阅读说明：本技术 基于单目视觉的平面位姿动态测量方法 (Monocular vision-based dynamic plane pose measurement method ) 是由 吴斌 张振 康杰虎 薛婷 于 2020-06-09 设计创作，主要内容包括：本发明涉及一种基于单目视觉的平面位姿动态测量方法,包括如下的步骤：采用两个一维旋转台、一个工业相机和一个激光器构建视觉测量装置,将两个转台分别命名为：水平转台、竖直转台；将两个转台回到零位,利用激光跟踪仪和半球形靶标对两个转台的旋转轴、激光束的空间位姿进行标定,确定方向向量及固定点,并将方向向量归一化；利用激光点深度信息与图像坐标的对应关系,建立深度恢复数学模型；确定激光跟踪仪测量坐标系与测量装置测量坐标系的转换关系；视觉测量装置对目标平面进行扫描,一处位置拍摄一幅图像,同时记录两转台的旋转角度,扫描点个数保证三个以上；由扫描点三位空间坐标进行平面拟合,确定目标平面空间姿态。(The invention relates to a monocular vision-based dynamic plane pose measurement method, which comprises the following steps: the vision measuring device is constructed by adopting two one-dimensional rotating tables, an industrial camera and a laser, and the two rotating tables are respectively named as: a horizontal turntable, a vertical turntable; returning the two rotary tables to zero positions, calibrating the space poses of the rotary shafts of the two rotary tables and the laser beams by using the laser tracker and the hemispherical targets, determining direction vectors and fixed points, and normalizing the direction vectors; establishing a depth recovery mathematical model by utilizing the corresponding relation between the depth information of the laser points and the image coordinates; determining the conversion relation between the measurement coordinate system of the laser tracker and the measurement coordinate system of the measurement device; the vision measuring device scans a target plane, one image is shot at one position, the rotating angles of the two rotary tables are recorded simultaneously, and the number of scanning points is more than three; and carrying out plane fitting by the three-dimensional space coordinates of the scanning points to determine the space attitude of the target plane.)

1. A monocular vision-based dynamic plane pose measurement method comprises the following steps:

(1) the vision measuring device is constructed by adopting two one-dimensional rotating tables, an industrial camera and a laser, and the two rotating tables are respectively named as: horizontal revolving stage, vertical revolving stage.

(2) Returning the two rotary tables to zero positions, calibrating the space poses of the rotary shafts of the two rotary tables and the laser beams by using the laser tracker and the hemispherical targets, determining direction vectors and fixed points, and normalizing the direction vectors;

(3) acquiring actual distances L of a plurality of groups of laser points and image point pixel coordinates u thereof as control points by utilizing the corresponding relation between the depth information of the laser points and the image coordinates, and establishing a depth recovery mathematical model;

(4) establishing a measurement coordinate system of a vision measurement device according to the calibration parameters, and determining the conversion relation between the measurement coordinate system of the laser tracker and the measurement coordinate system of the measurement device;

(5) the vision measuring device scans a target plane, one image is shot at one position, the rotating angles of the two rotary tables are recorded simultaneously, and the number of scanning points is more than three;

(6) according to the calibration parameters and the measurement data, the motion process is divided into two independent processes of horizontal turntable rotation and vertical turntable rotation based on an Euler angle and quaternion method, and the three-dimensional space coordinates of the scanning points are obtained through calculation by using a depth recovery mathematical model;

(7) and carrying out plane fitting by the three-dimensional space coordinates of the scanning points to determine the space attitude of the target plane.

Technical Field

The invention relates to a measuring device based on vision, in particular to a plane pose dynamic measuring method based on monocular vision.

Background

According to different technical requirements and application environments, various measuring methods and devices are produced, wherein the measuring system based on monocular vision is simple in structure, strong in real-time performance and environment adaptability, free of problems such as three-dimensional matching and the like, and is widely concerned. The paper document (i.e., a line structure light vision method [ J ] astronavigation academic newspaper for measuring pose of a non-cooperative large target, 2012,33(6:728-, the difference between the structural mode and the measurement principle is not large, the mapping relation between the characteristic information and the image is established by utilizing the information between the characteristic points, and the pose of the target is solved, but the method is mostly suitable for specific occasions and lacks adaptability and universality.

The granted chinese patent 201410850221.8 (measuring method based on non-orthogonal axis laser theodolite) proposes a "three-axis" architecture of non-orthogonal axis laser theodolite, and two rotary tables and laser beams do not have orthogonal structure requirements, and also do not need to intersect at a point in space, and two one-dimensional rotary tables, one industrial camera and one laser are used to construct a vision measuring device. Compared with the traditional pose measurement mode based on monocular vision, the device has the advantages of simple structure, easiness in installation and lower cost, can realize pose measurement through one-time calibration after assembly, does not need cooperation of cooperative targets in the measurement process, has wide application range and provides a new way for measuring the plane pose.

Disclosure of Invention

The invention provides a monocular vision-based pose measuring method based on an existing non-orthogonal shafting laser theodolite. Because the measuring device adopts a shafting structure, the measuring range is fully expanded by matching the two one-dimensional rotary tables. In addition, the laser point is used as an imaging mark point, and the method does not need to set a special cooperation target any more, so that the application range of the measuring device is wider. Meanwhile, due to the characteristics of the structure and the principle of the device, a camera does not need to be calibrated in the measuring process, and the device is convenient and practical. The technical scheme is as follows:

a monocular vision-based dynamic plane pose measurement method comprises the following steps:

(1) the vision measuring device is constructed by adopting two one-dimensional rotating tables, an industrial camera and a laser, and the two rotating tables are respectively named as: a horizontal turntable, a vertical turntable;

(2) returning the two rotary tables to zero positions, calibrating the space poses of the rotary shafts of the two rotary tables and the laser beams by using the laser tracker and the hemispherical targets, determining direction vectors and fixed points, and normalizing the direction vectors;

(3) acquiring actual distances L of a plurality of groups of laser points and image point pixel coordinates u thereof as control points by utilizing the corresponding relation between the depth information of the laser points and the image coordinates, and establishing a depth recovery mathematical model;

(4) establishing a measurement coordinate system of a vision measurement device according to the calibration parameters, and determining the conversion relation between the measurement coordinate system of the laser tracker and the measurement coordinate system of the measurement device;

(5) the vision measuring device scans a target plane, one image is shot at one position, the rotating angles of the two rotary tables are recorded simultaneously, and the number of scanning points is more than three;

(6) according to the calibration parameters and the measurement data, the motion process is divided into two independent processes of horizontal turntable rotation and vertical turntable rotation based on an Euler angle and quaternion method, and the three-dimensional space coordinates of the scanning points are obtained through calculation by using a depth recovery mathematical model;

(7) and carrying out plane fitting by the three-dimensional space coordinates of the scanning points to determine the space attitude of the target plane.

The invention can construct an omnibearing non-contact dynamic measuring device for the plane pose in order to realize a plane space pose measuring method with higher precision and simple and convenient operation.

Drawings

Fig. 1 is a schematic structural view of a pose measuring device based on monocular vision, wherein 1 is a horizontal rotary table, 2 is a vertical rotary table, 3 is a laser, 4 is a camera, 5 is a vertical rotary shaft, 6 is a horizontal rotary shaft, and 7 is a laser beam.

FIG. 2 is a schematic diagram of the principle of monocular visual depth information based recovery, wherein 8 is the perspective projection center of the camera imaging system, 9 is the laser beam fixed point, 10 is the laser point, 11 is the imaging plane, 12 is the symmetry plane of the imaging plane, 13 is the laser point imaging point, 14 is the imaging plane center, u is the pixel coordinate of the laser point in the horizontal direction, u is the u laser point₀The pixel coordinate of the center of the image plane in the horizontal direction, f is the effective focal length, D is the distance between the optical axis and the laser beam in the horizontal direction, and L is the laser spot distance.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

The direction vectors and fixed points of the horizontal rotary table rotating shaft, the vertical rotary table rotating shaft and the laser beam are key parameters for realizing accurate measurement of the measuring device, and for the convenience of expression, the directions are shown in table 1:

TABLE 1

(1) Two one-dimensional rotating tables, an industrial camera and a laser are adopted to construct a measuring device, and the two rotating tables are named as: horizontal turntable, vertical turntable, as shown in fig. 1;

(2) and the rotary table returns to the zero position, and the spatial pose of the rotary table and the laser beam is calibrated by using the laser tracker and the hemispherical target. 1) Firstly, a rotating shaft of a rotary table is calibrated, a target ball of a laser tracker is fixed on the rotary table, the rotary table is rotated for multiple times at equal intervals, coordinates of the target ball are measured simultaneously, and space circle fitting is carried out on obtained data to obtain a direction vector and a fixed point of a straight line where the rotating shaft is located. 2) Then calibrating a laser beam, adjusting the position of the hemispherical target through a precision displacement platform until the laser aims at the sphere center of the semicircular target ball, replacing the semicircular target ball with a target ball of a laser tracker to obtain the coordinate of the sphere center, repeating the steps, and performing straight line fitting on the obtained data to obtain a direction vector and a fixed point of the straight line where the laser beam is located;

(3) as shown in FIG. 2, according to the laser triangulation, the depth information of the laser spot is calculated in such a manner that

Wherein d is_xAs for the pixel size, the distance information of the laser spot corresponds to the pixel coordinates thereof one by one as can be seen from the equation (1). According to the depth recovery method, the depth recovery mathematical model is established by utilizing the corresponding relation between a plurality of groups of distance information L and the pixel coordinate u in the horizontal direction, compared with the laser triangulation principle, the complicated calibration process is avoided, and the depth information can be directly calculated according to the pixel coordinate of the laser point image point. The realization method comprises the following steps: according to the calibration process of the laser beam in the step (2), acquiring a plurality of groups of laser point actual distances L and image point pixel coordinates u thereof as control points, and determining parameters (a, b, c and d) of a depth recovery mathematical model by using a least square method, wherein the parameters comprise:

L＝au³+bu²+cu+d (2)

(4) setting a measurement coordinate system o-xyz of the laser tracker, establishing a measurement coordinate system o '-x' y 'z' of the measurement device according to calibration parameters, taking a fixed point of a vertical rotating shaft as an original point o ', and the positive direction of the vertical rotating shaft as a z' axis, and determining a conversion matrix M of the measurement coordinate system of the laser tracker and the measurement coordinate system of the measurement device;

[x′ y′ z′ 1]^T＝M[x y z 1]^T(3)

(5) the measuring device scans a target plane, an image is shot at one position, the rotating angles of the two rotary tables are recorded at the same time, the rotating angle of the horizontal rotary table is recorded as alpha, the rotating angle of the vertical rotary table is recorded as beta, and the number of scanning points is more than three;

(6) and analyzing the motion state of the laser beam based on an Euler angle and quaternion method according to the calibration parameters and the measurement data, wherein the motion process is divided into two independent processes of horizontal turntable rotation and vertical turntable rotation.

1) Horizontal rotation alphaThe conversion relation before and after the rotation of the laser beam can be represented by a rotation matrix R_HRepresents:

after horizontal rotation, the horizontal rotation axis is an initial direction vector, a fixed point and a laser beam direction vector, wherein the fixed point is respectively expressed as:

2) vertical rotation beta, rotation matrix R_VExpressed in quaternion form:

wherein q is₀＝cos(β/2)，q₁＝v_xsin(β/2)，q₂＝v_ysin(β/2)，q₃＝v_zsin(β/2)，

After vertical rotation, the laser beam direction vector, the fixed point, is represented as:

through image processing, extracting a laser point, acquiring a laser point pixel coordinate, and obtaining distance information L of a scanning point P by a depth recovery mathematical model, namely determining a space coordinate of the laser point, wherein the space coordinate comprises the following steps:

(7) and performing plane fitting on the scanning points based on the space coordinates of the scanning points to determine a normal vector (l, m, n) of the target plane, namely the space attitude information of the target plane.