CN103817699A - Quick hand-eye coordination method for industrial robot - Google Patents

Abstract

A quick hand-eye coordination method for an industrial robot is characterized by comprising the following steps of (1) calibrating a depth camera by using a calibrating method of Zhengyou Zhang, and acquiring internal and external parameters of the depth camera; (2) acquiring the depth information of a detection target point by using a triangulation survey principle of the depth camera; (3) acquiring coordinates of the target point in a camera coordinate system by using a camera forming model; (4) describing the relation between the camera coordinate system and a robot world coordinate system by using a Bursa model; and (5) solving parameters of the Bursa model by using an indirect balancing model so as to acquire a hand-eye coordination parameter of the industrial robot. The quick hand-eye coordination method for the industrial robot has the advantages that the internal and internal parameters of the depth camera can be quickly calculated; the three-dimensional coordinates of the depth camera can be effectively extracted; the hand-eye relation of the industrial robot can be quickly calculated; and the operation method is simple and convenient.

Description

One is industrial robot hand eye coordination method fast

Technical field

The present invention relates to robot vision field, be specifically related to Robot Hand-eye coordination approach.

Background technology

Computer vision refers to and replaces human eye that target is identified, followed the tracks of and measures with camera, and as the current study hotspot in forward position the most, the Robotics based on computer vision is one of key technology.Vision can provide abundant environment, target information for robot, for judgement, the decision-making of robot provide foundation.The hand eye coordination system of robot, is mainly divided into two large class: Eye-to-Hand and Eye-in-Hand at present.Camera is fixed on the machine-independent people in certain workspace by the former, and the latter is generally fixed on end effector of robot by camera.

For Eye-in-Hand hand eye coordination, there are many different hand and eye calibrating methods, there is researcher that camera and end effector of robot are done to as a whole modeling, so just robot standard error cannot be distinguished, also there is researcher to adopt the normal derivative method of optical flow field, but thisly rotatablely move and calculate translation vectors by two, can will reduce to a great extent computational accuracy., there is a fixing homogeneous transformation relation in Eye-to-Hand hand eye coordination system, imaging and the robot motion on camera is irrelevant for target object between camera coordinate system and industrial robot coordinate system.Traditional Eye-to-Hand hand eye coordination, employing be that the parallax information of left and right between two cameras calculates depth information, binocular vision Stereo Matching Algorithm is not also very ripe, amount of calculation is very large, the error of depth information is also larger.

Therefore, the defect that existing Robot Hand-eye coordination approach exists is: method of operating complexity, and precision is generally not high, and computational methods are too complicated.

Summary of the invention

In order to overcome existing industrial robot hand eye coordination method deficiency, the present invention proposes a kind of Robot Hand-eye coordination technique based on degree of depth camera and boolean Sha model, show by experiment, this Robot Hand-eye coordination approach is simple, convenient, can effectively carry out vision guide task.

The technical solution adopted for the present invention to solve the technical problems comprises the following steps:

1), first adopt the scaling method of Zhang Zhengyou to demarcate degree of depth camera, obtain its inside and outside parameter.

2) principle of triangulation, by degree of depth camera obtains the depth information that detects impact point, has following formula:

$D=\frac{b*f}{d}---\left(1\right)$

Wherein, D is depth information, described degree of depth camera comprises " infrared camera " and " infrared projection machine ", " infrared camera " and " infrared projection machine;, be horizontal positioned, b is the length of the horizontal base line between " infrared camera " and " infrared projection machine "; f is the focal length of " infrared camera ", d is two parallaxes between camera;

3), obtain and detect the coordinate of impact point at camera coordinate system by camera imaging model, as a certain point coordinates (X in camera coordinate system _c, Y _c, Z _c) can calculate and obtain by principle of triangulation, there is following formula:

$X_C=\frac{(a-a_0)Z_C}{f_x}=\frac{(a-a_0)*D}{f_x}$

$Y_C=\frac{(b-b_0)Z_C}{f_y}=\frac{(b-b_0)*D}{f_y}$

Z _C＝D (2)

A certain point coordinates (X in camera coordinate system _c, Y _c, Z _c), its subpoint coordinate in imaging plane pixel coordinate system is (a, b), and the point coordinates that corresponding imaging plane physical coordinates is is (x, y), and each pixel unit is dx and dy at x and actual range corresponding to y direction, f _x=f/dx, f _y=f/dy, the intersecting point coordinate of camera coordinate system and imaging plane pixel coordinate system is (a ₀, b ₀).

4), the relation of camera coordinate system and robot world's coordinate system is described with boolean Sha model, has following formula:

$\left[\begin{array}{c}{X}_W\\ Y_W\\ Z_W\end{array}\right]=(1+m)\left[\begin{array}{c}{X}_C\\ Y_C\\ Z_C\end{array}\right]+\left[\begin{array}{ccc}0& {\mathrm{\ϵ}}_z& -{\mathrm{\ϵ}}_y\\ -{\mathrm{\ϵ}}_z& 0& {\mathrm{\ϵ}}_{\text{x}}\\ {\mathrm{\ϵ}}_y& -{\mathrm{\ϵ}}_x& 0\end{array}\right]\left[\begin{array}{c}{X}_C\\ Y_C\\ Z_C\end{array}\right]+\left[\begin{array}{c}\mathrm{\Δx}\\ \mathrm{\Δy}\\ \mathrm{\Δz}\end{array}\right]---\left(3\right)$

Wherein, Δ x, Δ y, Δ z are two translational movements between rectangular coordinate system in space, ε _x, ε _y, ε _zbe rotation parameter, m is scale parameter, (X _c, Y _c, Z _c) be any coordinate of the camera coordinate system that obtains in formula (2), (X _w, Y _w, Z _w) be coordinates of targets value under robot coordinate system, obtain by the manual teaching of robot teach box.

5), utilize indirect adjustment model to solve the parameter of boolean Sha model, thereby obtain parameter (Δ x, Δ y, Δ z, the ε of industrial robot hand eye coordination _x, ε _y, ε _z, m).

Advantage of the present invention is: the inside and outside parameter of compute depth camera fast; Can effectively extract the three-dimensional coordinate of degree of depth camera; Can calculate industrial robot trick relation fast, method of operating is simple and convenient.

Accompanying drawing explanation

Fig. 1 is industrial robot fast hand eye coordinate method schematic diagram of the present invention.

Point 1, point 23, point 34, degree of depth camera 5, industrial robot 6, operating platform in the drawings, 1,

The specific embodiment

Below in conjunction with accompanying drawing, the invention will be further described.As Fig. 1, the present invention is achieved in that degree of depth camera 4 is arranged on as left and right, 1 meter of 6 top of Fig. 1 operating platform.Adopt the scaling method of Zhang Zhengyou to obtain degree of depth camera inside and outside parameter, by principle of triangulation and the camera imaging model of degree of depth camera, in computing platform, put one, point two and the coordinate of point three under degree of depth camera coordinate system.Point one, point two and the coordinate of point three in robot coordinate system, obtained by the robot manual teaching of 5 teach box.Therefore, camera coordinate system coordinate and robot coordinate system's coordinate of the point one of acquisition, point two and point three correspondences are updated to boolean Sha model, utilize indirect adjustment model to solve the parameter of boolean Sha model, thereby obtain parameter (Δ x, Δ y, Δ z, the ε of industrial robot hand eye coordination _x, ε _y, ε _z, m), degree of depth camera and industrial robot just can co-ordinations.For whether boolean Sha model of verifying proposition is applicable to industrial robot hand and eye calibrating fast, 3 points of under we selected robot coordinate system 3 and corresponding camera coordinate system, as table 1:

Table 1 robot coordinate system and visual coordinate are corresponding 3 point coordinates

Carry out error analysis by choosing multiple points, what error ratio was larger mainly concentrate on, and x and y sit on target value, by asking the mean value of multiple points, can obtain some the mean error between point apart from d _average(mm) be:

$d_{\mathrm{average}}=\sqrt{1.557+2.220+1.493+1.786+1.449+1.674+2.383+1.662}=1.78$

Therefore, good result can be obtained by boolean Sha model, the function such as welding, spraying of robot can be met completely.

Claims (1)

1. an industrial robot hand eye coordination method fast, is characterized in that hand eye coordination method is:

1), first adopt the scaling method of Zhang Zhengyou to demarcate degree of depth camera, obtain its inside and outside parameter.

2) principle of triangulation, by degree of depth camera obtains the depth information that detects impact point, has following formula:

$D=\frac{b*f}{d}---\left(1\right)$

Wherein, D is depth information, described degree of depth camera comprises " infrared camera " and " infrared projection machine ", " infrared camera " and " infrared projection machine " is horizontal positioned, b is the length of the horizontal base line between " infrared camera " and " infrared projection machine ", f is the focal length of " infrared camera ", and d is two parallaxes between camera;

3), obtain and detect the coordinate of impact point at camera coordinate system by camera imaging model, as a certain point coordinates (X in camera coordinate system _c, Y _c, Z _c) can calculate and obtain by principle of triangulation, there is following formula:

$X_C=\frac{(a-a_0)Z_C}{f_x}=\frac{(a-a_0)*D}{f_x}$

$Y_C=\frac{(b-b_0)Z_C}{f_y}=\frac{(b-b_0)*D}{f_y}$

Z _C＝D (2)

A certain point coordinates (X in camera coordinate system _c, Y _c, Z _c), its subpoint coordinate in imaging plane pixel coordinate system is (a, b), and the point coordinates that corresponding imaging plane physical coordinates is is (x, y), and each pixel unit is dx and dy at x and actual range corresponding to y direction, f _x=f/dx, f _y=f/dy, the intersecting point coordinate of camera coordinate system and imaging plane pixel coordinate system is (a ₀, b ₀).

4), the relation of camera coordinate system and robot world's coordinate system is described with boolean Sha model, has following formula:

$\left[\begin{array}{c}{X}_W\\ Y_W\\ Z_W\end{array}\right]=(1+m)\left[\begin{array}{c}{X}_C\\ Y_C\\ Z_C\end{array}\right]+\left[\begin{array}{ccc}0& {\mathrm{\ϵ}}_z& -{\mathrm{\ϵ}}_y\\ -{\mathrm{\ϵ}}_z& 0& {\mathrm{\ϵ}}_{\text{x}}\\ {\mathrm{\ϵ}}_y& -{\mathrm{\ϵ}}_x& 0\end{array}\right]\left[\begin{array}{c}{X}_C\\ Y_C\\ Z_C\end{array}\right]+\left[\begin{array}{c}\mathrm{\Δx}\\ \mathrm{\Δy}\\ \mathrm{\Δz}\end{array}\right]---\left(3\right)$

Wherein, Δ x, Δ y, Δ z are two translational movements between rectangular coordinate system in space, ε _x, ε _y, ε _zbe rotation parameter, m is scale parameter, (X _c, Y _c, Z _c) be any coordinate of the camera coordinate system that obtains in formula (2), (X _w, Y _w, Z _w) be coordinates of targets value under robot coordinate system, obtain by the manual teaching of robot teach box.

5), utilize indirect adjustment model to solve the parameter of boolean Sha model, thereby obtain parameter (Δ x, Δ y, Δ z, the ε of industrial robot hand eye coordination _x, ε _y, ε _z, m).