Multi-sensor furnace lining thickness measuring method for intermediate frequency furnace

Abstract

A multi-sensor furnace lining thickness measuring method for an intermediate frequency furnace adopts a visual algorithm, is realized based on a structured light camera and a laser radar, and comprises the following steps: firstly, calibrating an RGB camera and an infrared camera to obtain respective internal and external parameter matrixes; secondly, aligning the RGB camera with the infrared camera to obtain three-dimensional information; recognizing a reference target through a target detection model, extracting a plurality of groups of three-dimensional feature points in a target area, and calculating to obtain a structured light camera transformation matrix, namely, realizing pose estimation; joint calibration is carried out on the structured light camera and the laser radar; and finally, enabling a camera pose estimation result and a joint calibration result to act on laser radar data under multiple times of measurement, and enabling the two times of radar data to be in the same coordinate system, thereby realizing thickness measurement and monitoring of the furnace lining. According to the invention, under the condition that the furnace lining is not damaged, no harm is caused to the environment, the influence of the environment is small, and meanwhile, the thickness of the furnace lining can be rapidly measured and monitored in a large-scale and high-precision manner.

Description

Technical Field

[0001] The present invention relates to the fields of computer vision and artificial intelligence, and in particular to a multi-sensor lining thickness measurement method for a medium frequency furnace. Background Art

[0002] The change of furnace lining thickness and the detection of furnace lining defects are closely related to the stability and safe production of the furnace. Therefore, the development and application of furnace lining erosion monitoring technology are of great significance to metallurgical industry personnel and related researchers. In recent years, through the continuous development and innovation of domestic and foreign researchers, certain results have been achieved in furnace lining thickness monitoring technology. There are many types of furnace lining thickness measurement technologies, and the principles are also different. So far, they can be mainly divided into two categories: contact and non-contact. Contact thickness measurement mainly includes capacitance thickness measurement, resistance thickness measurement, thermocouple thickness measurement and ultrasonic thickness measurement. However, these methods require that special sensors be placed inside the furnace lining, which will cause certain damage to the furnace lining, and once damage occurs, it is difficult to repair. At the same time, the measurement range is fixed and inflexible, and the accuracy is low; and non-contact thickness measurement mainly includes radiographic thickness measurement and infrared thermal imaging thickness measurement. Although these methods are non-destructive measurements, radiographic thickness measurement will cause harm to the human body and the environment, and infrared thermal imaging thickness measurement is susceptible to environmental interference. Therefore, there are areas that need to be improved in the current furnace lining thickness detection methods. Summary of the invention

[0003] In order to overcome the shortcomings of the above-mentioned furnace lining thickness measurement method, the present invention provides a multi-sensor furnace lining thickness measurement method for medium frequency furnaces. By using a structured light camera and a laser radar, the furnace lining thickness can be measured quickly, over a large range, and with high precision, without damaging the furnace lining and with little environmental impact.

[0004] The technical solution adopted by the present invention to solve its technical problem is:

[0005] A multi-sensor lining thickness measurement method for a medium frequency furnace; the method is implemented based on a structured light camera and a laser radar, the structured light camera is composed of an RGB camera and an infrared camera, and the laser radar selects a single-line laser radar; the method comprises the following steps:

[0006] Step 1) calibrate the RGB camera and the infrared camera to obtain their respective internal and external parameter matrices, which represent the coordinate transformation information inside the camera and between cameras;

[0007] Step 2) The image data collected by the RGB camera and the infrared camera are not in the same coordinate system, so the color image and the depth image are aligned to achieve the unification of the image coordinates and the acquisition of the three-dimensional coordinates of the pixel points;

[0008] Step 3) Using deep learning methods to identify the specific area of ​​the target in the two previous and next benchmark target images taken by the structured light camera, extract multiple groups of three-dimensional feature points in the area and match them, calculate the front and back pose transformation matrix of the structured light camera, and realize camera pose estimation;

[0009] Step 4) Jointly calibrate the structured light camera and the lidar to obtain the position transformation relationship between the two sensors and unify the radar data into the camera coordinate system;

[0010] Step 5) Using the camera pose estimation results, the radar data collected by the two laser radars are unified into the same camera coordinate system for comparison to realize the thickness of the furnace lining and its changes before and after.

[0011] Furthermore, in the step 1), a checkerboard calibration plate is selected and stably placed within the field of view of the structured light camera, and an infrared fill light is placed at a suitable distance to illuminate the checkerboard calibration plate, and the infrared emitter of the infrared camera is completely blocked with black tape; a plurality of pictures of the checkerboard calibration plate at different angles are taken with the structured light camera; the pictures are calibrated to obtain the internal and external parameters of the RGB camera and the infrared camera.

[0012] Furthermore, in step 2), it is assumed that there is a point in the space. Since the RGB camera and the infrared camera are at different spatial positions, the coordinates of the point in the RGB camera coordinate system and the infrared camera coordinate system are different; according to the pinhole imaging model, the projection relationship and the intrinsic parameters of each camera, the pixel coordinates of the point in the RGB camera and the infrared camera can be obtained; then, according to the transformation relationship between the two coordinate systems, the conversion from the pixel coordinates of the infrared camera to the pixel coordinates of the RGB camera can be realized; and when calculating the transformation matrix connecting the two coordinate systems, the external parameter matrices of the RGB camera and the infrared camera are used, and then the transformation matrix connecting the two camera coordinate systems is calculated according to the camera model transformation relationship formula; finally, the color image and the depth image are aligned according to the internal and external parameters and the transformation matrix relationship.

[0013] Furthermore, in the step 2), the alignment operation can only align the color image to the depth image, because the color image does not contain depth information and cannot realize the transformation from pixel coordinates to point cloud coordinates. Therefore, the depth image is needed to align to the color image, because the depth image contains depth information to achieve alignment.

[0014] Furthermore, in the step 3), the front and rear two benchmark target images taken by the camera are first processed by steps 1) and 2) to obtain the color and depth information of the images; the front and rear two images are then input into the benchmark target object detection model to obtain accurate pixel coordinates of the region of interest (benchmark target region); feature points are extracted in the region of interest, and the Euclidean distance between the feature descriptor vectors corresponding to the feature points is used as a metric for determining whether the feature points are similar; the extracted feature points are matched to form a feature correspondence relationship; the mismatched point deletion operation is performed on these matching point pairs, and the matching point pairs with larger errors are deleted. The processed matching point pairs are roughly aligned using the ICP (Iterative Closest Point) method, and a transformation matrix is ​​calculated based on the feature point pairs with known pairing relationships, and this result is used as the initial value in the iterative operation; the transformation matrix value calculated by the ICP rough alignment is used as the initial value, and the original feature point set is iteratively calculated, the optimal corresponding point pair is selected under the current pose transformation, and the optimal pose transformation is calculated; then the transformation is applied, and the optimal corresponding point pair is found again, and a new optimal pose transformation is calculated, until the convergence accuracy requirement of the correct alignment is met or the maximum number of iterations is reached, and the final pose transformation matrix is ​​obtained.

[0015] Furthermore, in the step 4), since the actual spatial positions of the structured light camera and the laser radar do not overlap, the data obtained are not in the same coordinate system, and the corresponding relationship between the same features recognized by the two sensors is used for joint calibration to obtain a transformation matrix between the two; the structured light camera calculates the coordinates of the calibration plate plane in the camera coordinate system by identifying the characteristic corner points on the checkerboard calibration plate; the laser radar laser beam is hit on the calibration plate, and the coordinates of the laser point in the laser coordinate system and the coordinates in the camera coordinate system are used to construct a plane constraint to solve the external parameters; finally, the nonlinear least squares method is used to solve and calculate the transformation relationship between the structured light camera and the laser radar, thereby realizing the conversion of the radar data into the camera coordinate system;

[0016] Furthermore, in step 5), since the positions of the structured light camera and the laser radar remain relatively fixed, the structured light camera and the laser radar only need to be calibrated once to obtain the transformation matrix between the two; before the laser radar collects the furnace lining data, the structured light camera is required to take a picture of the reference target, and the camera posture transformation matrix is ​​calculated based on the two pictures before and after;

[0017] Furthermore, in step 5), when the laser radar undergoes rotational and translational changes compared to the position of the last acquisition, the two radar data collected are not in the same coordinate system, so the data have no meaning for comparison. Therefore, after the laser radar acquisition is completed, the two radar data of the furnace lining collected can be unified into the same coordinate system after going through steps 1) to 4), thereby realizing radar data comparison and finally obtaining the furnace lining thickness and its changes.

[0018] The beneficial effects of the present invention are mainly manifested in that: the present invention uses a multi-sensor fusion method of a structured light camera and a lidar, and utilizes a visual algorithm to convert the furnace lining data collected twice before and after into the same coordinate system, thereby realizing the thickness measurement of the furnace lining and its changes; compared with the existing method, the method uses a visual method to fuse multi-sensor information to achieve the unification of the coordinate system of the collected data, and ultimately can quickly, over a large range, and with high precision measure and monitor the thickness of the furnace lining without damaging the furnace lining, causing no harm to the environment, and being less affected by the environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of equipment involved in a multi-sensor lining thickness measurement method for a medium frequency furnace according to the present invention;

[0020] Figure 2 This is an overall scheme framework diagram of a multi-sensor lining thickness measurement method for a medium frequency furnace according to the present invention;

[0021] Figure 3 It is a color image and depth image alignment framework diagram of a multi-sensor lining thickness measurement method for a medium frequency furnace according to the present invention;

[0022] Figure 4 It is a camera pose estimation framework diagram of a multi-sensor lining thickness measurement method for a medium frequency furnace according to the present invention;

[0023] Figure 5 It is a joint calibration framework diagram of a multi-sensor lining thickness measurement method for a medium frequency furnace according to the present invention;

[0024] Figure 6 6a is a schematic diagram of the radar data of the furnace lining twice before and after, and 6b is a schematic diagram of the radar data of the furnace lining twice after processing;

[0025] Figure 7 This is a diagram of experimental results of a multi-sensor lining thickness measurement method for a medium frequency furnace described in the present invention. DETAILED DESCRIPTION

[0026] In order to make the technical solution of the present invention clearer, the technical solution provided by the present invention will be described in detail in conjunction with specific embodiments, and the present invention will be further described in conjunction with the accompanying drawings. It can be understood that the specific embodiments described herein are only partial embodiments of the present invention, which are used to explain the present invention, rather than to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] Example

[0028] Figure 1 The present invention is a schematic diagram of equipment involved in a multi-sensor lining thickness measurement method for a medium frequency furnace according to an embodiment of the present invention, including a structured light camera and a laser radar. The structured light camera is used to photograph a reference target to achieve camera pose estimation; the laser radar is used to perform joint calibration with the structured light camera;

[0029] The overall solution framework of this example is as follows Figure 2 As shown, the overall solution includes the following steps:

[0030] Step 1) calibrate the RGB camera and the infrared camera to obtain their respective internal and external parameter matrices; select a checkerboard calibration plate and place it stably within the field of view of the structured light camera, place the infrared fill light at a suitable distance to illuminate the checkerboard calibration plate, and use black tape to completely block the infrared emitter of the infrared camera; use the structured light camera to take multiple pictures of the checkerboard calibration plate at different angles; calibrate the pictures to obtain the internal and external parameters of the RGB camera and the infrared camera.

[0031] Step 2) The RGB camera and the infrared camera are not in the same coordinate system, so we first need to align the color image and the depth image to obtain the three-dimensional information of the pixels in the image. Figure 3 shown.

[0032] Let P ir is the spatial coordinate of a point under the infrared camera coordinates, p ir is the projection coordinate of the point on the image plane, H ir is the internal parameter matrix of the infrared camera. According to the pinhole imaging model, they satisfy the following relationship:

[0033]

[0034] Suppose P rgb is the spatial coordinate of the same point in RGB camera coordinates, p rgb is the projection coordinate of the point on the RGB image plane, H rgb is the intrinsic parameter matrix of the RGB camera; since the coordinates of the infrared camera are different from those of the RGB camera, they need to be connected by a rotation and translation transformation, that is:

[0035] P rgb =RP ir +T (2)

[0036] Where R is the rotation matrix and T is the translation vector. Finally, use H rgb P rgb Projection can get the RGB pixel coordinates corresponding to the point:

[0037] p rgb =H rgb P rgb (3)

[0038] Therefore, to calculate the rotation matrix and translation vector that connect the two coordinate systems, the camera's extrinsic parameters are required;

[0039] The external parameter matrix is ​​actually composed of a rotation matrix R ir (R rgb ) and the translation vector T ir (T rgb ), which means transforming a point P in the global coordinate system to the camera coordinate system. The infrared camera and RGB camera are transformed respectively, and there are the following relationships:

[0040]

[0041] P ir =R ir P+T ir P in P ir , R ir 、T ir Express and substitute into P rgb =R rgb P+T rgb

[0042] Available Combined with formula (2), we get:

[0043]

[0044] Therefore, we only need to obtain the external parameter matrix of the checkerboard calibration plate relative to the infrared camera and the RGB camera in the same scene to calculate the transformation matrix that connects the coordinate systems of the two cameras; after obtaining the transformation matrix of the RGB camera and the infrared camera, the color image and the depth image are aligned. Assume that the pixel coordinates of a point P in space in the depth image taken by the infrared camera are (U ir , V ir ), the point cloud coordinates of P in the infrared camera can be calculated according to the inverse projection relationship of the pinhole imaging model:

[0045]

[0046] Where K ir is the intrinsic parameter matrix of the infrared camera, and Z is the value of the pixel in the depth image, that is, the depth value; then according to the transformation matrix relationship from the infrared camera to the RGB camera, the point cloud coordinates of the point in the RGB camera are obtained:

[0047]

[0048] Among them, R and T are the rotation matrix and translation matrix in the transformation matrix from infrared camera to RGB camera; finally, the point cloud coordinates are projected to pixel coordinates according to the intrinsic parameter matrix of RGB camera:

[0049]

[0050] Where K rgb is the intrinsic parameter matrix of the RGB camera; finally, the depth image and the color image are aligned to obtain the three-dimensional information of the point.

[0051] Step 3) Identify the reference target through deep learning methods, and calculate the structured light camera transformation matrix by proposing multiple groups of three-dimensional feature points in the target area, thus realizing pose estimation. Figure 4 shown.

[0052] The structured light camera takes two pictures of the benchmark target, and does not directly extract feature points in the entire picture. Instead, the feature point extraction area is determined near the benchmark target area. The two pictures are input into the benchmark target object detection model to obtain accurate pixel coordinates of the region of interest, that is, the position range of the benchmark target in the picture; feature points are extracted in the region of interest, and the Euclidean distance between the feature descriptor vectors corresponding to the feature points is used as a metric to determine whether the feature points are similar; it is assumed that in the two collected pictures: the previous picture is image 1, and the next picture is image 2; the extracted feature points are matched to form a corresponding relationship; among them, the feature matching adopts FLANN fast nearest neighbor search; each feature point P in image 1 i In image 2, the two feature points P with the closest descriptor vector distance i1 , P i2 , P i With P i1 , P i2 The distance is denoted as d(P i1 ), d(P i2 ), given a threshold r, if If true, then it can be considered that the feature point P in image 2 is the same as that in image 1 i Descriptor vector is closest to point P i1 is the feature point P in image 1 i; perform mismatching point deletion on these matching point pairs, and use RANSAC to delete the matching point pairs with large errors; after obtaining the matching point pairs after deleting the mismatching, combine the depth information after the color image and the depth image are aligned, convert the pixel coordinates into 3D point cloud coordinates, and use the ICP coarse registration method for the matching point pairs after point cloud processing, that is, calculate the feature point pairs with known pairing relationships to obtain the pose transformation of the two pictures before and after:

[0053] This is minimized using the iterative closest point algorithm:

[0054]

[0055] P i (1) is the feature point in image 1 in the matching point pair, P i (2) For the feature points in the corresponding image 2, R, t is solved by singular value decomposition: Calculate the centroid: Calculate the centroid of the two groups of point clouds for centering operation; Center the data: Subtract the centroid of the point cloud data to translate it to the origin; Calculate the covariance matrix: Use the centered point cloud data to calculate the covariance matrix; Singular value decomposition: Perform singular value decomposition on the covariance matrix; Calculate the rotation matrix: Extract the rotation matrix from the singular value decomposition; Calculate the translation vector: Use the rotation matrix to map the centroid of one point cloud to the centroid of another point cloud to calculate the translation vector; Get the transformation matrix, and use this transformation matrix as the initial value in the iterative operation.

[0056] The transformation matrix value obtained by ICP rough registration is used as the initial value, and the original feature point set is iteratively calculated. The optimal corresponding point pair is selected under the current pose transformation, and the optimal pose transformation is calculated. Then the transformation is applied, and the optimal corresponding point pair is found again, and a new optimal pose transformation is calculated until the convergence accuracy requirement of correct registration is met or the maximum number of iterations is reached; the final camera pose transformation matrix is ​​obtained.

[0057] Step 4) Joint calibration of structured light camera and LiDAR. The overall framework is as follows: Figure 5 As shown. Since the positions of the structured light camera and the laser radar do not overlap, the data obtained are not in the same coordinate system. The corresponding relationship between the same features recognized by the two sensors is used for joint calibration to obtain the external parameter matrix between the two. The structured light camera calculates the coordinates of the calibration plate plane in the camera coordinate system by identifying the characteristic corner points on the checkerboard calibration plate. The laser beam hits the calibration plate, and the coordinates of the laser point in the laser coordinate system and the coordinates in the camera coordinate system are used to construct plane constraints to solve the external parameters.

[0058]

[0059] Among them, vector n is the three-dimensional normal vector of the label plane, and d is the origin of the camera coordinate system O c Distance to the label plane, point P in laser coordinate system O l The coordinates below are P l , in the camera coordinate system O c The coordinates below are P c , point P from the laser coordinate system O l Transform to camera coordinate system O c For: P c =R T (P l -t), substitute P l =(x, y, 0) T ,have to:

[0060]

[0061] The matrix H is treated as a new unknown quantity and a nonlinear least squares solution is performed to calculate the rotation matrix and translation vector between the structured light camera and the lidar.

[0062] Step 5) When actually collecting radar data of the furnace lining, it cannot be guaranteed that the two devices are in the same position. Therefore, the two collected radar data need to be processed and placed in the same camera coordinate system for comparison. Therefore, each time the collection is performed, it is necessary to take a picture of the benchmark target to obtain the camera pose change matrix of the two times; for the two collected radar data, the radar data is unified to the corresponding camera coordinate system using the joint calibration results of the camera and the lidar; then the latter radar data is processed by the camera pose estimation matrix and unified to the previous camera coordinate system, so that the radar data of the two furnace linings are in the same camera coordinate system; finally, the lidar data of the two furnace linings are experimentally compared to realize the thickness measurement of the furnace lining and its changes.

[0063] like Figure 6 As shown in FIG6a, the radar data of the furnace lining are placed in a unified coordinate system, so the data cannot be compared. After the above steps, as shown in FIG6b, the radar data of the furnace lining are placed in a unified coordinate system, so the data are meaningful for comparison.

[0064] The comparison results are as follows Figure 7 As shown, the overall lining data is divided into 10 areas, and the changes in the front and rear radar data of each area, that is, the changes in the lining thickness, are obtained.

[0065] Finally, the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A multi-sensor lining thickness measurement method for medium frequency furnace, characterized by: The thickness measurement method is implemented based on a structured light camera and a laser radar. The structured light camera includes an RGB camera and an infrared camera. The method includes the following steps: Step 1) calibrate the RGB camera and the infrared camera to obtain their respective internal and external parameter matrices, which represent the coordinate transformation information inside the camera and between cameras; Step 2) Align the color image and depth image collected by the RGB camera and the infrared camera to unify the coordinates of the image and obtain the three-dimensional coordinates of the pixel points; Step 3) For the two images of the benchmark target taken by the structured light camera, the specific area of ​​the target in the image is identified by using a deep learning method, multiple groups of three-dimensional feature points in the area are extracted and matched, and the front and back pose transformation matrix of the structured light camera is calculated to realize camera pose estimation; Step 4) Jointly calibrate the structured light camera and the lidar to obtain the position transformation relationship between the two sensors and unify the radar data into the camera coordinate system; Step 5) Using the camera pose estimation results, the radar data collected by the two laser radars are unified into the same camera coordinate system for comparison to realize the thickness of the furnace lining and its changes before and after.

2. A multi-sensor lining thickness measurement method for a medium frequency furnace as claimed in claim 1, characterized in that: The specific process of step 1) is as follows: 1.1) Select a checkerboard calibration plate and place it stably within the field of view of the structured light camera. Set the infrared fill light to illuminate the checkerboard calibration plate at a set distance and completely block the infrared emitter of the infrared camera with black tape. 1.2) Use a structured light camera to take multiple pictures of the checkerboard calibration plate at different angles; 1.3) Calibrate the image to obtain the internal and external parameters of the RGB camera and infrared camera.

3. A multi-sensor lining thickness measurement method for a medium frequency furnace as claimed in claim 1, characterized in that: The specific process of step 2) is as follows: 2.1) Assume that there is a point in space. Since the RGB camera and the infrared camera are in different spatial positions, the coordinates of this point in the RGB camera coordinate system and the infrared camera coordinate system are different. 2.2) According to the pinhole imaging model, projection relationship and internal parameters of each camera, the pixel coordinates of the point under the RGB camera and infrared camera are obtained; 2.3) According to the transformation relationship between the two coordinate systems, the pixel coordinates of the infrared camera are converted into the pixel coordinates of the RGB camera; 2.4) When calculating the transformation matrix connecting the two coordinate systems, the external parameter matrices of the RGB camera and the infrared camera are used to calculate the transformation matrix connecting the two camera coordinate systems according to the camera model conversion relationship formula; 2.5) Finally, the color image and the depth image are aligned according to the internal and external parameters and the transformation matrix relationship.

4. A multi-sensor lining thickness measurement method for a medium frequency furnace as claimed in claim 1, characterized in that: The specific process of step 3) is as follows: 3.1) The two images of the reference target taken by the camera are processed in steps 1) and 2) to obtain the color and depth information of the images; 3.2) Input the previous and next pictures into the benchmark target object detection model to obtain the accurate pixel coordinates of the region of interest; 3.3) Extract feature points in the region of interest, and use the Euclidean distance between the feature descriptor vectors corresponding to the feature points as a metric to determine whether the feature points are similar; 3.4) Match the extracted feature points to form feature correspondence; Perform mismatch point deletion on these matching point pairs, and delete the matching points with larger errors in the matching point pairs; 3.5) The processed matching point pairs are subjected to the iterative closest point ICP coarse registration method, and a transformation matrix is ​​calculated based on the feature points of the known pairing relationship, and this result is used as the initial value in the iterative operation; 3.6) The transformation matrix value obtained by ICP rough registration is used as the initial value, and the original feature point set is iteratively calculated. The optimal corresponding point pair is selected under the current pose transformation, and the optimal pose transformation is calculated; 3.7) Apply the pose transformation, find the optimal corresponding point pair, and calculate the new optimal pose transformation until the convergence accuracy requirement for correct registration is met or the maximum number of iterations is reached to obtain the final pose transformation matrix.

5. A multi-sensor lining thickness measurement method for a medium frequency furnace as claimed in claim 1, characterized in that: In step 4), since the actual spatial positions of the structured light camera and the laser radar do not overlap, the data obtained are not in the same coordinate system; the specific process is as follows: 4.1) Use the correspondence between the same features identified by the two sensors to perform joint calibration and obtain the transformation matrix between the two; 4.2) The structured light camera calculates the coordinates of the calibration plate plane in the camera coordinate system by identifying the characteristic corner points on the checkerboard calibration plate; 4.3) The laser radar laser beam hits the calibration plate, and the coordinates of the laser point in the laser coordinate system and the coordinates in the camera coordinate system are used to construct plane constraints to solve the external parameters; 4.4) Finally, nonlinear least squares is used to solve and calculate the transformation relationship between the structured light camera and the lidar, so as to realize the conversion of radar data into the camera coordinate system.

6. A multi-sensor lining thickness measurement method for a medium frequency furnace as claimed in claim 1, characterized in that: In the step 5), based on the relative position of the structured light camera and the laser radar, the structured light camera and the laser radar only need to be jointly calibrated once to obtain the transformation matrix between the two; before the laser radar collects the furnace lining data, the structured light camera is required to take a picture of the reference target, and the camera posture transformation matrix is ​​calculated based on the two pictures.

Images