accompanying drawing explanation

Fig. 1 is panorama three-dimensional laser sensor model schematic diagram;

Fig. 2 is three-dimensional laser sensor 3 angular error schematic diagram one;

Fig. 3 is three-dimensional laser sensor 3 angular error schematic diagram two;

Fig. 4 is three-dimensional laser sensor 3 angular error schematic diagram three;

Fig. 5 is caliberating device schematic diagram;

Fig. 6 is caliberating device angle point grid schematic diagram;

Fig. 7 is three-dimensional laser sensor construction schematic diagram;

Fig. 8 is panorama three-dimensional laser sensor;

Fig. 9 is for correcting panorama three-dimensional laser point cloud data experimental result picture in cup;

Figure 10 is the cloud data experimental result picture after correcting.

Embodiment

Below in conjunction with drawings and Examples, the utility model is described in further detail.

The maximum innovative point of the utility model is the angle comprehensively considering may exist in installation process and site error, is applicable to multiple panorama three-dimensional laser sensor parameters and demarcates.Between two-dimensional laser coordinate system and three-dimensional laser coordinate system three angle parameter and one are deflected heart distance parameter as parameter to be calibrated by the utility model for this reason, through plane fitting, laser data binaryzation, angle point grid process, the unique point belonging to caliberating device is extracted respectively from two parts three-dimensional laser data gathered, adopt iterative optimization method to treat calibrating parameters to solve, thus the angle obtained between two-dimensional laser coordinate system and three-dimensional laser coordinate system and location parameter.

The utility model is using the angle parameter of 3 between two-dimensional laser coordinate system and three-dimensional laser coordinate system and 1 location parameter as parameter to be calibrated, utilize the phenomenon that the angle point in two parts laser data on caliberating device can not overlap, adopt iterative optimization method to solve calibrating parameters, the data that this calibrating parameters can meet multiple panorama three-dimensional laser sensor correct demand.Concrete steps are as follows:

A) be placed in Experimental Area by caliberating device, the OK range of the spacing of caliberating device and three-dimensional laser sensor is between 2m-10m;

B) scan-data of SICKLMS200 laser sensor is pressed y _lbe positive and negatively equally divided into two parts, every part scanning angle is 90 degree, and when rotary head drives LMS200 laser rotary 360 degree time, positive and negative two parts laser data just in time forms two space three-dimensional cloud datas;

C) according to the putting position of caliberating device, filter out the data in caliberating device region, parallel planes matching of going forward side by side, foundation caliberating device region laser spots data carry out binary conversion treatment to the distance value d of institute's fit Plane to area data

$I_i=\left\{\begin{array}{cc}255& d_i<\stackrel{\&OverBar;}{d}\\ 0& d_i>=\stackrel{\&OverBar;}{d}\end{array}\right.$

Wherein I _ifor the value after i point binaryzation in region, for experience distance threshold, after binaryzation, the data in hollow out circular hole region are 0, and other area datas are 255, and by the distribution histogram of counting of row statistics 0 and 255, statistics obtains row coordinate by row

$\left\{\begin{array}{c}i=median\left(\arg \underset{i}{max}\underset{j}{\overset{bin}{\&Sigma;}}1\left(I_{ij}=0\right)\right)\\ j=median\left(\arg \underset{j}{max}\underset{i}{\overset{bin}{\&Sigma;}}1\left(I_{ij}=0\right)\right)\end{array}\right.$

Wherein i, j represent ranks subscript respectively, and function 1 () is 0-1 function, if condition is very in bracket, being worth is 1, otherwise be 0, median () are intermediate value choice function, I _ijrepresent binaryzation value, thus determine hollow out centre coordinate, and can further according to four neighborhood relationships determination black and white lattice angular coordinates;

D) distance that in two parts three-dimensional laser point cloud, corresponding angle point is right is calibration process is optimization distance formula

$D=\underset{\mathrm{\&alpha;},\mathrm{\&beta;},\mathrm{\&gamma;},l}{min}\underset{k=1}{\overset{n}{\&Sigma;}}\left|\right|P_P^{k+}-P_P^{k-}\left|\right|=\underset{\mathrm{\&alpha;},\mathrm{\&beta;},\mathrm{\&gamma;},l}{min}\underset{k=1}{\overset{n}{\&Sigma;}}\left|\right|(R_{LP}^{k+}{\mathrm{RP}}_L^{k+}+T_{LP}^{k+})-(R_{LP}^{k-}{\mathrm{RP}}_L^{k-}+T_{LP}^{k-})\left|\right|$

Wherein R _lPand T _lPbe respectively rotation and translation matrix that two-dimensional laser coordinate is tied to three-dimensional laser ordinate transform, l is the site error between two-dimensional laser coordinate system and three-dimensional laser coordinate system,

$R=\left[\begin{array}{ccc}\cos \mathrm{\&beta;}\cos \mathrm{\&gamma;}-\sin \mathrm{\&alpha;}\sin \mathrm{\&beta;}\sin \mathrm{\&gamma;}& -\cos \mathrm{\&beta;}\sin \mathrm{\&gamma;}-\sin \mathrm{\&alpha;}\sin \mathrm{\&beta;}\cos \mathrm{\&gamma;}& -\cos \mathrm{\&alpha;}\sin \mathrm{\&beta;}\\ \cos \mathrm{\&alpha;}\sin \mathrm{\&gamma;}& \cos \mathrm{\&alpha;}\cos \mathrm{\&gamma;}& -\sin \mathrm{\&alpha;}\\ \sin \mathrm{\&beta;}\cos \mathrm{\&gamma;}+\sin \mathrm{\&alpha;}\cos \mathrm{\&beta;}\sin \mathrm{\&gamma;}& -\sin \mathrm{\&beta;}\sin \mathrm{\&gamma;}+\sin \mathrm{\&alpha;}\cos \mathrm{\&beta;}\cos \mathrm{\&gamma;}& \cos \mathrm{\&alpha;}\cos \mathrm{\&beta;}\end{array}\right]$

α, β, γ are the spaced winding x of two-dimensional laser coordinate system and three-dimensional laser coordinate system _laxle, y _laxle, z _lthe error angle that axle rotates, utilizes iteration optimization algorithms, obtains calibration result, be i.e. the actual value of α, β, γ and l.

Specific embodiments of the present utility model is as follows:

1. panorama three-dimensional laser sensor characteristic is analyzed

The panorama three-dimensional laser sensor be made up of two-dimensional laser sensor and rotary head, its system model as shown in Figure 1, can use formula to describe:

Wherein, with be respectively the coordinate of a laser spots in two-dimensional laser coordinate system and three-dimensional laser coordinate system, be the The Cloud Terrace anglec of rotation corresponding to i-th group of laser data, with be respectively rotation and translation matrix that two-dimensional laser coordinate is tied to three-dimensional laser ordinate transform.Different owing to often organizing the corresponding The Cloud Terrace anglec of rotation of laser data, so rotation matrix and translation matrix are not fixed.What above formula described is ideal coordinates system relation, but in actual applications, owing to there is alignment error, above-mentioned ideal relationship formula cannot be met, rotation between needing equipment and translation relation are estimated, namely systematic parameter is demarcated, need to demarcate the rotation relationship between two-dimensional laser coordinate system and three-dimensional laser coordinate system and translation relation.

Owing to there is alignment error, need consideration three angular errors when carrying out coordinate system rotational transform, as shown in figs. 2 to 4.Definition rotation matrix R, by laser spots from there is the laser scanning Plane Rotation of error to ideal situation, after namely rotating, can there are not α, β and γ tri-angular errors in this matrix between laser scanning plane and rotary head plane.Coordinate transform formula after rectification is expressed as:

$P_P^{ij}=R_{LP}^i{\mathrm{RP}}_L^{ij}+T_{LP}^i---\left(2\right)$

So need the parameter of demarcating to have 4: α, β, γ and l.

2. the design of caliberating device

Different medium has different reflectivity to the laser beam that laser sensor is launched, wherein the reflectivity of glossy object is higher than matt object, same different incidence angles is on the impact also highly significant of range finding, but target object surface Color pair range finding impact is little.For above-mentioned laser characteristics, develop three-dimensional laser caliberating device (as shown in Figure 5), the material of device is glossy printing paper, to ensure having high reflectance to laser.Go out the square black and white lattice of 4 × 5 in printing paper surface air brushing, the length of side of each black and white lattice is 25cm.Hollow out is carried out, the circle of hollow out shape to be radius be 5cm in each black and white lattice center.Through experimental verification, select the circular impact being robustness in order to ensure algorithm as hollow out shape and avoiding Laser edge effect to produce.

3. demarcate the extraction of required laser data unique point

(1) the rough position scope of caliberating device is determined.According to the putting position of caliberating device, the coordinate parameters of caliberating device in three-dimensional laser coordinate system is set in scaling method, filters out the data in this region according to this parameter in conjunction with laser spots three-dimensional coordinate.

(2) according to scope determination plane.According to the data point fit Plane in region, namely

$a=\arg \underset{a}{min}\underset{i=1}{\overset{n}{\&Sigma;}}\left|\right|a^T{q}_i|{|}^2---\left(3\right)$

Wherein a=[a ₁, a ₂, a ₃] ^tfor planar process vector, q _i=[x, y, z] ^tfor area coordinate point.In order to improve counting yield, directly plane equation can be determined according to four of a device in-scope boundary point.N is the laser spots number in region.

(3) laser ranging data binaryzation.Foundation caliberating device region laser spots data carry out binary conversion treatment to the distance value d of institute's fit Plane to area data.

$I_i=\left\{\begin{array}{cc}255& d_i<\stackrel{\&OverBar;}{d}\\ 0& d_i>=\stackrel{\&OverBar;}{d}\end{array}\right.---\left(4\right)$

Wherein I _ifor the value after i point binaryzation in region, for experience distance threshold, =15cm.After binaryzation, the data in hollow out circular hole region are 0, and other area datas are 255, thus can use for reference image processing method further and carry out angle point grid.

(4) angle point is extracted.Caliberating device simulation reduced graph as shown in Figure 6,7, in the black and white lattice of 4 × 5, circular expression void region, the cross at center represents hollow out center, and the round dot between void region represents the angle point of black and white lattice, and plane evenly divides by dotted line.To utilize in (3) the data obtained after binaryzation, first by the distribution histogram of counting of row statistics 0 and 255, and according to the row-coordinate at peak point determination hollow out center, in like manner also can add up by row and obtain row coordinate, specifically see following formula:

$\left\{\begin{array}{c}i=median\left(\arg \underset{i}{max}\underset{j}{\overset{bin}{\&Sigma;}}1\left(I_{ij}=0\right)\right)\\ j=median\left(\arg \underset{j}{max}\underset{i}{\overset{bin}{\&Sigma;}}1\left(I_{ij}=0\right)\right)\end{array}\right.---\left(5\right)$

Wherein i, j represent ranks subscript respectively, and function 1 () is 0-1 function, if condition is very in bracket, being worth is 1, otherwise is 0.Median () is intermediate value choice function, I _ijrepresent binaryzation value.Thus determine the hollow out centre coordinate of 4 × 5, and the black and white lattice angular coordinate of 3 × 4 can be determined further according to four neighborhood relationships.

4. calculate calibrating parameters value

Due to the existence of alignment error, as shown in Figure 8, when The Cloud Terrace rotating 360 degrees gathers laser data time, the panorama three-dimensional laser point cloud of composition can produce distortion.In order to calculate calibrating parameters, laser data is divided into two parts by region, laser spots y _l>0, uses P ⁺represent, laser spots y _l<0, uses P ^-represent.When The Cloud Terrace rotating 360 degrees, build two three-dimensional point clouds, there is one-to-one relationship from two groups of angle points two some clouds, the distance of corresponding angle point is calibration process is corresponding laser spots range formula between optimization two some clouds:

$D=\underset{\mathrm{\&alpha;},\mathrm{\&beta;},\mathrm{\&gamma;},l}{min}\underset{k=1}{\overset{n}{\&Sigma;}}\left|\right|P_P^{k+}-P_P^{k-}\left|\right|=\underset{\mathrm{\&alpha;},\mathrm{\&beta;},\mathrm{\&gamma;},l}{min}\underset{k=1}{\overset{n}{\&Sigma;}}\left|\right|(R_{LP}^{k+}{\mathrm{RP}}_L^{k+}+T_{LP}^{k+})-(R_{LP}^{k-}{\mathrm{RP}}_L^{k-}+T_{LP}^{k-})\left|\right|---\left(6\right)$

In order to two groups of angle points as input, can adopt the optimized algorithm such as least square method or Newton tangential method, according to formula ${\&lsqb;{\mathrm{\&alpha;}}^{k+1},{\mathrm{\&beta;}}^{k+1},{\mathrm{\&gamma;}}^{k+1},l^{k+1}\&rsqb;}^T={\&lsqb;{\mathrm{\&alpha;}}^k,{\mathrm{\&beta;}}^k,{\mathrm{\&gamma;}}^k,l^k\&rsqb;}^T+\frac{D}{D^{\&prime;}}$ Iterative computation α, β, γ and l, k and k+1 represents kth and kth+1 iteration respectively, and D' is the derivative of D to α, β, γ and l.

In order to verify the validity of this method, the sensing system constructed by Fig. 6 is utilized to carry out the checking of scaling method.Panorama three laser sensor is made up of SICKLMS200 type laser sensor and rotary head, and wherein laser sensor plane scan angles is 0-180 degree, and the range of application of the frequency of The Cloud Terrace stepper motor is 500-2500Hz.Driven by motor laser sensor is utilized to obtain the three-dimensional laser ranging data of scene.

Be placed on by caliberating device in Experimental Area, the OK range of the spacing of caliberating device and three-dimensional laser sensor is between 2m-10m.The scan-data of LMS200 laser sensor is pressed y _lbe positive and negatively equally divided into two parts, every part scanning angle is 90 degree, and when rotary head drives LMS200 laser rotary 360 degree time, positive and negative two parts laser data just in time forms two space three-dimensional cloud datas.After plane fitting, laser data binaryzation, angle point grid process, the Corresponding matching point pair of two groups of data can being obtained, to point to carrying out Iterative matching, three-dimensional laser transducer calibration parameter can be calculated:

$\left\{\begin{array}{c}\mathrm{\&alpha;}=0.0095rad\\ \mathrm{\&beta;}=0.0017rad\\ \mathrm{\&gamma;}=-0.0353rad\\ l=0.004m\end{array}\right.$

From angle analysis qualitatively, by judging whether two groups of laser data overlap visual verification is carried out to calibration result.Ideally, two parts laser data of synchronization collection should overlap completely, but owing to there is error, as shown in Figure 9, there is certain deviation in two three-dimensional laser point clouds, utilizes calibrating parameters to correct three-dimensional laser data, and Figure 10 shows two groups of laser point cloud datas after rectification, can find out that laser data is more accurate, can description three-dimensional scenic strictly according to the facts.

Claims (3)

1. a panorama three-dimensional laser sensing data calibrating installation, is characterized in that, comprises three-dimensional laser sensor and caliberating device; Described three-dimensional laser sensor comprises rotary head and is arranged at the two-dimensional laser sensor on rotary head; Described caliberating device is positioned at the setpoint distance in three-dimensional laser sensor front.

2. a kind of panorama three-dimensional laser sensing data calibrating installation according to claim 1, it is characterized in that, described caliberating device is printing paper, square black square and white square that printing paper surface air brushing has n × m mutually to replace; There is hollow out the center of each black square and white square.

3. a kind of panorama three-dimensional laser sensing data calibrating installation according to claim 2, is characterized in that, the shape of described hollow out is circular.