CN100557634C - A kind of camera marking method based on double 1-dimension drone - Google Patents

Abstract

The present invention discloses a kind of camera marking method based on double 1-dimension drone, and this method may further comprise the steps: two 1-dimension drones of arbitrarily placing are set; Video camera is taken the target image from different perspectives, sets up camera coordinate system, the plane of delineation coordinate system of video camera under different camera sites respectively, and sets up world coordinate system; After respectively the target image of taking being carried out distortion correction, find the solution intrinsic parameters of the camera and external parameter.Adopt scaling method of the present invention, with two 1-dimension drones as demarcating thing, after the field of view of video camera is placed two 1-dimension drones arbitrarily, take the target image from different perspectives, only need extract three or three above unique points to every width of cloth target image can carry out camera interior and exterior parameter and demarcate, need not utility appliance, simple to operate, and can improve camera calibration precision, expansion camera calibration scope.

Description

A kind of camera marking method based on double 1-dimension drone

Technical field

The present invention relates to computer vision technique, be specifically related to a kind of camera marking method based on double 1-dimension drone.

Background technology

Camera calibration is an important and crucial problem in computer vision and the photogrammetry, so-called camera calibration just is meant the position relation of determining between camera review pixel and the corresponding fields sight spot, be specially: according to camera model, the inner parameter and the external parameter of finding the solution camera model by the image coordinate and the world coordinates of known features point.

Early stage R.Y.Tsai has proposed to demarcate based on the known three-dimensional of structure the camera marking method of thing in article " An efficient and accurate camera calibration techniquefor 3D machine vision; Proc.of IEEE Conference of Computer Vision and PatternRecognition; pp364-374; 1986 ", though adopt this scaling method can access higher stated accuracy, but the three-dimensional thing difficulty on making of demarcating is bigger, so this method is restricted in application in engineering.

Z.Zhang has proposed the camera marking method based on plane target drone in " A flexible new technique for camera calibration; IEEE Trans.onPattern Analysis and Machine Intelligence; 22 (11): pp1330-1334; 2000 ", this method is by the image of camera plane target at two above diverse locations, just can carry out camera calibration not needing to know under the situation of relative movement parameters between plane target drone and the video camera, but when adopting this scaling method to carry out camera calibration, stated accuracy depends on the precision of plane target drone to a great extent, and camera calibration generally requires target to occupy in the visual field big area of trying one's best, when measurement range is bigger, large scale, high-precision plane target drone is made relatively difficulty, and precision is difficult to ensure, thereby can causes stated accuracy to reduce.

Because one dimension demarcates thing and has simple structure, be easy to processing, do not have characteristics such as self block, and in big visual field test was used, it was more much easier than the two dimensional surface target or the three-dimensional target of an equal yardstick of processing to process a large-sized 1-dimension drone.So, become the focus that people pay close attention to gradually based on the camera marking method of 1-dimension drone.

Z.Zhang has proposed a kind of camera marking method based on 1-dimension drone in " Camera calibration with one-dimensional objects; IEEETransactions on Pattern Analysis and Machine Intelligence; 2004; VOL.26; NO.7; 892～899 ", when adopting this scaling method to carry out camera calibration, need a fixedly end of 1-dimension drone, taking any picture that rotates of several 1-dimension drones demarcates, yet this point is difficult to accurately be achieved on engineering, complex operation and can influence stated accuracy.Wu Fuchao etc. have proposed a kind of 1-dimension drone scaling method based on plane motion in " Camera calibration with moving one-dimensional objects.Pattern Recognition; 2005; 38 (5); 755～765 ", but wherein the realization of plane motion depends on the support of motion platform, and promptly adopting this scaling method to carry out camera calibration needs utility appliance.Wang Liang etc. have proposed a kind of utilization and have done the scaling method that the 1-dimension drone of any rigid motion carries out camera calibration in " multiple-camera of demarcating thing based on one dimension is demarcated; the robotization journal; the 33rd volume; the 3rd phase; 2007; 225～231 ", but this scaling method only is applicable to the demarcation multiple-camera, and can not demarcate single camera.

As can be seen, in the prior art, though in the camera calibration of big visual field test, have many advantages based on the camera marking method of 1-dimension drone, but still have that stated accuracy is lower, complex operation, need deficiencies such as utility appliance, calibration range be limited.

Summary of the invention

In view of this, fundamental purpose of the present invention is to provide a kind of camera marking method based on double 1-dimension drone, and can improve stated accuracy, enlarge calibration range, need not utility appliance, and simple to operate.

For achieving the above object, technical scheme of the present invention is achieved in that

A kind of camera marking method based on double 1-dimension drone, this method comprises:

A, two 1-dimension drones of arbitrarily placing are set;

B, video camera are taken the target image from different perspectives, set up camera coordinate system, the plane of delineation coordinate system of video camera under different camera sites respectively, and set up world coordinate system;

C, respectively the target image of taking is carried out distortion correction after, find the solution intrinsic parameters of the camera and external parameter.

After the step c, this method further comprises: intrinsic parameters of the camera and external parameter are carried out the optimum solution that nonlinear optimization obtains intrinsic parameters of the camera and external parameter.

The target image of the described shooting of step b comprises two 1-dimension drones at least, and each 1-dimension drone comprises the unique point of three conllinear at least.

The described world coordinates of setting up of step b is: with the camera coordinate system of first camera site of video camera as world coordinate system.

The described intrinsic parameters of the camera of finding the solution of step c is: the coordinate of asking for the shadow point that disappears of latter two 1-dimension drone unique point place straight line of every width of cloth target image distortion correction respectively; Spatially angle is constant according to two 1-dimension drones, finds the solution intrinsic parameters of the camera in conjunction with the coordinate of the resulting shadow point that disappears.

The described video camera external parameter of finding the solution of step c is:

C1, an optional width of cloth target image, utilize behind the coordinate, distortion correction of the shadow point that disappears of 1-dimension drone unique point place straight line the space length between two-end-point characteristic point coordinates in the target image and two-end-point unique point, try to achieve the coordinate of 1-dimension drone two-end-point unique point under camera coordinate system;

C2, obtain the coordinate of two 1-dimension drone end points unique points under world coordinate system;

C3, try to achieve video camera when taking the described target image of step c1 under the present position according to coordinate united law of the same name, world coordinates is tied to the transformation matrix of camera coordinate system, as the external parameter of video camera when this position.

Describedly the video camera internal and external parameter is carried out nonlinear optimization be: set up with intrinsic parameters of the camera, external parameter and two objective functions that the 1-dimension drone relative position is a parameter; The intrinsic parameters of the camera and the external parameter of trying to achieve with step c are initial value, optimize the optimum solution that obtains intrinsic parameters of the camera and external parameter.

Camera marking method based on double 1-dimension drone provided by the invention, with two 1-dimension drones as demarcating thing, field of view at video camera is placed two 1-dimension drones arbitrarily, take the target image from different perspectives, only need extract three or three above unique points to every width of cloth target image just can carry out camera interior and exterior parameter and demarcate, need not utility appliance, simple to operate, and improved stated accuracy, enlarged calibration range.

Description of drawings

Fig. 1 is the camera marking method process flow diagram that the present invention is based on double 1-dimension drone;

Fig. 2 is the 1-dimension drone structural representation;

Fig. 3 is a calibration principle synoptic diagram of the present invention;

Fig. 4, Fig. 5, Fig. 6 are respectively three width of cloth target images that video camera is taken.

Embodiment

Basic thought of the present invention is: with two 1-dimension drones as demarcating thing, field of view at video camera is placed two 1-dimension drones arbitrarily, take the target image from different perspectives, only need extract three or three above unique points to every width of cloth target image and carry out camera interior and exterior parameter and demarcate.

Below in conjunction with specific embodiment and accompanying drawing the present invention is described in further detail.

Fig. 1 is the camera marking method process flow diagram that the present invention is based on double 1-dimension drone, and as shown in Figure 1, the camera marking method that the present invention is based on double 1-dimension drone may further comprise the steps:

Step 11: two 1-dimension drones of arbitrarily placing are set.

Here, described 1-dimension drone is the linear pattern target of monumented point conllinear, and Fig. 2 is the 1-dimension drone structural representation, as shown in Figure 2, it is provided with 3～10 round dot marks, and the round dot diameter is 3～20mm, precision is 0.001～0.01mm, and the round dot center is a unique point, each unique point conllinear.

Step 12: video camera is taken the target image from different perspectives, sets up camera coordinate system, the plane of delineation coordinate system of video camera under different camera sites respectively, and sets up world coordinate system.

Here, the target image that video camera is taken need comprise two 1-dimension drones, and each 1-dimension drone comprises the unique point of three conllinear at least.

Fig. 3 is a calibration principle synoptic diagram of the present invention, as shown in Figure 3, sets up camera coordinate system O according to the putting position of video camera _c-x _cy _cz _cAnd the plane of delineation coordinate system O-UV of video camera.

When setting up world coordinate system, can be world coordinate system, be without loss of generality that the present invention gets the camera coordinate system of video camera under first camera site as world coordinate system with the coordinate of video camera under arbitrary camera site.

Step 13: respectively the target image of taking is carried out distortion correction, and ask for behind the distortion correction coordinate of the shadow point that disappears of two 1-dimension drone unique point place straight lines in every width of cloth target image.

When the target image is carried out distortion correction, need to extract the coordinate of 1-dimension drone unique point under image coordinate system in every width of cloth target image, in the present embodiment, the unique point on the target is a round dot.According to perspective projection transformation, circle video camera as the plane on imaging be approximately oval.So, determine elliptical center by the method for marginal point being carried out ellipse fitting in the present embodiment, can extract the unique point on the target image.Concrete extracting method has a detailed description in doctor's Wei Zhenzhong paper " based on the online flexible three-dimensional coordinates measurement systematic study of machine vision, the doctoral candidate of BJ University of Aeronautics ﹠ Astronautics academic dissertation ", repeats no more here.

It is prior art that the target image is carried out distortion correction, and the distortion model of video camera is as follows:

$\left\{\begin{array}{c}u=(u_u-u_0)(1+k_1{r_u}^2+k_2{r_u}^4)\\ v=(v_u-v_0)(1+k_1{r_u}^2+k_2{r_u}^4)\\ {r_u}^2=({u_u}^2+{v_u}^2)\end{array}\right.---\left(1\right)$

Wherein, (u v) is (u _u, v _u) through the real image coordinate after the distortion, (u ₀, v ₀) be video camera principal point coordinate, k ₁, k ₂Be the video camera coefficient of radial distortion.When carrying out distortion correction, can set up objective function earlier, find the solution optimum distortion factor by nonlinear optimization then based on collinear feature point linearity in the image.Concrete distortion correction method has a detailed description in the article " Nonmetric Calibration of Camera LensDistortion:Differential Methods and Robust Estimation; IEEE Transactions onPattern Analysis and Machine Intelligence; 2005; VOL.14; NO.8,1215～1230 " of Moumen Ahmed.

Utilize mutual distance on the 1-dimension drone known three or three above unique points can be in the hope of the coordinate of the shadow point that disappears of space line in image, the method that is provided in " R.Harley; A.Zisserman; A Multiple View Geometry in Computer Vision.Cambridge:CambridgeUniversity Press, 2000 " that concrete acquiring method can adopt R.Harley to show.

Step 14: constant at space angle according to two 1-dimension drones, the coordinate of the integrating step 13 resulting shadow points that disappear is found the solution intrinsic parameters of the camera.

The intrinsic parameters of the camera matrix can be expressed as:

$K=\left[\begin{array}{ccc}{f}_x& \mathrm{\α}& u_0\\ 0& f_y& v_0\\ 0& 0& 1\end{array}\right]---\left(2\right)$

Wherein, f _x, f _yBe respectively the scale factor of plane of delineation coordinate system U axle and V axle, α is the out of plumb factor of U axle and V axle.

Because f _x≈ f _y, α ≈ 0, u ₀≈ 0.5N _u, v ₀≈ 0.5N _v, N wherein _u, N _vBe respectively image at U axle and the axial pixel count of V.If f _x≈ f _y=f moves to principal point with coordinate origin, and the inner parameter matrix can be reduced to:

$\tilde{K}=\left[\begin{array}{ccc}f& 0& 0\\ 0& f& 0\\ 0& 0& 1\end{array}\right]---\left(3\right)$

Again because, if disappear shadow point the homogeneous coordinates under image coordinate system of two 1-dimension drones in image are respectively v ₁=(x ₁, y ₁, 1) ^TAnd v ₂=(x ₂, y ₂, 1) ^T, two the 1-dimension drone unique point place angles of straight line in the space are θ, then have:

$\cos \mathrm{\θ}=\frac{v_1^T\mathrm{\ω}{v}_2}{\sqrt{\left(v_1^T\mathrm{\ω}{v}_1\right)\left(v_2^T\mathrm{\ω}{v}_2\right)}}---\left(4\right)$

Wherein, ω=(KK ^T) ^-1

After coordinate origin moved to principal point, formula (4) became:

$\cos \mathrm{\θ}=\frac{{\tilde{v}}_1^T\widetilde{\mathrm{\ω}}{\tilde{v}}_2}{\sqrt{\left({\tilde{v}}_1^T\widetilde{\mathrm{\ω}}{\tilde{v}}_1\right)\left({\tilde{v}}_2^T\widetilde{\mathrm{\ω}}{\tilde{v}}_2\right)}}---\left(5\right)$

Wherein, $\widetilde{\mathrm{\ω}}={\left(\tilde{K}{\tilde{K}}^T\right)}^{-1},$ ${\tilde{v}}_1={(x_1-u_0,y_1-v_0,1)}^T,$ ${\tilde{v}}_2={(x_2-u_0,y_2-v_0,1)}^T$ Be respectively the homogeneous coordinates of shadow point under coordinate system after the translation that disappear of two 1-dimension drones.

If in m width of cloth image and the n width of cloth image, the shadow point that disappears of two 1-dimension drones is respectively

With

Then have in that space angle is constant according to two 1-dimension drones:

$\frac{{\left({\tilde{v}}_{m,1}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{m,2}\right)}^2}{\left({\tilde{v}}_{m,1}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{m,1}\right)\left({\tilde{v}}_{m,2}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{m,2}\right)}=\frac{{\left({\tilde{v}}_{n,1}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{n,2}\right)}^2}{\left({\tilde{v}}_{n,1}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{n,1}\right)\left({\tilde{v}}_{n,2}^T\widetilde{\mathrm{\ω}}{\tilde{v}}_{n,2}\right)}---\left(6\right)$

According to formula (3) and $\widetilde{\mathrm{\ω}}={\left(\tilde{K}{\tilde{K}}^T\right)}^{-1},$ Formula (6) can be written as about

Simple cubic equation, separate this equation, get positive root, can try to achieve f.

Then, set up about all parameter f in the inner parameter matrix K _x, f _y, α, u ₀, v ₀Objective function as follows:

$\underset{m=1}{\overset{M}{\mathrm{\Σ}}}{\left|\right|{\cos \mathrm{\θ}}_m-\mathrm{\μ}\left(\cos \mathrm{\θ}\right)\left|\right|}^2=\min ---\left(7\right)$

Wherein, ${\cos \mathrm{\θ}}_m=\frac{v_{m,1}^T\mathrm{\ω}{v}_{m,2}}{\sqrt{\left(v_{m,1}^T\mathrm{\ω}{v}_{m,1}\right)\left(v_{m,2}^T\mathrm{\ω}{v}_{m,2}\right)}},$ $\mathrm{\μ}\left(\cos \mathrm{\θ}\right)=\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\cos {\mathrm{\θ}}_i,m=\mathrm{1,2},...,M,$ The target figure film size number of M for taking.

Again with f _x=f, f _y=f, α=0, u ₀=0.5N _u, v ₀=0.5N _vAs initial value, adopt the Levenberg-Marquardt nonlinear optimization algorithm to be optimized to five variablees, just can obtain the intrinsic parameters of the camera matrix K.

Step 15:, find the solution the video camera external parameter according to the space length between two-end-point characteristic point coordinates and two-end-point unique point in the target image behind the coordinate of the described shadow point that disappears of step 13, the distortion correction.

For any one 1-dimension drone in any width of cloth target image, all have system of equations (8) to set up:

$\left\{\begin{array}{c}{s}_1{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">p</figure-callout>}}_1={\mathrm{KP}}_1\\ s_2{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">p</figure-callout>}}_2={\mathrm{<figure-callout\; id="2"\; label="KP"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">KP</figure-callout>}}_2\\ \left|\right|P_1-{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">P</figure-callout>}}_2\left|\right|=L\\ (p_1-p_2)\&times;v=0\end{array}\right.---\left(8\right)$

Wherein, P ₁=(x _C1, y _C1, z _C1, 1) ^T, P ₂=(x _C2, y _C2, z _C2, 1) ^TFor 1-dimension drone two-end-point unique point at camera coordinate system O _c-x _cy _cz _cUnder three-dimensional coordinate, p ₁=(u ₁, v ₁, 1) ^T, p ₂=(u ₂, v ₂, 1) ^TBe the image coordinate of two end points unique points of 1-dimension drone, v=(x, y, 1) ^TBe the coordinate of the shadow point that disappears of 1-dimension drone unique point place straight line, L is the space length between 1-dimension drone two-end-point unique point, s ₁, s ₂Be and be not equal to zero constant.

Here, an optional width of cloth target image, utilize behind the coordinate, distortion correction of the shadow point that disappears of 1-dimension drone unique point place straight line the space length between two-end-point characteristic point coordinates in the target image and two-end-point unique point, by solving equation group (8), obtain 1-dimension drone two-end-point unique point at camera coordinate system O _c-x _cy _cz _cUnder three-dimensional coordinate P ₁And P ₂

Try to achieve two all unique points of target according to system of equations (8) and be in coordinate under the local measurement coordinate system of diverse location at video camera.Under the present position, world coordinates is tied to the transformation matrix of camera coordinate system, i.e. the external parameter of video camera when this position when trying to achieve video camera shooting target image according to coordinate united law of the same name then." machine vision " that concrete acquiring method is write at Zhang Guangjun (ISBN:7-03-014717-0) the 186th page to the 187th page have a detailed description.

Step 16: intrinsic parameters of the camera and external parameter are carried out nonlinear optimization, obtain the optimum solution of intrinsic parameters of the camera and external parameter.

Here, be initial value with the intrinsic parameters of the camera of step 14 gained and the video camera external parameter of step 15 gained, be that objective function carries out nonlinear optimization to each parameter with the quadratic sum of the projection error of all unique points on the plane of delineation on two targets.

At first, set up with intrinsic parameters of the camera, external parameter and two objective functions that the 1-dimension drone relative position is a parameter:

Wherein, m represents m width of cloth image, and r represents r target, and j represents j unique point on the target,

Be the image coordinate of obtaining by camera model, x _{M, r, j}Be the real image coordinate.P _{1, r, 1}Represent the position of first end points under world coordinate system on r the target, θ _rBe r the target angle that the projection in the xy plane is become with the x axle under world coordinate system,

Be r the angle that target is become with the xy plane under world coordinate system.

Then, the camera interior and exterior parameter that step 14 and step 15 are tried to achieve utilizes the Levenberg-Marquardt optimization method to find the solution this nonlinear optimal problem as initial value, can obtain the optimum solution of intrinsic parameters of the camera and external parameter.

If the video camera of being demarcated is a Cannon EOS-5D digital camera, camera lens adopts 50mm F/4 camera lens, and cmos sensor resolution is 4368 * 2912.Operating distance is 200mm, and measurement range is 200mm * 160mm.Get six unique points on two 1-dimension drones respectively, unique point is spaced apart 12.728mm.Video camera has been taken totally three width of cloth target images from different perspectives respectively as Fig. 4, Fig. 5, shown in Figure 6, according to the described method of step 13, obtains that the unique point coordinate data sees Table 1 in the target image:

Table 1

The 13 described methods data of utilizing target image shown in Figure 4 to obtain are tried to achieve distortion factor k set by step ₁=0.55, k ₂=-3.93.

It is as follows that the described method of 13～step 14 is tried to achieve the intrinsic parameters of the camera matrix set by step:

$K=\left[\begin{array}{ccc}6131.02& 0& 2187.86\\ 0& 6111.01& 1469.64\\ 0& 0& 1\end{array}\right]$

It is as follows that 15 described methods are tried to achieve the video camera external parameter of each width of cloth image correspondence set by step:

The video camera external parameter of target image correspondence shown in Figure 4 is:

${\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">R</figure-callout>}}_1=\left[\begin{array}{ccc}0.638& 0.684& 0.351\\ 0.684& -0.714& 0.146\\ 0.351& 0.147& -0.924\end{array}\right],$ ${\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">T</figure-callout>}}_1=\left[\begin{array}{c}135.35\\ -81.36\\ 625.81\end{array}\right]$

The video camera external parameter of target image correspondence shown in Figure 5 is:

${\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">R</figure-callout>}}_2=\left[\begin{array}{ccc}0.723& 0.669& -0.173\\ 0.688& -0.721& 0.082\\ -0.070& -0.179& -0.981\end{array}\right],$ ${\mathrm{<figure-callout\; id="2"\; label="T"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">T</figure-callout>}}_2=\left[\begin{array}{c}121.67\\ -107.93\\ 654.70\end{array}\right]$

The video camera external parameter of target image correspondence shown in Figure 6 is:

${\mathrm{<figure-callout\; id="3"\; label="R"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">R</figure-callout>}}_3=\left[\begin{array}{ccc}0.590& 0.791& -0.161\\ 0.686& -0.596& -0.417\\ -0.426& 0.136& -0.895\end{array}\right],$ ${\mathrm{<figure-callout\; id="3"\; label="T"\; filenames="C20081010297800152.png,C20081010297800161.png"\; state="\{\{state\}\}">T</figure-callout>}}_3=\left[\begin{array}{c}131.32\\ -84.15\\ 672.25\end{array}\right]$

At last, 16 described methods are optimized the inside and outside parameter of video camera set by step, obtain optimum solution and see Table 2:

Table 2

Calculated characteristics point is 0.263 pixel by the root-mean-square error of camera model projection gained image coordinate and real image coordinate.

The above is preferred embodiment of the present invention only, is not to be used to limit protection scope of the present invention.

Claims (4)

1, a kind of camera marking method based on double 1-dimension drone is characterized in that, this method comprises:

A, two 1-dimension drones of arbitrarily placing are set;

B, video camera are taken the target image from different perspectives, set up camera coordinate system and the plane of delineation coordinate system of video camera under different camera sites respectively, and set up world coordinate system;

C, respectively the target image of taking is carried out distortion correction after, find the solution intrinsic parameters of the camera and external parameter;

The target image of described shooting comprises two 1-dimension drones, and each 1-dimension drone comprises the unique point of three conllinear at least;

The described detailed process of finding the solution intrinsic parameters of the camera is: the coordinate of asking for the shadow point that disappears of latter two 1-dimension drone unique point place straight line of every width of cloth target image distortion correction respectively; Spatially angle is constant according to two 1-dimension drones, finds the solution intrinsic parameters of the camera in conjunction with the coordinate of the resulting shadow point that disappears;

The described detailed process of finding the solution the video camera external parameter is: an optional width of cloth target image, utilize behind the coordinate, distortion correction of the shadow point that disappears of 1-dimension drone unique point place straight line the space length between two-end-point characteristic point coordinates in the target image and two-end-point unique point, try to achieve the coordinate of 1-dimension drone two-end-point unique point under camera coordinate system; Obtain the coordinate of two 1-dimension drone end points unique points under world coordinate system; Try to achieve video camera when taking described optional target image under the present position according to coordinate united law of the same name, world coordinates is tied to the transformation matrix of camera coordinate system, as the external parameter of video camera when this position;

Described 1-dimension drone i.e. the linear pattern target of three or three above unique point conllinear.

2, method according to claim 1 is characterized in that, after the step c, this method further comprises: intrinsic parameters of the camera and external parameter are carried out the optimum solution that nonlinear optimization obtains intrinsic parameters of the camera and external parameter.

3, method according to claim 1 is characterized in that, the described world coordinates of setting up of step b is: with the camera coordinate system of first camera site of video camera as world coordinate system.

4, method according to claim 2 is characterized in that, describedly the video camera internal and external parameter is carried out nonlinear optimization is: set up with intrinsic parameters of the camera, external parameter and two objective functions that the 1-dimension drone relative position is a parameter; The intrinsic parameters of the camera and the external parameter of trying to achieve with step c are initial value, optimize the optimum solution that obtains intrinsic parameters of the camera and external parameter;

Described objective function is:

Wherein, m represents m width of cloth image, and the span of m is: m=1, and 2 ..., M, the target figure film size number of M for taking, M gets the positive integer more than or equal to 2; R represents r target, and the span of r is: r=1,2; J represents j unique point on the target, and the span of j is: j=1, and 2 ..., J, J are that the feature on the target is counted, J gets the positive integer more than or equal to 3;

Be the image coordinate of obtaining by camera model, x _{M, r, j}Be real image coordinate, P _{1, r, l}Represent the position of first end points under world coordinate system on r the target, θ _rBe r the target angle that the projection in the xy plane is become with the x axle under world coordinate system,

Be r the angle that target is become with the xy plane under world coordinate system, k ₁, k ₂The expression distortion factor, K represents the intrinsic parameters of the camera matrix of trying to achieve, T _m, R _mThe video camera external parameter of representing m width of cloth target image correspondence.

Images