CN106023213A - Multi-camera system calibration method based on cylindrical surface - Google Patents

Abstract

The invention relates to a multi-camera system calibration method based on a cylindrical surface. A plurality of characteristic points are constructed on the cylindrical surface, and cameras surrounding the cylindrical surface are enabled to shoot at least six characteristic points; three-dimensional coordinate systems are established with the axis and radiuses of the cylindrical surface serving as coordinate axes, and the coordinate systems based on different radius directions are marked; according to the construction method of the characteristic points and the size of the cylindrical surface, the three-dimensional coordinates of the characteristic points and the conversion relation among the coordinate systems based on different radius directions are calculated; each camera shoots the images of the cylindrical surface, the image coordinates of the characteristic points are extracted, calibration initial values of each camera are obtained by means of direct linear conversion, and all internal and external parameters of each camera are obtained by means of non-linear optimization; and according to the relative relation between each camera and the cylindrical surface, the relative relations among the cameras are calculated, and the calibration of the multi-camera system is completed. According to the invention, the defects of the prior art are overcome, the calibration of the large-view-field wide-base-line and nonsynchronous multi-camera system is facilitated.

Description

A kind of multi-camera system scaling method based on the face of cylinder

Technical field

The present invention relates to the scaling method of multi-camera system, be specifically related to a kind of multi-camera system based on the face of cylinder Scaling method, it is adaptable to computer vision system and technology.

Background technology

By experimental technique, the demarcation of multi-camera system, i.e. determines that the intrinsic parameter of each video camera and mutual position are closed System, is the multi-camera system key link that is applied to the field such as target following and three-dimensional reconstruction.Knot is demarcated accurately for obtaining Really, need in the common visual field of multiple-camera, arrange demarcation thing known to a physical dimension.According to the demarcation thing used not Four classes can be divided into, the scaling method of current multi-camera system:

Use luminous point for demarcating thing (T.Svoboda, D.Martinec, and T.Pajdla, A convenient multi-camera selfcalibration for virtual environments,PRESENCE:Teleoperators And Virtual Environments, 2005,14 (4): 407-422), in common visual field, mobile luminous point repeatedly, utilizes certainly Multi-camera system is demarcated by scaling method.This kind of method needs multiple video camera sync pulse jamming luminous points, it is impossible to be applied to Asynchronous multi-camera system, limits its application.

Employing one-dimension calibration thing (pay strong, full powers, Cai Kaiyuan, multiple-camera parameter based on freely-movable one-dimension calibration thing Scaling method and experiment, control theory and application, 2014,31 (8): 1018-1024), demarcate the most two-by-two, followed by Geometrical constraint between characteristic point on fundamental matrix and one-dimension calibration thing, estimates the inside and outside parameter of two video cameras, finally uses Short circuit shot and harness adjust and multi-camera system are carried out global calibration.This method needs in the common visual field of multiple-camera In to wave one-dimension calibration thing at random up to a hundred time, just can obtain preferable result, this is a time-consuming job.If multiple-camera The configuration of system changes, then needs to repeat such time consuming work, is not easy to joining flexibly of multi-camera system Put.

Use two-dimensional calibrations thing (T.Ueshiba and F.Tomita, Plane-based calibration algorithm for multi-camera systems via factorization of homography matrices, in Proc.of IEEE International Conference on Computer Vision,Nice,France, Oct.2003, pp.966-973), it is that single camera demarcates current most common method, but for many shootings of loop configurations Machine system, there is occlusion issue in this kind of method, is i.e. positioned at two-dimensional calibrations thing video camera behind and cannot see that the mark demarcated on thing Determine pattern, thus can not demarcate simultaneously.And, this kind of method needs mark to be varied multiple times during shooting uncalibrated image Earnest, relative to the attitude of video camera, needs artificial participation, is not easy to automatic Calibration.

Use three-dimensional scaling thing (Jae-Hean Kim, Bon-Ki Koo, Convenient calibration method For unsynchronized multi-camera networks using a small reference object, in Proc.of IEEE/RSJ International Conference on Intelligent Robots and Systems, Vilamoura, Algarve, Portugal, Oct.2012, pp.438-444), it is only necessary to shooting piece image can be taken the photograph more The demarcation of camera system.But this kind of three-dimensional scaling thing many employings polyhedral structure, needs when making to ensure between Different Plane Precise positional relationship, need accurate process equipment, be not easy to popularize.For single camera, plane reference figure can be used Scaling method (Jianhua Wang, Zhenyi Li, Fuxin Ren, the Zhao Liu, Aidi that case combines with the face of cylinder Shen, Calibration of Vision System Used on Unmanned Surface Vehicle, in Proc.of OCEANS ' 16MTS/IEEE Shanghai, April 2016), it is simple to the making of three-dimensional scaling thing, but this method only considers The demarcation of single camera, it is impossible to be applied to multi-camera system.

Along with the development of computer technology and popularizing of digital camera, the application of multi-camera system is increasingly extensive, because of This, the multi-camera system scaling method of simpler and more direct practicality is wished in this area.

Summary of the invention

The present invention is directed to the deficiency of the existing scaling method of multi-camera system, it is provided that a kind of multiple-camera based on the face of cylinder System calibrating method.Thing of demarcating in the method need not the process equipment of precision, it is simple to demarcates the making of thing；Timing signal is each Video camera only needs to shoot piece image, it is simple to the automatic Calibration of multi-camera system；When increasing shooting in multi-camera system When machine or change video camera, it is only necessary to demarcate the inside and outside parameter of this video camera, this video camera can be closed with other video camera Connection, completes the renewal of multi-camera system calibrating parameters, it is simple to the flexible configuration of multi-camera system.

For reaching above-mentioned purpose, the present invention adopts the following technical scheme that:

A kind of multi-camera system scaling method based on the face of cylinder, it is characterised in that described scaling method includes as follows Step:

(1) on the face of cylinder, multiple characteristic point is constructed so that the video camera around the described face of cylinder can photograph at least 6 Individual described characteristic point, preferably can photograph at least 18 described characteristic points；

(2) axis and radius with the described face of cylinder set up three-dimensional system of coordinate for coordinate axes, and to based on different radii side To coordinate system make labelling；

(3) according to building method and the size on the described face of cylinder of described characteristic point, described characteristic point is calculated described three Transformation relation between three-dimensional coordinate in dimension coordinate system, and described three-dimensional system of coordinate based on different radii direction；

(4) shoot the image on the described face of cylinder with each video camera, extract the image coordinate of described characteristic point, by direct line Property conversion obtain the demarcation initial value of each video camera, and obtained whole inside and outside parameter of each video camera by nonlinear optimization；

(5) according to the relativeness of described each video camera Yu the described face of cylinder, calculate between described each video camera is relative Relation, completes the demarcation of multi-camera system.

Instant invention overcomes the occlusion issue existed when using plane reference object, it is simple to big visual field, wide baseline, asynchronous many The automatic Calibration of camera chain.

Accompanying drawing explanation

The present invention is further illustrated below in conjunction with the drawings and specific embodiments.

Fig. 1 is demarcation thing and the video camera distribution schematic diagram of an embodiment.

Fig. 2 is the reference object image schematic diagram of video camera 200 shooting in embodiment illustrated in fig. 1.

Fig. 3 is the reference object image schematic diagram of video camera 300 shooting in embodiment illustrated in fig. 1.

Fig. 4 is the reference object image schematic diagram of video camera 400 shooting in embodiment illustrated in fig. 1.

Fig. 5 is the reference object image schematic diagram of video camera 500 shooting in embodiment illustrated in fig. 1.

Fig. 6 is the characteristic point schematic diagram extracted from reference object image shown in Fig. 3.

Fig. 7 is camera coordinate system and the schematic diagram demarcating article coordinate system in embodiment illustrated in fig. 1.

Detailed description of the invention

For the technological means making the present invention realize, creation characteristic, reach purpose and be easy to understand with effect, below knot Conjunction is specifically illustrating, and the present invention is expanded on further.

The occlusion issue existed when the present invention is directed to use plane reference object and the problem needing to shoot several different images, And the demarcation thing existed when using three-dimensional scaling thing makes difficult problem, it is provided that a kind of multi-camera system based on the face of cylinder Scaling method.The step of the method is as follows:

(1) on the face of cylinder, multiple characteristic point is constructed so that the video camera around the described face of cylinder can photograph at least 6 Individual described characteristic point, preferably can photograph at least 18 described characteristic points；

(2) axis and radius with the described face of cylinder set up three-dimensional system of coordinate for coordinate axes, and to based on different radii side To coordinate system make labelling；

(3) according to building method and the size on the described face of cylinder of described characteristic point, described characteristic point is calculated described three Transformation relation between three-dimensional coordinate in dimension coordinate system, and described three-dimensional system of coordinate based on different radii direction；

(4) shoot the image on the described face of cylinder with each video camera, extract the image coordinate of described characteristic point, by direct line Property conversion obtain the demarcation initial value of each video camera, and obtained whole inside and outside parameter of each video camera by nonlinear optimization；

(5) according to the relativeness of described each video camera Yu the described face of cylinder, calculate between described each video camera is relative Relation, completes the demarcation of multi-camera system.

Based on above-mentioned principle, the specific implementation process of the present invention is as follows:

(1) on the face of cylinder, multiple characteristic point is constructed so that the video camera around the described face of cylinder can photograph at least 6 Individual described characteristic point, preferably can photograph at least 18 described characteristic points.Fig. 1 is the demarcation thing of an embodiment and video camera divides Cloth schematic diagram, on the face of cylinder 110, printing is the pattern 120 of plane grid after launching, and constitutes cylinder and demarcates thing 100, white The intersection point 123 of grid 121 and black box 122 is as characteristic point.According to the working range of visual system, determine the half of the face of cylinder Footpath is 90.5mm, and the length of side of white boxes 121 and black box 122 is 23.5mm, the most respectively distribution white boxes 121 He Each 120 of black box 122 so that the video camera 200,300,400 and 500 around the face of cylinder 110 can photograph at least 6 Described characteristic point 123, preferably can photograph at least 18 described characteristic points 123.Fig. 2, Fig. 3, Fig. 4 and Fig. 5 are respectively video camera 200, the reference object image schematic diagram of 300,400 and 500 shootings, they all photograph at least 77 characteristic points 123.

(2) axis and radius with the described face of cylinder set up three-dimensional system of coordinate for coordinate axes, and to based on different radii side To coordinate system make labelling.In the embodiment shown in fig. 1, with the axis on the face of cylinder 110 as ξ_ZAxle, crosses the radius of characteristic point For ξ_XAxle, sets up three-dimensional cartesian coordinate system ξ according to right-hand rule_Xξ_Yξ_Z.So, cross each characteristic point and can set up a coordinate system, For distinguishing the coordinate system of different characteristic point, the face of cylinder 110 is made labelling 130, in this embodiment with numeral labelling not Same radial direction.In fig. 2, the ξ of set up coordinate system_XThe axle characteristic point by labelling 1 direction, is expressed as this coordinate system ξ₁, its three coordinate axess are respectively ξ_1X,ξ_1Y,ξ_1Z；In figure 3, the ξ of set up coordinate system_XThe axle feature by labelling 7 direction Point, is expressed as ξ by this coordinate system₇, its three coordinate axess are respectively ξ_7X,ξ_7Y,ξ_7Z；In the diagram, the ξ of set up coordinate system_XAxle leads to Cross the characteristic point in labelling 13 direction, this coordinate system is expressed as ξ₁₃, its three coordinate axess are respectively ξ_13X,ξ_13Y,ξ_13Z；At Fig. 5 In, the ξ of set up coordinate system_XThe axle characteristic point by labelling 19 direction, is expressed as ξ by this coordinate system₁₉, its three coordinate axess divide Wei ξ_19X,ξ_19Y,ξ_19Z。

(3) according to building method and the size on the described face of cylinder of described characteristic point, described characteristic point is calculated described three Transformation relation between three-dimensional coordinate in dimension coordinate system, and described three-dimensional system of coordinate based on different radii direction.At Fig. 3 Shown coordinate system ξ₇In, it is assumed that the length of side of plane grid is a, and the radius on the face of cylinder is r, in this embodiment a=23.5mm, r =90.5mm, (m n) represents that this feature point is at circumferencial direction and ξ to the two-dimensional coordinate of characteristic point_7XAxle and the circular arc of face of cylinder intersection point Distance is m times of length of side a, at axial and ξ_7XThe distance of axle and face of cylinder intersection point is n times of length of side a, then this feature point is at ξ₇In Three-dimensional coordinate be represented by:

$\left[\begin{array}{c}{\mathrm{\ξ}}_{7x}\\ {\mathrm{\ξ}}_{7y}\\ {\mathrm{\ξ}}_{7z}\end{array}\right]=\left[\begin{array}{c}rcos\left(\frac{a\·m}{r}\right)\\ rsin\left(\frac{a\·m}{r}\right)\\ a\·n\end{array}\right]---\left(1\right)$

From coordinate system ξ shown in Fig. 2₁To coordinate system ξ shown in Fig. 3₇, coordinate system ξ₁Around ξ_1ZThe angle, θ that axle rotates can be by turning over Number of squares w represents, in this embodiment w=7.

$\mathrm{\θ}=\frac{(w-1)a}{r}---\left(2\right)$

Therefore, from coordinate system ξ₁To ξ₇Spin matrixFor:

${R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_7}=\left[\begin{array}{ccc}cos\mathrm{\θ}& -sin\mathrm{\θ}& 0\\ sin\mathrm{\θ}& cos\mathrm{\θ}& 0\\ 0& 0& 1\end{array}\right]---\left(3\right)$

From coordinate system ξ₁To ξ₇Only have and rotate, not translation, so spin matrix is the transformation relation between Two coordinate system. In like manner, can calculate from coordinate system ξ₁To ξ₁₃Transformation relationAnd from coordinate system ξ₁To ξ₁₉Transformation relation

(4) shoot the image on the described face of cylinder with each video camera, extract the image coordinate of described characteristic point, by direct line Property conversion obtain the demarcation initial value of each video camera, and obtained whole inside and outside parameter of each video camera by nonlinear optimization.Fig. 6 is The characteristic point schematic diagram extracted from reference object image shown in Fig. 3, for the image 600 in Fig. 6, can use existing angle point grid Method (such as Harris C, Stephens M., A combined corner and edge detector, Proc.4th Alvey Vision Conference, Manchester, England, 1988,147-151.) extract the image of characteristic point 610 (u v), have chosen 77 characteristic points to coordinate in the embodiment shown in fig. 6.

Assume that camera lens does not distort, then the three-dimensional seat of thing characteristic point can be demarcated according to pinhole camera model Relation between mark and the image coordinate extracted:

$s\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=A_2\left[\begin{array}{cc}{R_{C_2}}_{{\mathrm{\ξ}}_7}& {T_{C_2}}_{{\mathrm{\ξ}}_7}\end{array}\right]\left[\begin{array}{c}{\mathrm{\ξ}}_{7x}\\ {\mathrm{\ξ}}_{7y}\\ {\mathrm{\ξ}}_{7z}\\ 1\end{array}\right]---\left(4\right)$

In formula, s is scale factor,For the inner parameter matrix of video camera 300, α₂,β₂For shooting The master of machine 300 is away from, (u₀₂,v₀₂) it is the principal point coordinate of video camera 300.Fig. 7 be in embodiment illustrated in fig. 1 camera coordinate system with Demarcate the schematic diagram of article coordinate system,For from video camera 300 coordinate system C₂To demarcating article coordinate system ξ₇Spin matrix, For from video camera 300 coordinate system C₂To ξ₇Translation vector.

Employing direct linear transformation's method (Ma Songde, Zhang Zhengyou, computer vision, Science Press, 1998, pp52- 59), the intrinsic parameter of video camera and outer parameter can be calculated, complete the linear calibration of video camera.At the specific embodiment shown in Fig. 6 In, the camera interior and exterior parameter drawn is as follows:

Camera intrinsic parameter:

$A_2=\left[\begin{array}{ccc}1783.8& 0& 533.6\\ 0& 1779.5& 321.8\\ 0& 0& 1\end{array}\right]$

Outer parameter:

Spin matrixTranslation vector

Owing to actual camera lens exists distortion, consider that the single order of camera lens is the most abnormal in the embodiment shown in fig. 6 Become, the coordinate (u after characteristic point distortion_d,v_d) with obtain according to pinhole camera model ideal coordinates (u, v) between relation It is represented by:

$\left[\begin{array}{c}{u}_d\\ v_d\end{array}\right]=\left[\begin{array}{c}u\\ v\end{array}\right]+k_{21}\[{(u-u_0)}^2+{(v-v_0)}^2\]\left[\begin{array}{c}u\\ v\end{array}\right]---\left(5\right)$

K in formula₂₁Single order coefficient of radial distortion for video camera 300 camera lens.

With I_dkCoordinate (the u of characteristic point after the distortion that expression is calculated by (5) formula_dk,v_dk), with I_exkRepresent and extract from image Fact characteristic point image coordinate, with distance sum between the two as object function, with above-mentioned linear calibration's result as initial value, And set k₂₁Initial value be 0, carry out nonlinear optimization according to following formula (6),

$\arg \min \underset{k=1}{\overset{num}{\Σ}}\left|\right|I_{dk}-I_{exk}\left|\right|---\left(6\right)$

Whole inside and outside parameter A of video camera 300 can be drawn₂, k₂₁,WithOptimal value, in formula, num is characterized Point number, in this embodiment num=77.In the specific embodiment shown in Fig. 6, show that the inside and outside parameter of video camera 300 is excellent Change value is as follows:

Camera intrinsic parameter:

$A_2=\left[\begin{array}{ccc}2401.0& 0& 400.1\\ 0& 2399.2& 311.3\\ 0& 0& 1\end{array}\right];k_{21}=-1.227\×{10}^{-6}$

Outer parameter:

Spin matrixTranslation vector

Whole inside and outside parameter A of video camera 200,400 and 500 can be drawn by same method₁, k₁₁, A₃, k₃₁,A₄, k₄₁,Complete the demarcation of single shooting.In the embodiment shown in fig. 1, demarcate The whole inside and outside parameter of video camera 200,400 and 500 gone out is as follows:

Camera intrinsic parameter:

$A_1=\left[\begin{array}{ccc}2295.7& 0& 399.8\\ 0& 2297.4& 304.2\\ 0& 0& 1\end{array}\right];k_{11}=-1.367\×{10}^{-6}$

$A_3=\left[\begin{array}{ccc}2702.8& 0& 399.6\\ 0& 2710.2& 299.9\\ 0& 0& 1\end{array}\right];k_{31}=-1.529\×{10}^{-6}$

$A_4=\left[\begin{array}{ccc}3481.2& 0& 400.1\\ 0& 3468.2& 310.3\\ 0& 0& 1\end{array}\right];k_{41}=-1.412\×{10}^{-6}$

Outer parameter:

${R_{C_1}}_{{\mathrm{\ξ}}_1}=\left[\begin{array}{ccc}0.0019& 0.9848& -0.1736\\ 0.0063& -0.1736& -0.9848\\ -1.0000& 0.0008& -0.0065\end{array}\right];{T_{C_1}}_{{\mathrm{\ξ}}_1}=\left[\begin{array}{c}0\\ 115.4\\ 2283.6\end{array}\right]$

${R_{C_3}}_{{\mathrm{\ξ}}_{13}}=\left[\begin{array}{ccc}-0.1694& 0.9262& -0.3369\\ -0.0011& -0.3420& -0.9397\\ -0.9855& -0.1589& 0.0589\end{array}\right];{R_{C_3}}_{{\mathrm{\ξ}}_{13}}=\left[\begin{array}{c}0\\ 120.1\\ 2707.5\end{array}\right]$

${R_{C_4}}_{{\mathrm{\ξ}}_{19}}=\left[\begin{array}{ccc}-0.0012& 0.9398& 0.3418\\ 0.1775& 0.3366& -0.9248\\ -0.9841& 0.0596& -0.1672\end{array}\right];{R_{C_4}}_{{\mathrm{\ξ}}_{19}}=\left[\begin{array}{c}0\\ 109.7\\ 3473.2\end{array}\right]$

(5) according to the relativeness of each video camera Yu the described face of cylinder, calculate the relativeness between each video camera, complete The demarcation of multi-camera system.

Fig. 7 is camera coordinate system and the schematic diagram of demarcation article coordinate system in embodiment illustrated in fig. 1, if characteristic point P_k? Camera coordinate system C₁In coordinate representation beDemarcating article coordinate system ξ₁In coordinate representation beThen characteristic point P_k? Coordinate system C₁And ξ₁In coordinate transform be represented by:

${P_{C_1}}_k={R_{C_1}}_{{\mathrm{\ξ}}_1}\·{P_{{\mathrm{\ξ}}_1}}_k+{T_{C_1}}_{{\mathrm{\ξ}}_1}---\left(7\right)$

In formulaWithObtain in step (4).In like manner can obtain camera coordinate system C_iWith demarcation article coordinate system ξ_j Transformation relation:

${P_{C_i}}_k={R_{C_i}}_{{\mathrm{\ξ}}_j}\·{P_{{\mathrm{\ξ}}_j}}_k+{T_{C_i}}_{{\mathrm{\ξ}}_j}---\left(8\right)$

In the embodiment shown in fig. 1, i=2,3,4, j=7,13,19,WithObtain in step (4).According to Step (3) gained demarcates article coordinate system ξ₁And ξ_jTransformation relation, can obtain

${P_{{\mathrm{\ξ}}_1}}_k={R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·{P_{{\mathrm{\ξ}}_j}}_k---\left(9\right)$

By (9) formula substitution (7) Shi Ke get:

${P_{C_1}}_k={R_{C_1}}_{{\mathrm{\ξ}}_1}\·{P_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·{P_{{\mathrm{\ξ}}_j}}_k+{T_{C_1}}_{{\mathrm{\ξ}}_1}---\left(10\right)$

By (8) Shi Ke get:

${P_{{\mathrm{\ξ}}_j}}_k=R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}\·{P_{C_i}}_k-R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}\·{T_{C_i}}_{{\mathrm{\ξ}}_j}---\left(11\right)$

By (11) formula substitution (10) Shi Ke get:

${P_{C_1}}_k={R_{C_1}}_{{\mathrm{\ξ}}_1}\·{R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}\·{P_{C_i}}_k+{T_{C_1}}_{{\mathrm{\ξ}}_1}-{R_{C_1}}_{{\mathrm{\ξ}}_1}\·{R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}\·{T_{C_i}}_{{\mathrm{\ξ}}_j}---\left(12\right)$

Thus obtain camera coordinate system C₁With C_iBetween transformation relation:

${P_{C_1}}_k={R_{C_1}}_{C_i}\·{P_{C_i}}_k+{T_{C_1}}_{C_i}---\left(13\right)$

In formula

${R_{C_1}}_{c_i}={R_{C_1}}_{{\mathrm{\ξ}}_1}\·{R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}---\left(14\right)$

${T_{C_1}}_{c_i}={T_{C_1}}_{{\mathrm{\ξ}}_1}-{R_{C_1}}_{{\mathrm{\ξ}}_1}\·{R_{{\mathrm{\ξ}}_1}}_{{\mathrm{\ξ}}_j}\·R_{{\mathrm{\ξ}}_j}^{-1}{}_{C_i}\·{T_{C_i}}_{{\mathrm{\ξ}}_j}---\left(15\right)$

In (14) and (15) formula, every on the right of equal sign draw in step (3) and (4), therefore can obtain each shooting Relativeness between machine, completes the demarcation of multi-camera system.

In the embodiment shown in fig. 1, the C drawn₁With C₂、C₃、C₃Between transformation relation be:

${R_{C_1}}_{c_2}=\left[\begin{array}{ccc}0.1393& 0.1728& -0.9751\\ -0.2043& 0.9685& 0.1424\\ 0.9689& 0.1794& 0.1702\end{array}\right];{T_{C_1}}_{C_2}=\left[\begin{array}{c}2321.3\\ -331.7\\ 1855.5\end{array}\right]$

${R_{C_1}}_{c_3}=\left[\begin{array}{ccc}-0.8533& 0.4999& 0.1481\\ 0.4936& 0.8661& -0.0794\\ -0.1680& 0.0054& -0.9858\end{array}\right];{T_{C_1}}_{C_3}=\left[\begin{array}{c}-461.1\\ 226.3\\ 4951.9\end{array}\right]$

${R_{C_1}}_{c_4}=\left[\begin{array}{ccc}-0.0564& -0.0137& 0.9983\\ -0.3309& 0.9437& -0.0058\\ -0.9420& -0.3307& -0.0577\end{array}\right];{T_{C_1}}_{C_4}=\left[\begin{array}{c}-3465.9\\ 31.9\\ 2520.3\end{array}\right]$

From the implementation process of technique scheme: first, demarcate the video camera of thing around the face of cylinder of the present invention Can photograph and demarcate the image of characteristic point on thing, therefore, overcome the occlusion issue existed when using two-dimensional calibrations thing, it is simple to The demarcation of wide baseline multi-camera system.Secondly, the method using the present invention, each video camera only needs to shoot described face of cylinder mark The piece image of earnest can complete the demarcation of multi-camera system, is not required to manually participate in, it is simple to whole calibration process automatic Change.3rd, for big visual field multi-camera system, for improving stated accuracy, need to make large-sized demarcation thing, if used The three-dimensional scaling thing of polyhedron type, the flatness of each side and relative position relation need large-scale precise machining equipment Can guarantee that, be not easy to the universal of application；And the metal of various diameter or plastic pipe can be easily obtained, and it is flat after cylinder face expansion Face, various plane patterns can print easily or paste on the face of cylinder, therefore, use the method phase of three-dimensional scaling thing with other Ratio, the demarcation thing of the present invention makes simple, it is simple to the demarcation of big visual field multi-camera system.It addition, the present invention demarcates on thing Characteristic point three-dimensional coordinate in demarcating article coordinate system keeps constant, therefore need not each video camera synchronous acquisition uncalibrated image, It is easy to the demarcation of asynchronous multi-camera system.

The ultimate principle of the present invention, principal character and advantages of the present invention have more than been shown and described.The technology of the industry Personnel, it should be appreciated that the present invention is not restricted to the described embodiments, simply illustrating this described in above-described embodiment and description The principle of invention, without departing from the spirit and scope of the present invention, the present invention also has various changes and modifications, and these become Change and improvement both falls within scope of the claimed invention.Claimed scope by appending claims and Equivalent defines.

Claims (1)

1. a multi-camera system scaling method based on the face of cylinder, it is characterised in that described scaling method includes walking as follows Rapid:

(1) on the face of cylinder, multiple characteristic point is constructed so that the video camera around the described face of cylinder can photograph at least 6 institutes State characteristic point, preferably can photograph at least 18 described characteristic points；

(2) axis and radius with the described face of cylinder set up three-dimensional system of coordinate for coordinate axes, and to based on different radii direction Coordinate system makes labelling；

(3) according to building method and the size on the described face of cylinder of described characteristic point, described characteristic point is calculated at described three-dimensional seat Three-dimensional coordinate in mark system, and the transformation relation between described three-dimensional system of coordinate based on different radii direction；

(4) shoot the image on the described face of cylinder with each video camera, extract the image coordinate of described characteristic point, become by the most linear Change the demarcation initial value obtaining each video camera, and obtained whole inside and outside parameter of each video camera by nonlinear optimization；

(5) according to the relativeness of described each video camera Yu the described face of cylinder, the relativeness between described each video camera is calculated, Complete the demarcation of multi-camera system.