CN105806318A - Visual measurement method for space three-dimensional information based on motion time quantity - Google Patents

Abstract

The invention relates to a visual measurement method for space three-dimensional information based on motion time quantity and belongs to the technical field of computer visual measurement. The visual measurement method comprises the following steps: firstly, fixing a camera on a high-precision linear guide rail, wherein the guide rail is utilized to ensure the high-precision uniform linear motion of the camera; solving a space three-dimensional coordinate by utilizing a sequence image feature in a motion process, thereby completing the measurement for the space three-dimensional information of a to-be-detected object. The first step is adopted for calibrating the camera, the second step is adopted for extracting a marking point feature and the third step is adopted for solving the three-dimensional information. According to the visual measurement method provided by the invention, the high-precision uniform linear motion of an electric control platform is utilized to convert the geometrical measurement for the measured space into the measurement for the time quantity of the speed, the feature in the image can be subjected to reconstruction of three-dimensional information, the measurement cost is lowered, the measurement efficiency and precision are increased and the quick measurement in full-visual field of a monocular camera can be realized.

Description

Space three-dimensional information vision measuring method based on movement time amount

Technical field

The invention belongs to computer vision measurement technical field, relate to a kind of space three-dimensional information vision measuring method based on movement time amount.

Background technology

Vision measurement technology suffers from extremely wide application in every field such as Aeronautics and Astronautics, military project, marine navigations.Vision measurement has the advantages such as noncontact, environmental suitability be strong, the pictorial information shot by camera gets final product the three-dimensional information of quick measurement space feature, and present vision measuring method is mostly only applied the image information of camera or the aggregate information in space and space three-dimensional information is measured, which has limited the precision of measurement and efficiency.

The patent of invention CN105184857A of Li Xiuzhi of Beijing University of Technology et al. application, " monocular vision based on structure light range finding rebuilds mesoscale factor determination method ", the method is from structured light, propose a kind of with structure light for auxiliary the European three-dimensional rebuilding method of monocular vision, including facula mass center centralized positioning, add RANSAC rejection space line matching, ask for space three-dimensional point coordinates.But, the method needs repeatedly to adjust the position of chessboard target repeatedly in measurement process, demarcates numerous and diverse and needs repeatedly to demarcate, and requiring higher to the locus of laser instrument and camera, and stability is difficult to ensure that.The patent of invention CN104864851A of Zhao Rujin of Photoelectric Technology Inst., Chinese Academy of Sciences et al. application, " a kind of monocular vision pose measuring method based on rectangular perimeter and area Weighted Constraint " the method constructs rectangle on measured object surface by 4 characteristic points, weighted error restriction relation, recycling rectangular area and girth error constraints is established to target measurement according to the geometric properties that rectangular target girth and area remain unchanged.Measured object surface configuration and size be there are certain requirements by this measuring method, therefore measure and are limited in scope；Secondly, in the process of structural feature rectangle, the geometric accuracy of rectangle is difficult to ensure that, measurement result can be produced large effect by rectangular shape scale error, causes that measurement result is more unstable.

Summary of the invention

The present invention not only improves the efficiency of vision measurement, and geometric measurement in vision measurement process is converted into time quantum and measures, improve the precision measured, invent a kind of space three-dimensional information vision measuring method based on movement time amount, simplify measurement procedure and improve measuring speed and precision.By traditional geological information vision measuring method based on image information or space, refer to kinematic parameter, improve measurement efficiency and reduce cost, it is achieved that monocular camera is for the quick measurement in full filed.

The technical solution adopted in the present invention is a kind of space three-dimensional information vision measuring method based on movement time amount, camera is first fixed on high accuracy line slideway by measuring method, guide rail is utilized to ensure that the high accuracy linear uniform motion of camera, then utilize sequential image feature in motor process that 3 d space coordinate is solved, complete the measurement of testee space three-dimensional information, specifically comprising the following steps that of measuring method

The first step carries out the demarcation of camera

Measuring method adopts the relatively-stationary mode of single camera, with the method that video camera is demarcated by precision machined target plate；Based on Zhang Zhengyou et al. camera marking method based on 2 dimensional plane targets proposed, and for the high-precision measurement requirement of relative distance in measurement process, it is achieved the high-precision calibrating of camera, camera imaging machine model is as follows:

$Z_c\left[\begin{array}{c}\hat{u}\\ \hat{v}\\ 1\end{array}\right]+Z_c\left[\begin{array}{ccc}{k}_1\hat{u}& k_2\hat{u}& 0\\ k_1\hat{v}& k_2\hat{v}& 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}{\hat{u}}^2+{\hat{v}}^2\\ {({\hat{u}}^2+{\hat{v}}^2)}^2\\ 1\end{array}\right]=M\[\begin{array}{cc}R& T\end{array}\]\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]---\left(1\right)$

Wherein, Z_cFor scale factor,For control point actual pixels coordinate in camera plane, k₁,k₂For distortion factor, M is camera Intrinsic Matrix, R, and T respectively camera coordinate system is relative to the rotation of world coordinate system and translation matrix, (X_w,Y_w,Z_w) for control point coordinate under world coordinate system.

The extraction of second step labelling point feature

First at electric control platform initial position, control camera by graphics workstation to shoot, obtain the first two field picture A1, then control electric control platform to advance 1s along y-axis with certain speed, continue shooting, obtain the second two field picture A2, finally still control electric control platform and advance 1s along y-axis with certain speed, it is thus achieved that the 3rd two field picture A3；Gradient centroid method is adopted to extract the feature on the first frame, the second frame, the 3rd two field picture A1, A2, A3 image respectively；Then, adopting Gauss first order differential operator that image is carried out convolution operation, to obtain the image gradient at each point, then the acquiring method for the gradient focus point coordinate of pattern place subregion is as follows:

$C=\underset{i=-h}{\overset{h}{\Σ}}\underset{j=-w}{\overset{w}{\Σ}}\[\left|G(i,j)\right|\·P(i,j)\]/\underset{i=-h}{\overset{h}{\Σ}}\underset{j=-w}{\overset{w}{\Σ}}\left|G(i,j)\right|---\left(2\right)$

Wherein: C is the labelling dot center pixel coordinate value that application gradient centroid method extracts, | and G (i, j) | for (i, j) gradient magnitude put, w, the width of h respectively target image and height, (i j) is (i, j) image coordinate put to P.

3rd step three-dimensional information solves

The movement velocity of known guide is v, the movement time of taken image is spaced apart Δ t, respectively at time t=0, t=Δ t, inscribe during t=2 Δ t tri-and carry out image procossing, obtain three, the space coordinate of characteristic point, first according to national forest park in Xiaokeng, the actual coordinate of spatial point and the relation of pixel planes coordinate can be obtained:

$x=\frac{zX}{f},y=\frac{zY}{f}---\left(3\right)$

Wherein, (x, y, z) be the actual coordinate of spatial point, and (X, Y) is characteristic point physical coordinates on picture.Below equation can be obtained according to formula (3) and spatial relation:

$\mathrm{\Δ}l=\sqrt{\frac{{(z_1{X}_1-z_2{X}_2)}^2+{(z_{1}Y-z_2{Y}_2)}^2}{f^2}+{(z_1-z_2)}^2}---\left(4\right)$

$\mathrm{\Δ}l=\sqrt{\frac{{(z_2{X}_2-z_3{X}_3)}^2+{(z_2{Y}_2-z_3{Y}_3)}^2}{f^2}+{(z_2-z_3)}^2}---\left(5\right)$

$\frac{z_1-z_2}{z_1{X}_1-z_2{X}_2}=\frac{z_2-z_3}{z_2{X}_2-z_3{X}_3}---\left(6\right)$

Wherein, v is the movement velocity of guide rail, and Δ t is movement time interval, and Δ l is the move distance of camera, (X in the Δ t time_i,Y_i) it is the pixel planes coordinate of the i-th frame spatial point, (x_i,y_i,z_i) for being separated by the actual coordinate of i × Δ t spatial point.(i=1,2,3), solve (x by simultaneous formula (4), (5), (6)₁,y₁,z₁), it is object point real space coordinate.

The invention has the beneficial effects as follows that the method only needs single camera collocation high accuracy straight line electric control platform, do not need auxiliary laser and accessory projector, also without the priori in known image, utilize the high accuracy linear uniform motion of electric control platform, the geometric measurement of measurement space is converted to the measurement of the time quantum of speed, the reconstruction of three-dimensional information can be carried out for the feature in image, reduce measurement cost, add measurement efficiency and precision, it is achieved that monocular camera is for the quick measurement in full filed.

Accompanying drawing explanation

Fig. 1 show the space three-dimensional information vision measuring method mounted cast figure based on movement time amount.1-high precision electric control platform, 2-camera, 3-testee.

Fig. 2 is based on the flow chart of the space three-dimensional information vision measuring method of movement time amount.

Detailed description of the invention

The specific embodiment of the present invention is described in detail below in conjunction with technical scheme and accompanying drawing.

Accompanying drawing 1 is the space three-dimensional information vision measuring method mounted cast figure based on movement time amount.This device drives camera by high-precision electric control platform, in identical interval, completes linear uniform motion, three photos of the identical time shooting in interval on straight-line trajectory, thus carrying out the accurate measurement of space object three-dimensional coordinate.

The mounting means of its device is as follows: camera is installed on high precision electric control platform.The present invention adopts high accuracy straight line electric control platform.Adopting the camera subject volume image with camera lens, camera model is FASTCAMUX50 video camera, resolution: 1248*1024, frame frequency: ensure in resolution 1248*1024 situation at most up to 2000fps.Camera lens model is that AF-S17-35mmf/2.8DIF-ED parameter is as follows, lens focus: f=17-35, APS focal length: 25.5-52.5, maximum ring: F2.8, minimum aperture: F22, camera lens weight: 745g, Lens: 82.5 × 106.Shooting condition is as follows: picture pixels is 1248 × 1024, and visual field is about 600mm × 600mm, and shooting speed is 100fps.

Accompanying drawing 2 is based on the flow chart of the space three-dimensional information vision measuring method of movement time amount, according to flow chart embodiment to be embodied as step as follows:

The first step carries out the demarcation of camera

The present invention adopts the relatively-stationary mode of video camera, with the method that video camera is demarcated by precision machined target plate.The method is based on the camera marking method based on 2D plane target drone that Zhang Zhengyou et al. proposes, and for the high-precision measurement requirement of relative distance in measurement process, using the reconstruction precision of monocular system space point as object function.And adopt Longguet-Higgins 8 algorithms of normalization proposed to calculate fundamental matrix, utilize formula (1) that the intrinsic parameter K of high speed camera, outer parameter [RT], distortion factor k and fundamental matrix R, t can be obtained；Focal length calibration result is f=1294.21658.

Adopt Zhang Shi scaling method accurate can ask for the inside and outside parameter of camera, add the reconstruction precision using monocular system space point, as object function, inside and outside parameter is carried out global optimization, improve the credibility of flag parameters under larger field further.

The extraction of second step labelling point feature

First at electric control platform initial position, control camera by graphics workstation to shoot, obtain the first two field picture A1, then control electric control platform to advance 1s along y-axis with speed v=20.209mm/s, continue shooting, obtain the second two field picture A2, finally still control electric control platform and advance 1s along y-axis with v=20.209mm/s speed, it is thus achieved that the 3rd two field picture A3.

Gradient centroid method is adopted to extract the pixel coordinate value of the upper labelling point of image A1, A2 that camera shoots at 3 diverse locations, A3, respectively [X respectively₁Y₁]=[469.58219291.43307], [X₂Y₂]=[440.40808292.30085], [X₃Y₃]=[410.74325292.68879].

Solving of 3rd step three-dimensional information

The movement velocity of known guide is v=20.209mm/s, and the movement time of taken image is spaced apart Δ t=1s, inscribes when time t=0, t=Δ t, t=2 Δ t tri-respectively and carries out image procossing, it is thus achieved that three, the space of characteristic point coordinate.Further according to national forest park in Xiaokeng, the actual coordinate of spatial point and the relation of pixel planes coordinate can be obtained, utilize the movement velocity of electric control platform and time to try to achieve: Δ l=v × Δ t=20.209mm/s × 1s=20.209mm

Then bring formula formula (3) into, (4), (5), (6) simultaneous solution go out three groups of space coordinatess, and result is as follows:

[x₁y₁z₁]=[-112.056153-162.0950091150.600906]

[x₂y₂z₂]=[-130.899183-160.7039101143.197584]

[x₃y₃z₃]=[-149.838016-159.3166371136.007787]

[-112.056153-162.0950091150.600906] is spatial point actual coordinate.

The method utilizes the high accuracy linear uniform motion of electric control platform, the geometric measurement of measurement space is converted to the measurement of the time quantum of speed, the reconstruction of three-dimensional information can be carried out for the feature in image, reduce measurement cost, add measurement efficiency and precision, it is achieved that monocular camera is for the quick measurement in full filed.

Claims (1)

1. based on a space three-dimensional information vision measuring method for movement time amount, it is characterized in that, camera is first fixed on high accuracy line slideway by measuring method, utilizes guide rail to ensure that the high accuracy linear uniform motion of camera；Then, utilize sequential image feature in motor process that 3 d space coordinate is solved, complete the measurement of testee space three-dimensional information；Specifically comprising the following steps that of measuring method

The first step carries out the demarcation of camera

Measuring method adopts the relatively-stationary mode of single camera, with the method that video camera is demarcated by precision machined target plate；Based on Zhang Zhengyou et al. camera marking method based on 2 dimensional plane targets proposed, and for the high-precision measurement requirement of relative distance in measurement process, it is achieved the high-precision calibrating of camera, camera imaging machine model is as follows:

$Z_c\left[\begin{array}{c}\hat{u}\\ \hat{v}\\ 1\end{array}\right]+Z_c\left[\begin{array}{ccc}{k}_1\hat{u}& k_2\hat{u}& 0\\ k_1\hat{v}& k_2\hat{v}& 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}{\hat{u}}^2+{\hat{v}}^2\\ {({\hat{u}}^2+{\hat{v}}^2)}^2\\ 1\end{array}\right]=M\[\begin{array}{cc}R& T\end{array}\]\left[\begin{array}{c}{X}_w\\ Y_w\\ Z_w\\ 1\end{array}\right]---\left(1\right)$

Wherein, Z_cFor scale factor,For control point actual pixels coordinate in camera plane, k₁,k₂For distortion factor, M is camera Intrinsic Matrix, R, and T respectively camera coordinate system is relative to the rotation of world coordinate system and translation matrix, (X_w,Y_w,Z_w) for control point coordinate under world coordinate system；

The extraction of second step labelling point feature

First at electric control platform initial position, control camera by graphics workstation to shoot, obtain the first two field picture A1, then control electric control platform to advance 1s along y-axis with certain speed, continue shooting, obtain the second two field picture A2, finally still control electric control platform and advance 1s along y-axis with certain speed, it is thus achieved that the 3rd two field picture A3；Gradient centroid method is adopted to extract the feature on the first frame, the second frame, the 3rd two field picture A1, A2, A3 image respectively；Then adopting Gauss first order differential operator that image is carried out convolution operation, to obtain the image gradient at each point, then the acquiring method for the gradient focus point coordinate of pattern place subregion is as follows:

$C=\underset{i=-h}{\overset{h}{\Σ}}\underset{j=-w}{\overset{w}{\Σ}}\[\left|G(i,j)\right|\·P(i,j)\]/\underset{i=-h}{\overset{h}{\Σ}}\underset{j=-w}{\overset{w}{\Σ}}\left|G(i,j)\right|---\left(2\right)$

Wherein: C is the labelling dot center pixel coordinate value that application gradient centroid method extracts, | and G (i, j) | for (i, j) gradient magnitude put, w, the width of h respectively target image and height, (i j) is (i, j) image coordinate put to P；

3rd step three-dimensional information solves

The movement velocity of known guide is v, the movement time of taken image is spaced apart Δ t, respectively at time t=0, t=Δ t, inscribe during t=2 Δ t tri-and carry out image procossing, obtain three, the space coordinate of characteristic point, first according to national forest park in Xiaokeng, obtain the actual coordinate of spatial point and the relation of pixel planes coordinate:

$x=\frac{zX}{f},y=\frac{zY}{f}---\left(3\right)$

Wherein, (x, y, z) be the actual coordinate of spatial point, and (X, Y) is characteristic point physical coordinates on picture；Below equation is obtained according to formula (3) and spatial relation:

$\mathrm{\Δ}l=\sqrt{\frac{{(z_1{X}_1-z_2{X}_2)}^2+{(z_{1}Y-z_2{Y}_2)}^2}{f^2}+{(z_1-z_2)}^2}---\left(4\right)$

$\mathrm{\Δ}l=\sqrt{\frac{{(z_2{X}_2-z_3{X}_3)}^2+{(z_2{Y}_2-z_3{Y}_3)}^2}{f^2}+{(z_2-z_3)}^2}---\left(5\right)$

$\frac{z_1-z_2}{z_1{X}_1-z_2{X}_2}=\frac{z_2-z_3}{z_2{X}_2-z_3{X}_3}---\left(6\right)$

Wherein, v is the movement velocity of guide rail, and Δ t is movement time interval, and Δ l is the move distance of camera, (X in the Δ t time_i,Y_i) it is the pixel planes coordinate of the i-th frame spatial point, (x_i,y_i,z_i) for being separated by the actual coordinate of i × Δ t spatial point；(i=1,2,3), solve (x by simultaneous formula (4), (5), (6)₁,y₁,z₁), it is object point real space coordinate.