CN107966135B

CN107966135B - Multi-view vision measurement method based on dome structure

Info

Publication number: CN107966135B
Application number: CN201711128717.4A
Authority: CN
Inventors: 陈国华; 张爱军; 郑玥; 邢健; 于洪杰; 王永生
Original assignee: Beijing University of Chemical Technology
Current assignee: Beijing University of Chemical Technology
Priority date: 2017-11-15
Filing date: 2017-11-15
Publication date: 2020-07-24
Anticipated expiration: 2037-11-15
Also published as: CN107966135A

Abstract

The invention relates to a dome structure-based multi-view vision measurement method, which adopts multiple cameras to construct a vision measurement model capable of realizing all-dimensional and multi-angle measurement, designs a full-view measurement scheme, and researches key technical problems in the full-view measurement scheme through theoretical analysis and experimental simulation so as to meet the pose measurement of a multi-degree-of-freedom serial mechanical arm in various motion tracks. The invention realizes the omnibearing synchronous real-time detection of the dynamic errors of each connecting rod and the end effector of the mechanical arm, deeply studies the error distribution and the source of each component of the mechanical arm in the operation process, constructs a novel measurement and study system of the dynamic errors of the mechanical arm on the basis of integral analysis, and can more comprehensively and accurately reflect the integral precision of the mechanical arm in series connection so as to improve the precision, the rapidity and the reliability of the control of the mechanical arm, thereby providing a theoretical basis for the research of the pose precision and the track precision of a mechanical arm system and providing data support for the error real-time correction of a servo control system.

Description

Multi-view vision measurement method based on dome structure

Technical Field

The patent relates to a multi-view visual field model based on a dome structure.

Background

The dome structure aims to enable cameras to be distributed more reasonably, a multi-view visual field model capable of realizing all-around and multi-angle measurement is constructed under the dome structure, key technical problems are researched through theoretical analysis and experimental simulation, and finally pose measurement of the multi-degree-of-freedom mechanical arm in different motion tracks is met.

Disclosure of Invention

The purpose of this patent lies in: the cameras are distributed more reasonably, and a multi-view visual field model capable of realizing all-around and multi-angle measurement is constructed under the dome structure.

In order to achieve the purpose, the technical scheme adopted by the patent is as follows: the dome structure is composed of a plurality of points on the same spherical surface, three adjacent points which are not collinear form a basic unit of the dome structure, and the distances between the points are fixed and the same. From the geometric characteristics of the spherical surface, the spherical surface has only one spherical center, and the distance between the spherical center and each point on the dome structure is the radius of the sphere. The basic unit of the dome structure is also a measuring unit for visual measurement, namely, the multi-vision with three cameras as a group is used for carrying out visual measurement on objects under the dome structure.

A multi-view vision measurement method based on a dome structure is characterized in that: a dome structure is constructed, the dome structure is composed of a plurality of points on the same spherical surface, three adjacent points which are not collinear form a basic unit of the dome structure, and the distances between the adjacent points are fixed and the same.

The target object is placed at the center of the sphere, the point on the dome structure represents the position of a certain camera, and the vector from the point on the dome structure to the center of the sphere represents the positive direction of the camera, i.e., the direction in which the target object can be photographed. A common area that can be captured by a plurality of cameras, for this purpose the measurement field under the dome structure. The basic unit of the dome structure is also a measuring unit for visual measurement, namely, the multi-vision with three cameras as a group is used for carrying out visual measurement on objects under the dome structure.

Further, P is constructed in the common area₁A sphere centered at R, representing the dome structure viewing area. The observation area of the dome structure not only needs to satisfy the observable distance of the camera, but also needs to be ensured in the common area of the camera, so that R takes a smaller value between the two, namely

R＝(d_R，f_max-f_min)_min

Wherein d is_RRepresents P₁Distance to the upper bound of the observable area.

Defining a binocular visual field model expression as

Wherein α represents the angle between the camera and the reference plane, the distance between adjacent cameras is 2d, the observable angle of the cameras is 2 β, and the images are takenMinimum viewing distance of machine is f_minThe farthest observation distance of the camera is f_max。

Further, in the present invention,

1) increasing d corresponds to increasing the distance between the cameras, and decreasing the area of intersection of the fields of view of the cameras in a limited space, i.e. decreasing the measurement field of the field of view model.

2) Increasing β corresponds to increasing the viewing angle of the camera, increasing the area of intersection of the field of view of the camera in a limited space, i.e., increasing the field of view model.

3) Increasing α corresponds to increasing the top view angle of the camera, decreasing the effective space, i.e., the intersection of the field of view of the camera, i.e., decreasing the field of view model measurement.

The geometric characteristic that points on the dome structure are all on the same spherical surface is utilized, and the spherical center of the spherical surface is marked as P₁，P₁Simultaneously, the position of the mechanical arm to be measured is also used for constructing a P₁A sphere centered at R represents the range of motion of the robot arm. Therefore, the observation distances of all the cameras on the dome structure to the mechanical arm to be measured are basically consistent, and the calculated amount can be reduced by utilizing various geometric characteristics of the ball.

For the active space of the robotic arm to be within both the viewable area of the base unit under the dome structure (i.e., the common portion of the viewing area of the camera) and the objective viewing distance of the camera itself (i.e., not less than the minimum viewing distance nor greater than the maximum viewing distance), the active space is defined as the visually measurable viewing area under the dome structure.

By analogy with a two-dimensional space model, P is constructed in an observable public area₁A circle centered at R, which represents the viewing area of the dome structure. In order to ensure that the observation area of the dome structure not only meets the observable distance of the camera, but also is ensured in the common area of the camera, R takes a smaller value between the two, namely

R＝(d_R，f_max-f_min)_min

Therefore, the multi-view visual field model expression is defined as

In the formula, V_tripleRepresenting the measuring field of the multi-view visual field model, α, β and d are field model parameters related to the actual placement position of the camera, f_min、f_maxThe parameters are intrinsic parameters of the camera and do not change along with the change of the field model.

In an actual environment, the center of the sphere represents the position of the target object, the point on the dome structure represents the position of a certain camera, and the vector from the point on the dome structure to the center of the sphere represents the positive direction of the camera, i.e., the direction in which the target object can be photographed. The common area which can be shot by a plurality of cameras is the measuring field under the dome structure.

Compared with the prior art, the advantage of this patent lies in: the dynamic errors of all connecting rods and end effectors of the mechanical arm are detected in an all-round and synchronous mode in real time, error distribution and sources of all components of the mechanical arm in the operation process are deeply researched, a novel measurement and research system of the dynamic errors of the mechanical arm is built on the basis of overall analysis, the overall precision of the mechanical arm in series connection can be reflected more comprehensively and accurately, the accuracy, the rapidity and the reliability of mechanical arm control are improved, a theoretical basis is provided for research of pose precision and track precision of a mechanical arm system, and data support can be provided for error real-time correction of a servo control system.

Drawings

FIG. 1 schematic view of a dome structure

FIG. 2 is a schematic view of a camera model in which the camera has a viewing angle of 2 β and a minimum viewing distance f_minThe farthest observation distance is f_max。

FIG. 3 binocular Vision schematic view, C₁、C₂Respectively showing a camera 1 and a camera 2, C₁、C₂At a distance of 2d, C₁、C₂Has an optical axis of S₁、S₂，S₁、S₂The angle to the reference plane is α, which means the angle the camera makes with the reference plane the observable area of the camera is a sector with a sector angle of 2 β.

FIG. 4 is a schematic view of a multi-view visual field measurement unit, wherein C₁、C₂、C₃Respectively showing a camera 1, a camera 2, a camera 3, C₁C₂、C₁C₃、C₂C₃All the distances between the two are 2d and C₁、C₂、C₃Has an optical axis of S₁、S₂、S₃，S₁、S₂、S₃The angle between the camera and the reference plane is α, namely the angle between the camera and the reference plane is represented, the observable area of the camera is a cone area, and the included angle between generatrices of the cone is 2 β.

Fig. 5 is a simulation of the basic unit under a dome structure using matlab.

Fig. 6 is a graph of the relationship between the field of measurement and α, d for the multi-ocular visual field model using matlab.

Detailed Description

Fig. 5 is a simulation of the basic unit under a dome structure using matlab, where the cone represents the viewing area of the camera, the apex of the cone represents the position of the camera on the dome structure, and the area where the cones intersect is a common part of the viewing area of the camera, i.e. the area where visual measurements can be made.

FIG. 6 is a diagram of the multi-view visual field model measurement domain versus α, d using matlab.

As can be seen from fig. 6, for the multi-view visual field model under the dome structure,

1) when the defined conditions are satisfied, V is held constant at α, β_tripleDecreasing with increasing d, i.e. the measurement field of the multi-ocular visual model decreases with increasing distance between the cameras.

2) When the defined conditions are satisfied, V is maintained at β d_tripleIt is seen from fig. 5 that the theory is analyzed in the actual scene,

3) Increasing α corresponds to increasing the top view angle of the camera, and decreasing the effective space, i.e., the intersecting area of the fields of view of the camera, i.e., decreasing the measurement field of the field of view model, intuitively corresponds to the camera gradually moving from the intersection of the optical axes to the parallel optical axes.

The above conclusion is in line with the field of view change in the actual scene, so that the correctness of the multi-view visual field model under the dome structure provided by the invention is verified.

Claims

1. A multi-view vision measurement method based on a dome structure is characterized in that: firstly, constructing a dome structure, wherein the dome structure is composed of a plurality of points on the same spherical surface, three adjacent points which are not collinear form a basic unit of the dome structure, and the distances between the adjacent points are fixed and the same;

the target object is placed at the center of a sphere, a point on the dome structure represents the position of a certain camera, and a vector from the point on the dome structure to the center of the sphere represents the positive direction of the camera, namely the direction of shooting the target object; a common area which can be shot by a plurality of cameras, namely a measurement area under a dome structure; the basic unit of the dome structure is also a measuring unit for visual measurement, namely, the multi-vision with three cameras as a group is used for carrying out visual measurement on objects under the dome structure.

2. The method of claim 1, wherein:

constructing P in common area₁A circle centered at R and having a radius representing the viewing area of the dome structure; the observation area of the dome structure not only needs to satisfy the observable distance of the camera, but also needs to be ensured in the common area of the camera, so that R takes a smaller value between the two, namely

R＝(d_R，f_max-f_min)_min

Wherein d is_RRepresents P₁Distance to the upper bound of the observable area;

defining a binocular visual field model expression as

Wherein α is the angle between the camera and the reference plane, the distance between adjacent cameras is 2d, the observation angle of the camera is 2 β, and the minimum observation distance of the camera is f_minThe farthest observation distance of the camera is f_max。

3. The method of claim 1, wherein:

1) increasing d corresponds to increasing the distance between the cameras, and the measurement field of the field model is reduced in a limited space, namely the intersecting area of the fields of view of the cameras is reduced;

2) increasing β corresponds to increasing the observable angle of the camera, and the intersection area of the fields of view of the camera is increased under the limited space, namely, the measurement field of the field of view model is increased;