阅读说明：本技术 基于结构光的空气预热器转子三维测量与可视化方法 (Air preheater rotor three-dimensional measurement and visualization method based on structured light ) 是由 刘君 杨延西 黄雪飞 邓毅 魏永贵 宋念龙 潘正权 蒲亨林 刘朝成 金明峰 易广 于 2020-04-21 设计创作，主要内容包括：本发明公开了一种基于结构光的空气预热器转子三维测量与可视化方法。该方法通过提取相邻环状点云交集部分进行ICP配准,省去了相机之间位姿关系的标定过程,避免了外参标定带来的误差,极大缩减了算法复杂度,并提高了测量结果的精确度。同时,使用圆点阵列光源,由于实际情况下圆点有可能缺损,从而导致圆心检测的偏差,本发明提出使用近邻估算的方法进行缺损圆点圆心计算修正,既保证了匹配的准确性,又简单快速。本发明为大型旋转物体,尤其是大型空预器转子的直接三维测量提供了一种新的简单、快速、准确的三维测量方法和新思路。(The invention discloses a structured light-based three-dimensional measurement and visualization method for an air preheater rotor. According to the method, ICP registration is carried out by extracting the intersection part of adjacent annular point clouds, so that the calibration process of the pose relationship between cameras is omitted, errors caused by external reference calibration are avoided, the algorithm complexity is greatly reduced, and the accuracy of the measurement result is improved. Meanwhile, the round dot array light source is used, and the round dots are possibly damaged under the actual condition, so that the deviation of circle center detection is caused. The invention provides a new simple, quick and accurate three-dimensional measurement method and a new idea for the direct three-dimensional measurement of a large rotating object, particularly a large air preheater rotor.)

1. The method for three-dimensional measurement and visualization of the rotor of the air preheater based on the structured light comprises the following steps:

s1: performing off-line calibration on all binocular cameras to obtain camera internal parameters for stereoscopic vision correction; a group of measuring units are formed by a group of eye cameras and a dot array light source;

s2: uniformly fixing mark blocks along the radius of the rotor at the cold end of the air preheater, determining the number of the mark blocks according to the radius size of a measured object, using the mark blocks as an initial mark for determining one circle of rotation of the rotor and as a boundary line of adjacent camera measuring areas, and providing an initial value for cloud splicing of rear end points; then arranging measurement units corresponding to the mark blocks one by one;

s3: turning on a dot array light source, supplementing light, enabling a rotating shaft of the rotor and all cameras to be in a state to be started, and synchronously acquiring images of one circle of rotation of the rotor by all the cameras after the rotating shaft enters a uniform rotation state; splicing the acquired images to form an annular image, and obtaining homography H between two adjacent frames;

s4: the air preheater rotor continues to rotate, all cameras start to collect images containing laser dot arrays, and when the cameras rotate for one circle, laser sweeps the surface of the whole air preheater rotor;

s5: calculating the local three-dimensional space coordinates of the rotor by utilizing a stereoscopic vision principle, calculating the space point coordinates of the rotor surface in the current camera view field range according to the epipolar constraint and the dot array characteristics of the collected left and right laser images, and obtaining the local three-dimensional point cloud of the rotor;

s6: splicing the local three-dimensional point cloud obtained in the step S5 by using the homography H obtained in the step S3, and restoring the real position of each frame of image measurement result to obtain annular point cloud;

s7: repeating the steps S5 and S6 on all images collected by the binocular camera to obtain corresponding annular point clouds;

s8: extracting the intersection part of two adjacent annular point clouds, and obtaining a pose relation R, t between the two groups of point clouds by utilizing a nearest iteration algorithm through iteration minimization matching errors to complete the splicing of the two groups of annular point clouds;

s9: and (5) repeating the step S8 on all the annular point clouds to complete the three-dimensional point cloud measurement of the whole air preheater rotor, and performing three-dimensional visual display on a display terminal.

2. The structured light based air preheater rotor three dimensional measurement and visualization method of claim 1, wherein: the number of the measuring units is n groups, and n is more than or equal to 3.

3. The structured light based air preheater rotor three dimensional measurement and visualization method of claim 2, wherein: the arrangement of the measurement units in step S2 is specifically: n groups of measuring units are uniformly arranged in front of a radius on the surface of the rotor, so that the conditions are met: the n groups of measuring units can shoot a complete radius of the rotor, and two adjacent measuring units can observe the same mark block.

4. The structured light based air preheater rotor three dimensional measurement and visualization method of claim 3, wherein: the step of stitching the acquired images described in step S3 specifically includes:

A. detecting feature points from two adjacent frames of images;

B. matching the characteristic points to obtain a matching point set;

C. solving homography H through the matching point set to obtain a mapping relation between two adjacent frames of images; D. splicing two adjacent frames, wherein a homography transformation formula is shown as a formula (1):

in the formula [ u ]_k，v_k]^TAnd [ u ]_k`，v<sub>k</sub>`]^TThe pixel coordinates of the same feature point on two adjacent frames are respectively.

5. The structured light based air preheater rotor three dimensional measurement and visualization method of claim 3, wherein: the extraction of the dot array feature in step S5 is as follows:

A. performing region segmentation on the image, and judging whether the circular dots are defective or not according to the shape of the region; storing the complete dot coordinates and the corresponding sequence positions, and carrying out next step on the defective dots;

B. according to the sequence position of the defect round point and the stored complete round point coordinate and sequence position, and according to the sequence position of the defect round point and the complete round point coordinate and sequence position of the adjacent round point, interpolation estimation is carried out, and the method specifically comprises the following steps:

estimating the difference value of defective dots at the edge of the dot array according to the centers of 4 complete dots adjacent to each other; interpolation estimation is carried out on defective dots in the dot array and 4 complete dot circle centers a, b, d and e adjacent to four corners of the defective dots, and the coordinate calculation formula of the defective dots c is shown as formula (2) and formula (3):

in the formula I₁、l₂Is a straight line where the defective dots are located; (x)_a，y_a)、(x_b，y_b)、(x_d，y_d)、(x_e，y_e) The coordinates of the circle centers of a, b, d and e are respectively; solving simultaneous equations (2) and (3) to obtain the coordinate of the defective dot c;

C. the feature that the dot arrays are sequentially arranged in the left view and the right view is utilized to correspondingly match the left view and the right view in sequence, and the parallax is directly calculated to obtain the three-dimensional coordinates [ X, Y, Z ] of each dot]^T。

6. The structured light based air preheater rotor three dimensional measurement and visualization method of claim 3, wherein: the step of stitching the annular point clouds in the step S8 specifically includes:

A. the point sets of the intersection of the inner ring and the outer ring of two groups of adjacent annular point clouds are respectively as follows:

and

B. define the error term for the jth pair of points:

e_j＝P_m-(RP_m`+t) (j＝1，2，...，m) (5)

wherein R and t are respectively a point setPoint-to-point set ofRotating matrix translation vectors corresponding to the midpoints;

C. constructing a least square problem:

and (5) obtaining R and t when the error term is minimum through iterative solution, and splicing two adjacent annular point clouds.

7. The structured light based air preheater rotor three-dimensional measurement and visualization method as recited in any one of claims 1 to 6, wherein: the mark block is a cross-shaped reflective ceramic block.

Technical Field

The invention relates to a three-dimensional measurement technology, in particular to a structured light three-dimensional measurement method based on a rotary table.

Background

The rotary air preheater is one of the main auxiliary machines of large industrial boilers, and is called an air preheater for short. The device is a key device for realizing efficient combustion of a boiler and safe and economic operation of a unit in a thermal power plant. As the flue gas (high temperature is near 400 ℃) is input into the upper part (hot end) of the air preheater, and the air (low temperature is about 20 ℃) is blown into the lower part (cold end), the whole air preheater is unevenly heated, mushroom-shaped deformation is generated, a large amount of combustion-supporting air is leaked into a flue, the power consumption of a fan is increased, and the combustion heat efficiency of the boiler is reduced.

Therefore, how to detect the appearance and the thermal deformation of the air preheater so as to guide the design of the air preheater is very important for effectively reducing air leakage of the air preheater and improving the boiler efficiency. At present, no device for measuring the whole surface appearance and radial deformation exists, so that the design of the air preheater depends on theoretical calculation and numerical simulation modes. For example, a method of reserving a clearance value is adopted, but the clearance value changes with load change in actual production, and the radius ratio of the air preheater is large, and the deformation in the radial direction is different, so that the reservation is generally large, and the air leakage is large. In addition, because the air preheater has no light when operating in a sealed environment, an operator cannot see the operation condition inside, and the thermal state operation and maintenance are inconvenient.

Three-dimensional measurement techniques have matured in recent years, but there has been no great progress in directly performing three-dimensional measurements on a rotating object in an operating state without interfering with the operation of the apparatus.

The air preheater belongs to a large industrial part, and in the process of three-dimensional measurement of the large industrial part, due to the field range of a single camera, the three-dimensional reconstruction and measurement of the geometric shape of an object are difficult to realize by a traditional computer vision method based on structured light (such as documents: Songlong, Liguang Hui, Liu Bao, and the like; calibration of an effective visual area based on a binocular measurement system based on structured light [ J ]. mechanical engineer, 2007, (9): 24-26.).

Z L201610443919.7 patent, a three-dimensional full-field deformation measurement method of blade, although it can solve the three-dimensional measurement problem of rotating object, the method carries on complex rotation axis position and attitude calibration and space point calculation, but most practical situations do not meet the rotation axis calibration requirement, so it is difficult to apply to the three-dimensional measurement of air preheater rotor.

Therefore, how to perform three-dimensional measurement on a large rotating air preheater rotor in an operating state still remains an unsolved problem in the field.

Disclosure of Invention

The invention provides a structured light-based three-dimensional measurement and visualization method for a rotor of an air preheater, which aims to realize efficient and accurate three-dimensional measurement of the rotor of a large-scale rotary air preheater in an operating state.

The technical scheme adopted by the invention is as follows: the method for three-dimensional measurement and visualization of the rotor of the air preheater based on the structured light comprises the following steps:

s1: performing off-line calibration on all binocular cameras to obtain camera internal parameters for stereoscopic vision correction; a group of measuring units are formed by a group of eye cameras and a dot array light source;

s2: uniformly fixing mark blocks along the radius of the rotor at the cold end of the air preheater, determining the number of the mark blocks according to the radius size of a measured object, using the mark blocks as an initial mark for determining one circle of rotation of the rotor and as a boundary line of adjacent camera measuring areas, and providing an initial value for cloud splicing of rear end points; then arranging measurement units corresponding to the mark blocks one by one;

s3: turning on a dot array light source, supplementing light, enabling a rotating shaft of the rotor and all cameras to be in a state to be started, and synchronously acquiring images of one circle of rotation of the rotor by all the cameras after the rotating shaft enters a uniform rotation state; splicing the acquired images to form an annular image, and obtaining homography H between two adjacent frames;

s4: the air preheater rotor continues to rotate, all cameras start to collect images containing laser dot arrays, and when the cameras rotate for one circle, laser sweeps the surface of the whole air preheater rotor;

s5: calculating the local three-dimensional space coordinates of the rotor by utilizing a stereoscopic vision principle, calculating the space point coordinates of the rotor surface in the current camera view field range according to the epipolar constraint and the dot array characteristics of the collected left and right laser images, and obtaining the local three-dimensional point cloud of the rotor;

s6: splicing the local three-dimensional point cloud obtained in the step S5 by using the homography H obtained in the step S3, and restoring the real position of each frame of image measurement result to obtain annular point cloud;

s7: repeating the steps S5 and S6 on all images collected by the binocular camera to obtain corresponding annular point clouds;

s8: extracting the intersection part of two adjacent annular point clouds, and obtaining a pose relation R, t between the two groups of point clouds by utilizing a nearest iteration algorithm through iteration minimization matching errors to complete the splicing of the two groups of annular point clouds;

s9: and (5) repeating the step S8 on all the annular point clouds to complete the three-dimensional point cloud measurement of the whole air preheater rotor, and performing three-dimensional visual display on a display terminal.

As a further improvement of the invention, the number of the measuring units is n groups, and n is more than or equal to 3.

More preferably, the arrangement of the measuring units in step S2 is specifically as follows: n groups of measuring units are uniformly arranged in front of a radius on the surface of the rotor, so that the conditions are met: the n groups of measuring units can shoot a complete radius of the rotor, and two adjacent measuring units can observe the same mark block.

As a further improvement of the present invention, the step of stitching the acquired images in step S3 specifically includes:

A. detecting feature points from two adjacent frames of images;

B. matching the characteristic points to obtain a matching point set;

C. solving homography H through the matching point set to obtain a mapping relation between two adjacent frames of images; D. splicing two adjacent frames, wherein a homography transformation formula is shown as a formula (1):

in the formula [ u ]_k，v_k]^TAnd [ u ]_k`，v<sub>k</sub>`]^TThe pixel coordinates of the same feature point on two adjacent frames are respectively.

As a further improvement of the present invention, the step of extracting the dot array feature in step S5 is as follows:

A. performing region segmentation on the image, and judging whether the circular dots are defective or not according to the shape of the region; storing the complete dot coordinates and the corresponding sequence positions, and carrying out next step on the defective dots;

B. according to the sequence position of the defect round point and the stored complete round point coordinate and sequence position, and according to the sequence position of the defect round point and the complete round point coordinate and sequence position of the adjacent round point, interpolation estimation is carried out, and the method specifically comprises the following steps:

estimating the difference value of defective dots at the edge of the dot array according to the centers of 4 complete dots adjacent to each other; interpolation estimation is carried out on defective dots in the dot array and 4 complete dot circle centers a, b, d and e adjacent to four corners of the defective dots, and the coordinate calculation formula of the defective dots c is shown as formula (2) and formula (3):

in the formula I₁、l₂Is a straight line where the defective dots are located; (x)_a，y_a)、(x_b，y_b)、(x_d，y_d)、(x_e，y_e) The coordinates of the circle centers of a, b, d and e are respectively; solving simultaneous equations (2) and (3) to obtain the coordinate of the defective dot c;

C. the feature that the dot arrays are sequentially arranged in the left view and the right view is utilized to correspondingly match the left view and the right view in sequence, and the parallax is directly calculated to obtain the three-dimensional coordinates [ X, Y, Z ] of each dot]^T。

As a further improvement of the present invention, the step of stitching the annular point clouds in step S8 specifically includes:

A. the point sets of the intersection of the inner ring and the outer ring of two groups of adjacent annular point clouds are respectively as follows:

B. define the error term for the jth pair of points:

e_j＝P_m-(RP_m`+t)(j＝1，2，...，m) (5)

wherein R and t are respectively a point setPoint-to-point set ofRotating matrix translation vectors corresponding to the midpoints;

C. constructing a least square problem:

and (5) obtaining R and t when the error term is minimum through iterative solution, and splicing two adjacent annular point clouds.

As a further improvement of the invention, the marking block is a cross-shaped reflecting ceramic block.

The invention has the beneficial effects that: 1) by extracting the intersection part of adjacent annular point clouds to carry out ICP (nearest iterative algorithm) registration, the calibration process of the pose relationship between cameras is omitted, errors caused by external reference calibration are avoided, the algorithm complexity is greatly reduced, and the accuracy of the measurement result is improved. 2) The invention uses the dot array light source, because the dot may be defective under the actual condition, thereby causing the deviation of the circle center detection, the invention provides the method for using the neighbor estimation to calculate and correct the circle center of the defective dot, thereby not only ensuring the matching accuracy, but also being simple and fast. 3) The invention provides a new simple, quick and accurate three-dimensional measurement method and a new idea for the direct three-dimensional measurement of a large rotating object, particularly a large air preheater rotor.

Drawings

Fig. 1 is a schematic view of a measurement unit.

FIG. 2 is a schematic diagram of a dot array light source.

Fig. 3 is a schematic diagram of a flag block and measurement unit arrangement.

Fig. 4 is a schematic image stitching diagram.

Fig. 5 is a model for stitching two adjacent frames of images.

FIG. 6 is a schematic diagram of extracting an intersection of annular point clouds.

FIG. 7 is a schematic diagram of annular point cloud stitching.

Fig. 8 is a schematic view of defect dot correction.

FIG. 9 is a partial three-dimensional point cloud display effect diagram of the rotor of the air preheater according to the embodiment.

Detailed Description

The invention is further illustrated with reference to the following figures and examples.