阅读说明：本技术 一种辊压机辊面在线三维成像方法 (Online three-dimensional imaging method for roll surface of roll squeezer ) 是由 倪章勇 辛斌杰 王益亮 于 2020-05-25 设计创作，主要内容包括：本发明公开了一种辊压机辊面在线三维成像方法,所述方法包括以下步骤：S1：系统标定；通过一个特定标定物对双目视觉系统进行标定,求解出双目相机的内参和外参；S2：局部点云重建；通过一对线激光照射的辊面图片,恢复单条线激光照射区域辊面的三维点云。本发明能够在辊压机工作状态下精确的对辊面进行三维重建,并对重建的数据进行测量,当发现辊面损毁严重的情况时,及时发出警报,提醒维护人员尽早对辊面进行修复,从而提高设备的运转率,降低维护成本,间接提高产品质量,能够适应恶劣的工作环境,对烟尘、噪音等因素具有很强的抗干扰性。(The invention discloses an online three-dimensional imaging method for a roll surface of a roll squeezer, which comprises the following steps: s1: calibrating a system; calibrating the binocular vision system through a specific calibration object, and solving the internal reference and the external reference of the binocular camera; s2: reconstructing local point cloud; and recovering the three-dimensional point cloud of the roller surface in the single line laser irradiation area through the roller surface pictures irradiated by the pair of line lasers. The invention can accurately carry out three-dimensional reconstruction on the roll surface in the working state of the roll squeezer, measure the reconstructed data, send out an alarm in time when the condition that the roll surface is seriously damaged is found, remind maintenance personnel to repair the roll surface as early as possible, thereby improving the operation rate of equipment, reducing the maintenance cost, indirectly improving the product quality, being capable of adapting to severe working environment, and having strong anti-interference performance on factors such as smoke dust, noise and the like.)

1. The online three-dimensional imaging method for the roll surface of the roll squeezer is characterized by comprising the following steps:

s1: calibrating a system; calibrating the binocular vision system through a specific calibration object, and solving the internal reference and the external reference of the binocular camera;

s2: reconstructing local point cloud; recovering three-dimensional point cloud of the roller surface in a single line laser irradiation area through the roller surface picture irradiated by the pair of line lasers;

s3: point cloud splicing; the binocular camera moves left and right on the lead screw guide rail at a specified speed, the roller surface images are shot at a certain frame rate and are opposite, three-dimensional coordinates of single line laser are recovered for each pair of images, a plurality of line point clouds are spliced into a surface, and then, when reconstruction algorithm is carried out, the roller surface rotates at a certain angular speed, and a plurality of circles of scanning data are spliced into complete roller surface data.

2. The online three-dimensional imaging method for the roll surface of the roller press as claimed in claim 1, wherein in S2, the local point cloud reconstruction step further comprises the following steps:

s21: correcting distortion; due to the influence of the shape of a lens of the camera and the assembly error in the camera, the binocular camera can generate certain radial distortion and tangential distortion in the imaging process, so that the image needs to be corrected by distortion parameters obtained by calibrating the camera, and the image completely conforms to a camera imaging geometric model;

s22: performing three-dimensional correction; the premise of recovering the three-dimensional point cloud of the roller surface according to the parallax is that two imaging planes of the binocular camera are completely aligned in a line mode, and due to errors existing in the hardware installation process, the absolute line alignment of the two camera planes is required to be achieved through a stereo correction algorithm;

s23: extracting a laser central line; the laser stripes have a certain width in the left and right images, and in order to improve the precision of the measuring system, the pixel coordinates of the central line of the laser stripe are required to be accurately extracted;

s23: stereo matching; line alignment of left and right images can be realized according to epipolar rectification, line laser is vertical, and matching of corresponding positions of the two images can be realized by searching the position of the laser center line of the two images in each line;

s24: calculating three-dimensional coordinates; and after the stereo matching is finished, calculating the parallax corresponding to the laser centers of the two images, and then calculating the coordinates of the space points of the line laser irradiation area according to the triangulation principle.

3. The on-line three-dimensional imaging method for the roll surface of the roll squeezer as recited in claim 1, wherein in S3, the implementation of the point cloud splicing part requires fixing four landmark points within the view angle, each pair of images can calculate the spatial coordinates of the four landmark points, and the point cloud is converted into a unified world coordinate system according to the four known spatial coordinates to implement the splicing of the point cloud of the line.

Technical Field

The invention relates to the technical field of roller presses, in particular to an online three-dimensional imaging method for a roller surface of a roller press.

Background

The roller press is widely applied to industries such as building material cement, metallurgical mine, chemical production and the like, has the advantages of high efficiency, energy conservation, environmental protection and the like, is commonly used for grinding hard materials such as steel slag, iron ore, quartz ore and the like, and the roller surface of the roller press is usually ensured to be wear-resistant by adopting a mode of embedding hard alloy studs.

The roller press mainly applies pressure to materials through relative rotation between two rollers, crushed materials are extruded to form a compact material bed, the next process is carried out from the lower parts of the two rollers, and the ground materials are high in hardness and high in abrasiveness, so that roller surface studs can be worn to different degrees after normally running for a period of time, too large metal impurities can directly damage the roller surface, and the roller surface is locally peeled off even in a large area.

If the damaged roll surface is not repaired or replaced in time, unqualified material processing can be caused, and the next process can be directly carried out, so that the product quality is seriously influenced, the roll surface is greatly abraded all the time, and if the roll surface is not timely detected and maintained, the equipment operation rate is low, and the product quality is reduced. Therefore, the key for improving the production quality and the production efficiency is to carry out timely detection and maintenance on the roll surface of the roll squeezer.

The traditional roll surface detection method adopts a manual mode based on human eye judgment or soft rope measurement, has the defects of low efficiency, poor precision, long time consumption, high cost and the like, and cannot meet the requirement of large-scale equipment detection in the existing factory, so improvement is needed.

Disclosure of Invention

The invention aims to solve the defects in the prior art and provides an online three-dimensional imaging method for a roll surface of a roll squeezer.

In order to achieve the purpose, the invention adopts the following technical scheme:

an online three-dimensional imaging method for a roll surface of a roll squeezer comprises the following steps:

s1: calibrating a system; calibrating the binocular vision system through a specific calibration object, and solving the internal reference and the external reference of the binocular camera;

s2: reconstructing local point cloud; recovering three-dimensional point cloud of the roller surface in a single line laser irradiation area through the roller surface picture irradiated by the pair of line lasers;

s3: point cloud splicing; the binocular camera moves left and right on the lead screw guide rail at a specified speed, the roller surface images are shot at a certain frame rate and are opposite, three-dimensional coordinates of single line laser are recovered for each pair of images, a plurality of line point clouds are spliced into a surface, and then, when reconstruction algorithm is carried out, the roller surface rotates at a certain angular speed, and a plurality of circles of scanning data are spliced into complete roller surface data.

Preferably, in S2, the local point cloud reconstruction step further includes the following steps:

s21: correcting distortion; due to the influence of the shape of a lens of the camera and the assembly error in the camera, the binocular camera can generate certain radial distortion and tangential distortion in the imaging process, so that the image needs to be corrected by distortion parameters obtained by calibrating the camera, and the image completely conforms to a camera imaging geometric model;

s22: performing three-dimensional correction; the premise of recovering the three-dimensional point cloud of the roller surface according to the parallax is that two imaging planes of the binocular camera are completely aligned in a line mode, and due to errors existing in the hardware installation process, the absolute line alignment of the two camera planes is required to be achieved through a stereo correction algorithm;

s22: extracting a laser central line; the laser stripes have a certain width in the left and right images, and in order to improve the precision of the measuring system, the pixel coordinates of the central line of the laser stripe are required to be accurately extracted;

s23: stereo matching; line alignment of left and right images can be realized according to epipolar rectification, line laser is vertical, and matching of corresponding positions of the two images can be realized by searching the position of the laser center line of the two images in each line;

s24: calculating three-dimensional coordinates; and after the stereo matching is finished, calculating the parallax corresponding to the laser centers of the two images, and then calculating the coordinates of the space points of the line laser irradiation area according to the triangulation principle.

Preferably, in S3, the implementation of the point cloud splicing part needs to fix four landmark points within the view, each pair of images can calculate the spatial coordinates of the four landmark points, and the point cloud is converted into a unified world coordinate system according to the four known spatial coordinates to implement the splicing of the line point cloud.

Compared with the prior art, the invention has the beneficial effects that:

1. the invention can accurately carry out three-dimensional reconstruction on the roll surface under the working state of the roll squeezer, measure the reconstructed data, and send out an alarm in time when the situation that the roll surface is seriously damaged is found, thereby reminding maintenance personnel to repair the roll surface as early as possible, improving the operation rate of equipment, reducing the maintenance cost and indirectly improving the product quality.

2. The method is suitable for measuring the abrasion of the roll surfaces (stud roll surfaces, groove roll surfaces, herringbone roll surfaces, welding roll surfaces and the like) of various forms and materials, in addition, the roll surfaces are irradiated in a laser mode to obtain binocular images, the method can adapt to severe working environments, has strong anti-interference performance on factors such as smoke dust, noise and the like, and the algorithm design comprises a large number of compensation algorithms, so that tiny errors in the installation process of the device can be corrected, and more accurate measurement is realized.

In conclusion, the roller surface three-dimensional reconstruction method can accurately carry out three-dimensional reconstruction on the roller surface in the working state of the roller press, measures the reconstructed data, sends out an alarm in time when the condition that the roller surface is seriously damaged is found, and reminds maintenance personnel to repair the roller surface as early as possible, so that the operation rate of equipment is improved, the maintenance cost is reduced, the product quality is indirectly improved, the roller surface three-dimensional reconstruction method can adapt to severe working environment, and has strong anti-interference performance on factors such as smoke dust, noise and the like.

Drawings

FIG. 1 is a flow chart of the method of the present invention for on-line three-dimensional imaging of the roll surface of a roll press;

FIG. 2 is a schematic view of binocular image stereo rectification;

FIG. 3 is a schematic illustration of laser fringes after binocular threshold segmentation;

FIG. 4 is a schematic diagram of the laser line center extraction result;

fig. 5 is a principle diagram of binocular vision space point coordinate restoration based on parallax.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.

Referring to fig. 1-5, an online three-dimensional imaging method for a roll surface of a roll squeezer comprises the following steps:

s1: calibrating a system; the binocular vision system is calibrated through a specific calibration object, and the internal reference and the external reference of the binocular camera are solved.

Wherein, the internal reference matrix K of the binocular camera is only related to the internal structure of the camera parts and consists of the focal length f in the x and y directions_x，f_yAnd camera optical center coordinates u₀，v₀Determining that the external parameters of the binocular camera are a world coordinate system to camera coordinate system down-conversion matrix M₂It consists of a rotation matrix R of 3 × 3 and a three-dimensional translation vector t, the transformation between a spatial point P (Xw, Yw, Zw) and the projected point coordinates (u, v) in the image is shown as follows:

wherein s is a scale factor, M is a 3 x 4 projection matrix, when the binocular camera is calibrated, the optical center of one camera is usually used as the origin of a world coordinate system, and the rotation matrix and the translation vector from the other camera pose to the camera pose are external parameters of the binocular vision system.

S2: reconstructing local point cloud; and recovering the three-dimensional point cloud of the roller surface in the single line laser irradiation area through the roller surface pictures irradiated by the pair of line lasers.

S3: point cloud splicing; the binocular camera moves left and right on the lead screw guide rail at a specified speed, the roller surface images are shot at a certain frame rate and are opposite, three-dimensional coordinates of single line laser are recovered for each pair of images, a plurality of line point clouds are spliced into a surface, and then, when reconstruction algorithm is carried out, the roller surface rotates at a certain angular speed, and a plurality of circles of scanning data are spliced into complete roller surface data.

The measuring system only calibrates the camera once, the calibrated internal and external parameters are kept unchanged, and because the position relation between the camera and the world coordinate system is relatively fixed, the position of the world coordinate system is continuously changed, and each optical strip is obtained under different world coordinate systems, the three-dimensional information of the whole measured object needs to be reflected, and the data under the different coordinate systems need to be converted into the same world coordinate system. In the measuring process, the position of the marking point is kept unchanged all the time, so that the conversion relation between each world coordinate system and the world coordinate system at the scanning initial position can be calculated according to a series of three-dimensional coordinates of the marking point in different world coordinate systems.

In S2, the local point cloud reconstruction step further includes the following steps:

s21: correcting distortion; due to the influence of the shape of the lens of the camera and the assembly error inside the camera, the binocular camera can generate certain radial distortion and tangential distortion in the imaging process, so that the image is corrected by distortion parameters obtained by calibrating the camera, and the image completely conforms to the imaging geometric model of the camera.

Because the linear camera model cannot accurately describe the imaging geometric relationship due to defects of the camera lens in the manufacturing process, positioning errors in the assembling process and the like, the linear camera model needs to be subjected to distortion correction after image acquisition, the invention mainly corrects the radial distortion, and the correction formula can be expressed by the following formula:

wherein(X_d,Y_d) Actual image plane image coordinates, k, shifted for lens distortion₁,k₂,k₃The distortion coefficient of the lens can be obtained when the camera is calibrated.

S22: performing three-dimensional correction; the premise of recovering the three-dimensional point cloud of the roller surface according to the parallax is that the two imaging planes of the binocular camera are completely aligned in a line mode, and due to errors existing in the hardware installation process, the absolute line alignment of the two camera planes is required to be achieved through a stereo correction algorithm.

The invention adopts a Bouguet method to carry out epipolar rectification, the rectification process is shown in figure 4, according to the epipolar geometrical relationship, a space point and the optical centers of a left camera and a right camera form a plane which is called as an epipolar plane, the intersection line of the plane and an image plane is called as an epipolar line, and the image plane in actual conditions is like pi_R,π_TAs shown, a certain included angle exists, after the stereo correction, the two image planes are completely parallel, and the corresponding polar lines are aligned.

S23: extracting a laser central line; the laser stripes have a certain width in the left and right images, and in order to improve the precision of the measurement system, the central line pixel coordinates (sub-pixel level) of the line laser stripes need to be accurately extracted.

The method comprises the following steps of firstly carrying out denoising and filtering on a light bar image, removing isolated noise in a target and a background in the image, and simultaneously well keeping the detail information of laser stripes, processing the laser stripe image by adopting value filtering, sequencing pixels in a neighborhood according to gray levels, and then selecting a group of intermediate values as output pixel values, wherein the following formula is as follows:

g(m,n)＝Median_A{f_A(m+k,n+l),(k,l)∈R} (3)

in the formula of Median_AThe algorithm is a middle value of the window A, can eliminate isolated noise points, reduces the influence of smoke dust in a working environment on an image, then performs self-adaptive threshold segmentation on the image, extracts the outline positions of laser stripes and mark points in the image, finally extracts the center coordinates of the mark points and refines the light stripes, removes edge pixel points of the outline with a certain width successively, and obtains a center skeleton of a single pixel of the light stripe (as shown in figure 4) under the condition of not influencing the connectivity of the light stripe.

S24: stereo matching; the line alignment of the left image and the right image can be realized according to the epipolar rectification, the line laser is vertical, and the matching of the corresponding positions of the two images can be realized by searching the position of the laser central line of the two images in each line.

Step S25: calculating three-dimensional coordinates; and after the stereo matching is finished, calculating the parallax corresponding to the laser centers of the two images, and then calculating the coordinates of the space points of the line laser irradiation area according to the triangulation principle.

The schematic diagram of binocular stereo vision is shown in fig. 7, and according to similarity of triangles, the coordinates of the obtained space points are as follows:

wherein B is a baseline distance, which is the distance between the optical centers of the two cameras, and D ═ x_T-x_RAnd is the disparity between the corresponding matched feature points.

In S3, the implementation of the point cloud splicing part needs to fix four landmark points in the view of the scene, each pair of images can calculate the spatial coordinates of the four landmark points, and the point cloud is converted into a uniform world coordinate system according to the four known spatial coordinates to implement the splicing of the line point cloud.

The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.