A coil shape detection device and method based on three-dimensional point cloud of hot-rolled steel coil end surface

Abstract

The present invention discloses a coil shape detection device and method based on a three-dimensional point cloud of the end face of a hot-rolled steel coil. The device includes a three-dimensional laser radar, a mounting bracket, an air cooler, and a server. The three-dimensional laser radar and the air cooler are both mounted within the mounting bracket, which is placed on the operating side at a preset distance from the stationary position of the steel coil when it comes off the production line. The three-dimensional laser radar is connected to the server, and the air cooler is used to dissipate heat from the three-dimensional laser radar. When the steel coil reaches a designated position, the three-dimensional laser radar begins scanning the end face of the steel coil on the operating side, obtaining a point cloud of the steel coil end face. The obtained point cloud of the steel coil end face is transmitted to the server, which extracts and identifies the steel coil features based on the point cloud of the steel coil end face to achieve coil shape detection. The technical solution of the present invention can achieve online detection and automatic control of the coil shape of the hot-rolled steel coil, improving the intelligent level of warehouse logistics.

Description

Technical Field

[0001] The present invention relates to the field of intelligent metallurgical detection technology, and in particular to a coil shape detection device and method based on a three-dimensional point cloud of a hot-rolled steel coil end face. Background Art

[0002] Coil shape quality is the most intuitive product attribute during hot-rolled coil production, drawing particular attention from users. Coil shape is the most crucial quality indicator in the hot-rolling coiling area. Coils with poor coil shape during coiling at a mill's coiler can impact subsequent production and use. Therefore, accurately assessing hot-rolled coil shape quality and reducing coil shape defects are crucial for improving economic efficiency and meeting cold-rolling raw material requirements. Coil shape inspection is a crucial step in the process of coil roll-off and storage. This process inspects coil shape upon entering the storage area after hot-rolled strip coils have been rolled off the production line. This process provides scheduling information for overhead cranes, ensuring coil shape quality for incoming and outgoing coils.

[0003] Currently, the field of hot-rolled steel coil shape inspection mainly relies on traditional manual methods. Since the hot-rolled steel coil is still at a relatively high temperature when it comes off the line and the on-site environment is relatively complex, it may pose certain safety hazards to workers. In addition, relying on manual judgment of the coil shape may lead to misjudgment and omission due to different evaluation standards of each person. Summary of the Invention

[0004] The present invention provides a coil shape detection device and method based on the three-dimensional point cloud of the end face of a hot-rolled steel coil, so as to solve the technical problems that the traditional manual coil shape detection method may cause certain safety hazards to workers, and the manual judgment of the coil shape may lead to misjudgment and missed judgment.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] On the one hand, the present invention provides a coil shape detection device based on a three-dimensional point cloud of a hot-rolled steel coil end face, comprising: a three-dimensional laser radar, a mounting bracket, an air cooler, and a server; wherein,

[0007] The three-dimensional laser radar and the air cooler are both installed in the mounting bracket, and the mounting bracket is placed on the operating side at a preset distance from the static position when the steel coil is off the line; the three-dimensional laser radar is connected to the server; the air cooler is used to dissipate heat from the three-dimensional laser radar;

[0008] When the steel coil arrives at the specified position, the three-dimensional laser radar starts to scan the end face of the steel coil operating side, obtains the steel coil end face point cloud, and transmits the obtained steel coil end face point cloud to the server. The server extracts and identifies the features of the steel coil based on the steel coil end face point cloud to realize coil shape detection.

[0009] Furthermore, the height of the mounting bracket is adjustable;

[0010] The mounting bracket includes: a fixed support fixed part, a support movable part, a rotating base plate, an upper shell and a conical panel; wherein, the three-dimensional laser radar is placed in the middle position of the upper shell, and its scanning surface faces outward toward the conical panel, and the air cooler is placed at the tail position of the upper shell for dissipating heat for the three-dimensional laser radar. The installation height of the three-dimensional laser radar is 30 to 50 cm lower than the height of the center line of the steel coil, and the yoz scanning plane of the three-dimensional laser radar is perpendicular to the center line direction of the steel coil.

[0011] Furthermore, the feature extraction and recognition of the steel coil is performed based on the point cloud of the steel coil end surface to realize coil shape detection, including:

[0012] Preprocessing the steel coil end surface point cloud obtained by the three-dimensional laser radar scanning;

[0013] Perform angle calibration on the pre-processed point cloud of the steel coil end face and obtain the point cloud after removing the inner ring of the steel coil;

[0014] Based on the calibrated point cloud, multiple radii on the end face of the steel coil are obtained through straight-through filtering, and the features of the point cloud radius are extracted to determine whether the hot-rolled steel coil is a flat coil and complete the coil type identification of the hot-rolled steel coil.

[0015] Furthermore, the point cloud of the steel coil end surface obtained by the 3D laser radar scanning is preprocessed, including:

[0016] According to the placement of the three-dimensional laser radar and the steel coil, the maximum radius R of the offline steel coil max The position of the mounting bracket from the steel coil is determined by removing points in the steel coil end face point cloud that are far away from the steel coil through straight-through filtering to obtain a first point cloud dataset; wherein the straight-through filtering range is shown in the following formula:

[0017]

[0018] Wherein, x, y, and z are the scanning directions of the three-dimensional laser radar respectively;

[0019] The first point cloud dataset is statistically filtered to remove points with an appearance probability less than 0.3174 to remove noise points and outliers, thereby obtaining a second point cloud dataset, as shown in the following formula:

[0020]

[0021] Among them, μ is the mean distance of the global points, σ is the standard deviation of the distance of the global points, d k is the average distance from the target point to its k neighboring points, if dk If the global standard deviation σ is less than 1, it is considered as a noise point and removed.

[0022] Furthermore, the pre-processed coil end face point cloud is angle-calibrated to obtain a point cloud after removing the inner ring of the coil, including:

[0023] The RANSAC plane fitting method is used to fit a plane to the second point cloud dataset, extract the plane point cloud dataset, and obtain the normal vector of the plane; the characteristic equation of the fitted plane is shown as follows:

[0024] A 11 +B 11 +C 11 +D 11 =0

[0025] performing indexing based on the points on the planar point cloud dataset, removing the points in the second point cloud dataset that are located in the planar point cloud dataset, to obtain a third point cloud dataset;

[0026] Performing cylindrical fitting on the third point cloud dataset to extract a cylindrical point cloud dataset;

[0027] performing indexing based on the points on the cylindrical point cloud dataset, removing the points in the second point cloud dataset that are located in the cylindrical point cloud dataset, to obtain a fourth point cloud dataset;

[0028] The fourth point cloud dataset and the cylindrical point cloud dataset are rotated according to the normal vector of the fitted plane so that the normal vector direction of the steel coil end surface is parallel to the x-axis direction of the three-dimensional laser radar scanning plane, thereby obtaining a calibrated fourth point cloud dataset; and the normal vector of the steel coil plane is perpendicular to the laser radar scanning plane, thereby obtaining a calibrated cylindrical point cloud dataset; wherein, the calibration matrix T is shown as follows:

[0029]

[0030]

[0031]

[0032] Among them, A 11 、B 11 、C 11 、D 11 are the four parameters of the plane equation of the steel coil end face.

[0033] Furthermore, multiple radii on the end face of the steel coil are obtained through direct filtering, and the features of the point cloud radius are extracted to determine whether the hot-rolled steel coil is a flat coil and complete the coil type identification of the hot-rolled steel coil, including:

[0034] Step 1: Based on the calibrated cylindrical point cloud dataset, index the point with the largest and smallest y-axis values on the cylinder, record the coordinate values of the indexed point cloud B1 = (x1, y1, z1) and B2 = (x2, y2, z2), and calculate the center coordinates of a point on the axis of the cylinder, as shown in the following formula;

[0035]

[0036] Step 2: Obtain i points within a range of 1 cm above and below z0 on the currently calibrated fourth point cloud dataset through straight-through filtering, and set all z-axis values of the obtained point cloud to 0, while keeping the x and y coordinate values unchanged, to obtain the fifth point cloud dataset;

[0037] Step 3: Index the fifth point cloud dataset and record the point C1 = (x3, y3, z3) with the smallest y coordinate among the points with y coordinate values greater than 0, and the point C2 = (x4, y4, z4) with the largest y coordinate among the points with y coordinate values less than 0. The value obtained by calculating y3-y4 is the inner diameter of the steel coil. If the calculated value of y3-y4 is less than 0.8D or greater than 1.2D, the current steel coil is determined to be a flat coil; where D is the inner radius of the standard coil.

[0038] Step 4: Obtain a point cloud radius dataset with y coordinate values greater than 0 through straight-through filtering;

[0039] Step 5: Rotate the calibrated fourth point cloud dataset by a certain angle to obtain a new calibrated fourth point cloud dataset, and re-execute steps 1 to 4 to obtain a new point cloud radius dataset.

[0040] Step 6: Iterate step 5 six times to obtain seven point cloud radius datasets. Sorting the seven point cloud radius datasets from large to small according to the y-axis value and removing the points with smaller x-axis values in the point cloud data with the same y-axis value to obtain the sorted point cloud radius dataset.

[0041] Step 7: Count the flat coil judgment results. If the flat coil is judged as flat coil more than twice, it is determined that the steel coil has a flat coil defect.

[0042] Step 8: Extract features of the radius corresponding to each point cloud radius data set to preliminarily determine the coil shape of the steel coil, and count the judgment results of each radius to obtain the final judgment result of the coil shape.

[0043] Furthermore, feature extraction is performed on the radius corresponding to each point cloud radius dataset to preliminarily determine the coil shape. The judgment results of each radius are statistically analyzed to obtain the final coil shape judgment result, including:

[0044] For each point cloud radius dataset, perform the following steps:

[0045] Detect the distance in the y-axis direction between two adjacent point clouds in the current point cloud radius dataset. If the distance between adjacent points is greater than 5mm, it is judged as loose roll;

[0046] Extract the point cloud of the inner 1 / 3 of the current point cloud radius dataset and calculate the maximum concave-convex value σ in the x-axis direction of the extracted point cloud max11 , the average value x of all points on the x-axis avg11 And the minimum value point in the x direction is away from x avg11 If the minimum value of the x direction point is away from x avg11 The distance is less than σ max11 / 2, it is judged to be a pyramidal roll; if the minimum value point in the x direction is away from x avg11 The distance is greater than σ max11 / 2, it is judged as an inner circle overflow volume;

[0047] Extract the point cloud of the outer 1 / 3 of the current point cloud radius dataset and calculate the point x with the minimum x value in the extracted point cloud. min12 The point x with the maximum value in the x direction max12 The distance x between point clouds max12 -x min12 , if the calculated x max12 -x min12 If the value is greater than the overflow standard of 10cm, it is judged as an outer circle overflow roll;

[0048] Extract the point cloud of the outer 1 / 3 of the current point cloud radius dataset, calculate the mean x-direction value of the extracted point cloud and calculate the variance δ. When δ1<δ<δ2, it is recorded as the outer circle interlaced volume, and when δ>δ2, it is recorded as the interlaced volume, where δ1 is the judgment threshold between the outer circle interlaced volume and the normal volume, and δ2 is the judgment threshold between the outer circle interlaced volume and the interlaced volume.

[0049] The judgment results corresponding to each radius are counted. If there are more than three radii that judge that the current steel coil has the same preset type of coil shape defect, it is finally judged that the current steel coil has the preset type of coil shape defect.

[0050] On the other hand, the present invention further provides a coil shape detection method based on a three-dimensional point cloud of a hot-rolled steel coil end surface, implemented by the coil shape detection device based on a three-dimensional point cloud of a hot-rolled steel coil end surface, comprising:

[0051] When the steel coil arrives at the designated position, a steel coil in-position signal is transmitted to the three-dimensional laser radar, which is then activated to scan the end face of the steel coil on the operating side to obtain a point cloud of the steel coil end face;

[0052] The point cloud of the steel coil end surface obtained by the three-dimensional laser radar scanning is transmitted to the server, and the server extracts and identifies the features of the steel coil based on the point cloud of the steel coil end surface to realize coil shape detection.

[0053] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0054] This invention captures the location information of hot-rolled steel coils, activates a three-dimensional laser radar to obtain a large amount of end surface point cloud data information of the hot-rolled steel coils, and uses a feature extraction algorithm to obtain the end surface point cloud features of the hot-rolled steel coils. This allows accurate identification of the coil shape information of the hot-rolled steel coils, determines the type of defects in the hot-rolled steel coils, and realizes online detection and automatic control of the coil shape of the hot-rolled steel coils, thereby improving the intelligent level of warehouse logistics. Specific advantages are as follows:

[0055] (1) Without any auxiliary calibration tools, the point cloud is calibrated by obtaining the characteristics of the target point cloud itself, thereby realizing the subsequent target detection process;

[0056] (2) The automatic online detection system can reduce on-site manual participation and greatly improve the safety of workers;

[0057] (3) Set a unified inspection standard to avoid misjudgment or omission of roll defects due to different judgment standards of different operators;

[0058] (4) The coil shape detection device of the present invention can perform information management on the collected data of the hot-rolled steel coils, and the stored production data can be used for statistical analysis of information such as the hot-rolled steel coil qualification rate and the rework rate.

[0059] (5) The roll shape detection device of the present invention has a simple and clear overall structure, is easy to install and maintain, has a strong algorithm robustness, and its accuracy meets the detection requirements of most production lines for roll shape defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0061] Figure 1 2 is a schematic structural diagram of a roll shape detection device provided by an embodiment of the present invention;

[0062] Figure 2 is a schematic diagram of a point cloud dataset A0 provided in an embodiment of the present invention;

[0063] Figure 3 This is the point cloud dataset A provided by the embodiment of the present invention. 平面 Schematic diagram of ′;

[0064] Figure 4 This is the point cloud dataset A provided by the embodiment of the present invention. 平面″Schematic diagram;

[0065] Figure 5 This is the point cloud dataset B provided by the embodiment of the present invention. 圆柱 Schematic diagram of ′;

[0066] Figure 6 is a schematic diagram of a point cloud dataset A2 provided by an embodiment of the present invention;

[0067] Figure 7 is a schematic diagram of a point cloud dataset A3 provided by an embodiment of the present invention;

[0068] Figure 8 is a schematic diagram of a point cloud dataset R0′ provided by an embodiment of the present invention;

[0069] Figure 9 Schematic diagram of a point cloud dataset R1′ provided in an embodiment of the present invention.

[0070] Description of reference numerals:

[0071] 1. 3D laser radar; 2. Mounting bracket; 3. Air cooler; 4. Server. DETAILED DESCRIPTION

[0072] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0073] First embodiment

[0074] This embodiment provides a coil shape detection device based on the three-dimensional point cloud of the end face of a hot-rolled steel coil. The device uses a three-dimensional laser radar to scan the end face of the offline steel coil to obtain three-dimensional point cloud data of the end face. The point cloud data obtained is used to calibrate and identify features of the point cloud, accurately identify the surface morphology of the steel coil, determine the defect type of the steel coil, realize online detection of the coil shape of the incoming steel coil, and improve the intelligence level of coil shape detection.

[0075] Specifically, if Figure 1 As shown, the device includes: a three-dimensional laser radar 1, a mounting bracket 2, an air cooler 3 and a server 4; wherein, the three-dimensional laser radar 1 and the air cooler 3 are both installed in the mounting bracket 2, and the mounting bracket 2 is placed on the operating side 4.5m away from the static position when the steel coil is off the line; the three-dimensional laser radar 1 is connected to the server 4; the air cooler 3 is used to dissipate heat for the three-dimensional laser radar 1.

[0076] When the steel coil arrives at the specified position, the three-dimensional laser radar 1 is activated and begins to scan the end face of the steel coil operating side to obtain the steel coil end face point cloud, and transmits the obtained steel coil end face point cloud to the server 4. The server 4 processes the steel coil end face point cloud, extracts and identifies the features of the steel coil, and realizes coil shape detection.

[0077] Specifically, in this embodiment, the height of the mounting bracket 2 is adjustable; it consists of a fixed support fixed part, a support movable part, a rotating base plate, an upper shell and a conical panel; wherein, the three-dimensional laser radar 1 is placed in the middle position of the upper shell, and its scanning surface faces outward toward the conical panel, and the air cooler 3 is placed at the tail position of the upper shell to dissipate heat for the three-dimensional laser radar 1. The installation height of the three-dimensional laser radar 1 is 30 to 50 cm lower than the height of the center line of the steel coil, and its yoz scanning plane is perpendicular to the center line direction of the steel coil.

[0078] Furthermore, the process of performing roll shape detection by the roll shape detection device of this embodiment is as follows:

[0079] S1, obtaining the point cloud of the steel coil end surface: When the steel coil is stationary on the off-line track and unloads to the position of the 3D laser radar 1, the steel coil in position signal is transmitted to the I / O interface of the 3D laser radar 1, and the 3D laser radar 1 is started to scan the end surface of the steel coil on the operating side to obtain the point cloud of the steel coil end surface;

[0080] S2, data preprocessing: The point cloud of the steel coil end surface obtained by the 3D laser radar 1 is transmitted to the server 4 via Ethernet, and the point cloud is preprocessed using a preset processing algorithm; the details are as follows:

[0081] S21, according to the placement position of the three-dimensional laser radar 1 and the steel coil, the maximum radius R of the offline steel coil max As well as the position of the mounting bracket 2 from the steel coil, remove the points far from the steel coil, and obtain the point cloud data set A0 through straight-through filtering. The straight-through filtering range is shown in formula (1):

[0082]

[0083] Wherein, x, y, and z are the scanning directions of the three-dimensional laser radar 1;

[0084] S22, perform statistical filtering on the point cloud dataset A0, delete noise points, and regard the point cloud distribution as a Gaussian distribution. If the average distance d from the target point to its k neighboring points is k If the global standard deviation σ is less than 1, it is considered as a noise point and removed. That is, points with an occurrence probability less than 0.3174 are removed to obtain the point cloud dataset A1 with outliers and noise points removed, as shown in Formula 2.

[0085]

[0086] Among them, μ is the mean distance of global points, and σ is the standard deviation of the distance of global points.

[0087] S3, Calibration: Perform angle calibration on the point cloud and obtain the end face point cloud after removing the inner ring of the steel coil; the details are as follows:

[0088] S31, use the RANSAC plane fitting method to fit the plane to the pre-processed steel coil end surface point cloud, and extract the point cloud A on the plane 平面 ′, and obtain the normal vector of the plane; the characteristic equation of the fitted plane is shown in formula (3):

[0089] A 11 +B 11 +C 11 +D 11 =0 (3)

[0090] Among them, A 11 、B 11 、C 11 、D 11 are the four parameters of the plane equation of the steel coil end face;

[0091] S32, based on point cloud dataset A 平面 ′, and remove the points in point cloud dataset A1 that are located in point cloud dataset A. 平面 ′, and obtain the point cloud dataset A after removing the plane 平面 ″;

[0092] S33, point cloud dataset A 平面 Perform cylindrical fitting and extract cylindrical point cloud dataset B 圆柱 ';

[0093] S34, based on point cloud dataset B 圆柱 ′, and remove the points in point cloud dataset A1 that are located in point cloud dataset B. 圆柱 ′, and obtain the coil end face point cloud dataset A2;

[0094] S35, calibrate the steel coil end face point cloud dataset A2 according to the parameters of the fitted plane point cloud, rotate it so that the normal vector direction of the steel coil end face is parallel to the x-axis direction of the scanning plane of the 3D laser radar 1, and obtain the calibrated steel coil end face point cloud dataset A3, so that the normal vector of the steel coil plane is perpendicular to the scanning plane of the laser radar, and calibrate the point cloud dataset B 圆柱 ' Perform the same calibration and obtain the calibrated cylindrical point cloud dataset B 圆柱 ; Among them, the calibration matrix T is shown in formula 4:

[0095]

[0096] S4, based on the calibrated point cloud, obtains multiple radii on the end face of the steel coil through straight-through filtering, extracts the features of the point cloud radius, determines whether the steel coil is a flat coil, and completes the coil type recognition; including:

[0097] S41, determine whether it is a flat coil: obtain two opposite radii on the end surface of the steel coil through straight-through filtering and determine whether it is a flat coil; the details are as follows:

[0098] S411, based on the obtained calibrated cylindrical point cloud dataset B 圆柱 , index the point with the largest y-axis value and the point with the smallest y-axis value on the cylinder, record the coordinate values of the indexed point cloud B1 = (x1, y1, z1) and B2 = (x2, y2, z2), and calculate the center coordinates of a point on the axis of the cylinder, as shown in formula (5);

[0099]

[0100] S412: Obtain i points with a height of z0±1 cm on the point cloud dataset A3 through straight-through filtering, and change all z-axis values of the obtained point cloud to 0, while keeping the x and y coordinate values unchanged, to obtain the point cloud dataset R0′, as shown in Formula 6.

[0101]

[0102] S413, index the point cloud dataset R0′ and record the point with the smallest y coordinate among the points with y coordinate values greater than 0, C1 = (x3, y3, z3), and the point with the largest y coordinate among the points with y coordinate values less than 0, C2 = (x4, y4, z4). The value obtained by calculating y3-y4 is the inner diameter of the steel coil. If the calculated value of y3-y4 is less than 0.8D or greater than 1.2D, the current steel coil is determined to be a flat coil; where D is the inner radius of the standard coil; as shown in formula (7);

[0103]

[0104] S414, obtaining a point cloud radius dataset R1 with a y coordinate value greater than 0 through straight-through filtering, as shown in formula (8);

[0105] y>0 (8)

[0106] S415: The point cloud dataset A3 is rotated six times, with the rotation angles being -Π / 6, -5Π / 6, -Π, -5Π / 4, -3Π / 2, and -7Π / 4, respectively. Steps S411 to S414 are recalculated after each rotation, resulting in a total of seven point cloud radius datasets R1′, R2′, R3′, R4′, R5′, R6′, and R7′ of the coil end surface.

[0107] S416 , sorting the point cloud radius datasets R1′, R2′, R3′, R4′, R5′, R6′, and R7′ from large to small according to the y-axis values and removing the points with smaller x-axis values in the point cloud data with the same y-axis value, to obtain the sorted point cloud datasets R1, R2, R3, R4, R5, R6, and R7;

[0108] S417, collecting statistics on the flat coil determination results. If the flat coil is determined to be flat for more than 2 times, the steel coil is determined to have a flat coil defect.

[0109] S42, determine the coil type: extract features of each radius to preliminarily determine the coil shape, calculate the determination results of each radius, and obtain the final determination result of the coil shape, as follows:

[0110] S421, detect the distance between two adjacent point clouds in the point cloud radius dataset R1 in the y-axis direction. If the distance between any adjacent points is greater than 5 mm, it is judged as loose coiling, as shown in formula (9);

[0111] |y i -y i+1 |>0.005 (9)

[0112] S422, extract the point cloud dataset R1 of the inner 1 / 3 of the point cloud radius dataset R 11 , calculate R 11 The maximum concave-convex amount σ in the x-axis direction max11 , if σ max11 Greater than 40mm, and the minimum value point x in the x direction min11 From point cloud dataset larger than R 11 The average value x of all points on the x-axis avg11 The distance is less than σ max11 / 2, it is judged to be a pyramidal roll; if it is greater than σ max11 / 2, it is judged as an inner circle overflow, as shown in formula (10),

[0113]

[0114] Among them, x max1 、x min1 is the maximum and minimum value of the x-axis in the point cloud dataset R1, x max11 、x min11 For point cloud dataset R 11 The maximum and minimum values along the x-axis;

[0115] S423, extract the point cloud dataset R of the outer 1 / 3 of the point cloud radius dataset R1 12 , calculate R 12 The point x with the minimum value in the x direction min12 The point x with the maximum value in the x direction max12 The distance x between point clouds max12 -x min12 If the calculated distance value is greater than the overflow standard of 10 cm, it is judged as an outer circle overflow roll; as shown in formula (11);

[0116] x max12 -xmin12 >0.010 (11)

[0117] S424, extracting the point cloud of the outer 1 / 3 of the current point cloud radius dataset, calculating the mean x-direction value of the extracted point cloud and calculating the variance δ, when δ1<δ<δ2, it is recorded as the outer circle interleaved volume, and when δ>δ2, it is recorded as the interleaved volume, where δ1 is the judgment threshold between the outer circle interleaved volume and the normal volume, and δ2 is the judgment threshold between the outer circle interleaved volume and the interleaved volume;

[0118] S425, repeating the judgment process of S421-S424 for each radius and recording the judgment result of each radius;

[0119] S426, counting the judgment results of each of the seven radii, if more than three radii judge that the steel coil has the same coil shape defect, then the steel coil is finally judged to have this defect.

[0120] In summary, the coil shape detection device of this embodiment scans the steel coil offline by a three-dimensional laser radar placed at the offline location of the steel coil, obtains a three-dimensional point cloud, and pre-processes the point cloud. The point cloud that is far away from the end face of the steel coil is filtered out by filtering, and the point cloud is processed to remove outliers and noise points. After pre-processing, the point cloud is plane fitted to extract plane points and obtain the normal vector of the plane. Then, the plane points are removed from the pre-processed point cloud, and the remaining point cloud is cylinder fitted to obtain the point cloud on the inner circle of the steel coil. The point cloud on the cylinder obtained by cylinder fitting is removed and the remaining point cloud is retained. Then, the pre-processed point cloud without the fitted cylinder and the fitted cylinder are rotated according to the normal vector calculated by plane fitting to complete the calibration. Finally, the radius of the pre-processed point cloud without the fitted cylinder and the calibrated cylindrical point cloud are taken according to the calibrated point cloud without the fitted cylinder and the calibrated cylindrical point cloud, and the characteristics of the point cloud radius are extracted to complete the coil shape recognition of the hot-rolled steel coil. Therefore, the use of laser radar to intelligently upgrade the detection of steel coil shape defects not only improves the efficiency and accuracy of coil shape defect detection and ensures the stability of steel coil quality, but also reduces the company's manpower and time costs as well as the occurrence of enterprise safety accidents. At the same time, the production data saved by this coil shape detection device can also be used for statistical analysis of information such as steel coil qualification rate and rework rate.

[0121] Second embodiment

[0122] This embodiment combines Figures 2 to 7 The process of roll shape detection by the roll shape detection device of the present invention is further explained with a practical application example and actual data, wherein the installation height of the three-dimensional laser radar is 40 cm lower than the height of the center line of the steel coil.

[0123] Specifically, the process of performing roll shape detection by the set roll shape detection device is as follows:

[0124] 1. When the steel coil stops on the off-line track and reaches the location of the 3D laser radar, the 3D laser radar is activated to scan the end face of the steel coil on the operating side to obtain a point cloud of the steel coil end face. The point cloud of the steel coil end face scanned by the 3D laser radar is then transmitted to the server via Ethernet.

[0125] 2. Preprocess the point cloud data using a straight-through filter and a statistical filter; including:

[0126] 2.1 According to the placement of the 3D laser radar and the steel coil, the maximum radius R of the offline steel coil max As well as the position of the mounting bracket from the steel coil, remove the points that are far away from the steel coil, and obtain the coarse filtered point cloud dataset A0 through straight-through filtering. The straight-through filtering range is shown in formula (1):

[0127]

[0128] Wherein, x, y, and z are the scanning directions of the three-dimensional laser radar 1; R max =1m.

[0129] 2.2 Set the statistical filter parameters based on the measurement accuracy of the point cloud, set the number of nearest neighbors for average distance estimation to 50, set the standard deviation multiplier for distance threshold calculation to 4.0, perform statistical filtering on the point cloud to remove outliers and noise points, and obtain the following: Figure 2 The preprocessed point cloud dataset A1 is shown in Formula 2;

[0130]

[0131] Among them, μ is the mean distance of global points, and σ is the standard deviation of the distance of global points.

[0132] 3. Use the RANSAC plane fitting method to fit the plane of the pre-processed steel coil end point cloud and extract the following Figure 3 Point cloud A on the plane shown 平面 ′, and obtain the normal vector of the plane; the characteristic equation of the fitted plane is shown in formula (3):

[0133] 0.99833x-0.00300y+0.05754z-3.81721=0 (3)

[0134] 4. According to point cloud dataset A 平面 ′, and remove the points in point cloud dataset A1 that are located in point cloud dataset A. 平面 ′, we get Figure 4 The point cloud dataset A after removing the plane is shown 平面 ″;

[0135] 5. Pair A平面 ″ Perform cylindrical fitting and extract Figure 5 The cylindrical point cloud dataset B shown 圆柱 ';

[0136] 6. Based on point cloud dataset B 圆柱 ′, and remove the points in point cloud dataset A1 that are located in point cloud dataset B. 圆柱 ′, we get Figure 6 The steel coil end surface point cloud dataset A2 is shown;

[0137] 7. Calibrate the coil end face point cloud dataset A2 according to the parameters of the fitted plane point cloud, and rotate it so that the normal vector direction of the coil end face is parallel to the x-axis direction of the 3D laser radar 1 scanning surface, and obtain the following: Figure 7 The calibrated steel coil end face point cloud dataset A3 is shown in the figure, so that the normal vector of the steel coil plane is perpendicular to the scanning plane of the laser radar. 圆柱 ' Perform the same calibration and obtain the calibrated cylindrical point cloud dataset B 圆柱 ; Among them, the calibration matrix T is shown in formula 4:

[0138]

[0139] 8. Based on the obtained calibrated cylindrical point cloud dataset B 圆柱 , index the point with the largest y-axis value and the point with the smallest y-axis value on the cylinder, and record the coordinate values of the indexed point cloud B1 = (x1, y1, z1) and B2 = (x2, y2, z2), where (x1, y1, z1) = (4.508, 0.427, 0.526) and (x2, y2, z2) = (3.741, -0.311, 0.230); calculate the center coordinates of a point on the axis of the cylinder, as shown in formula (5);

[0140]

[0141] 9. Obtain i points with a height of 1 cm above and below z0 without removing the cylinder point cloud, and change all the z-axis values of the obtained point cloud to 0, while keeping the x and y axis values unchanged, and obtain the following: Figure 8 The point cloud dataset R0′ shown is as shown in Formula 6;

[0142]

[0143] 10. Index R0′ and record the point with the largest y value among the points with y-axis values greater than 0, y3 = 0.427, and the point with the smallest y value among the points with y-axis values less than 0, y4 = -0.312; where D = 0.8m. The calculation for this steel coil is shown in formula (7);

[0144] y3-y4=0.427-(-0.312)=0.739 (7)

[0145] 11. Obtain the point cloud radius dataset R1 with y coordinate values greater than 0, as shown in formula (8);

[0146] y>0(8)

[0147] 12. The point cloud dataset A3 is rotated 6 times, with the rotation angles being -Π / 6, -5Π / 6, -Π, -5Π / 4, -3Π / 2, and -7Π / 4 respectively. The point cloud radius dataset is recalculated after each rotation, and a total of 7 point cloud radius datasets R1′, R2′, R3′, R4′, R5′, R6′, and R7′ of the coil end face are obtained; the point cloud dataset R1′ is as follows Figure 9 shown.

[0148] 13. Sort the point cloud radius dataset R1′, R2′, R3′, R4′, R5′, R6′, R7′ according to the y-axis value from large to small and remove the points with smaller x-axis values in the point cloud data with the same y-axis value to obtain the sorted point cloud dataset R1, R2, R3, R4, R5, R6, R7;

[0149] 14. Count the flat coil judgment results. If the coil is judged as flat coil more than twice, it is determined that the coil has a flat coil defect.

[0150] Specifically, in this embodiment, the results of y3-y4 of the seven radii are 0.738m, 0.755m, 0.740m, 0.739m, 0.756m, 0.751m, and 0.751m, respectively. Therefore, it is determined that the steel coil has no flat coil defect.

[0151] 15. Detect the distance between two adjacent point clouds in the point cloud radius dataset R1 in the y-axis direction. If the distance between adjacent points is greater than 5 mm, it is judged as loose coiling, as shown in formula (9);

[0152] |y i -y i+1 |>0.005 (9)

[0153] 16. Extract the point cloud dataset R1, which is 1 / 3 of the inner circle of the point cloud radius dataset R 11 , calculate R 11 The maximum concave-convex amount σ in the x-axis direction max11 , if σ max11 Greater than 40mm, and the minimum value point x in the x direction min11 From point cloud dataset larger than R 11 The average value x of all points on the x-axis avg11 The distance is less than σ max11 / 2, it is judged to be a pyramidal roll; if it is greater than σ max11 / 2, it is judged as an inner circle overflow volume, and the calculation result of the volume is shown in formula (10):

[0154]

[0155] 17. Extract the point cloud dataset R of the outer 1 / 3 of R1 12 , calculate R 12 The point x with the minimum value in the x direction min12 The point x with the maximum value in the x direction max12 The distance x between point clouds max12 -x min12 If the calculated distance value is greater than the overflow standard of 10 cm, it is judged as an outer circle overflow roll; the calculation result of the roll is shown in formula (11);

[0156] x max12 -x min12 =0.046 (11)

[0157] 18. Extract the point cloud of the outer 1 / 3 of R1, calculate the mean value of the x-direction value in the extracted point cloud and calculate the variance δ. When δ1<δ<δ2, it is recorded as the outer circle staggered volume. When δ>δ2, it is recorded as the staggered volume. Among them, δ1 is the judgment threshold between the outer circle staggered volume and the normal volume, and δ2 is the judgment threshold between the outer circle staggered volume and the staggered volume. Take δ1=30mm 2 The judgment threshold between the inner ring staggered roll and the normal roll is δ2 = 60mm 2 is the judgment threshold of inner circle interlaced volume and interlaced volume; the calculation result of this volume is shown in formula (12);

[0158]

[0159] 19. Repeat the above judgment process for each radius and record the judgment result for each radius;

[0160] 20. The judgment results of each of the 7 radii are counted. If more than 3 radii judge that the steel coil has the same coil shape defect, the steel coil is finally judged to have this defect.

[0161] Specifically, in the judgment results of each radius in this embodiment, 6 radii judge that the steel coil has a coil defect of inner ring overflow, and 1 radius judges that the steel coil has a tower defect. Therefore, it is finally judged that the steel coil has an inner ring overflow defect.

[0162] At this point, the coil shape detection operation of the steel coil is completed.

[0163] Furthermore, it should be noted that the present invention may be provided as a method, apparatus, or computer program product. Thus, embodiments of the present invention may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code.

[0164] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0166] It should also be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "comprises," "includes," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further restrictions, an element defined by the phrase "comprises a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0167] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be noted that although the preferred embodiment of the present invention has been described, it is clear that those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered as within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A coil shape detection device based on the three-dimensional point cloud of the end face of a hot-rolled steel coil, characterized in that: include: 3D laser radar, mounting bracket, air cooler and server; among them, The three-dimensional laser radar and the air cooler are both installed in the mounting bracket, and the mounting bracket is placed on the operating side at a preset distance from the static position when the steel coil is off the line; the three-dimensional laser radar is connected to the server; the air cooler is used to dissipate heat from the three-dimensional laser radar; When the steel coil reaches the designated position, the three-dimensional laser radar starts to scan the end face of the steel coil on the operating side, obtains a point cloud of the steel coil end face, and transmits the obtained point cloud of the steel coil end face to the server. The server extracts and recognizes the features of the steel coil based on the point cloud of the steel coil end face to realize coil shape detection; Extract and identify steel coil features based on the coil end surface point cloud to achieve coil shape detection, including: Preprocessing the steel coil end surface point cloud obtained by the three-dimensional laser radar scanning; Perform angle calibration on the pre-processed point cloud of the steel coil end face and obtain the point cloud after removing the inner ring of the steel coil; Based on the calibrated point cloud, multiple radii on the end face of the steel coil are obtained through straight-through filtering. The features of the point cloud radius are extracted to determine whether the hot-rolled steel coil is a flat coil and complete the coil type recognition of the hot-rolled steel coil; Preprocessing the steel coil end surface point cloud obtained by the three-dimensional laser radar scanning includes: According to the placement of the three-dimensional laser radar and the steel coil, the maximum radius R of the offline steel coil max The position of the mounting bracket from the steel coil is determined by removing points in the steel coil end face point cloud that are far away from the steel coil through straight-through filtering to obtain a first point cloud dataset; wherein the straight-through filtering range is shown in the following formula: Wherein, x, y, and z are the scanning directions of the three-dimensional laser radar respectively; The first point cloud dataset is statistically filtered to remove points with an appearance probability less than 0.3174 to remove noise points and outliers, thereby obtaining a second point cloud dataset, as shown in the following formula: Among them, μ is the mean distance of the global points, σ is the standard deviation of the distance of the global points, d k is the average distance from the target point to its k neighboring points, if d k If the global standard deviation is less than 1, it will be considered as a noise point and removed; Perform angle calibration on the pre-processed coil end face point cloud and obtain the point cloud after removing the inner ring of the coil, including: The RANSAC plane fitting method is used to fit a plane to the second point cloud dataset, extract the plane point cloud dataset, and obtain the normal vector of the plane; the characteristic equation of the fitted plane is shown as follows: A 11 +B 11 +C 11 +D 11 =0 performing indexing based on the points on the planar point cloud dataset, removing the points in the second point cloud dataset that are located in the planar point cloud dataset, to obtain a third point cloud dataset; Performing cylindrical fitting on the third point cloud dataset to extract a cylindrical point cloud dataset; performing indexing based on the points on the cylindrical point cloud dataset, removing the points in the second point cloud dataset that are located in the cylindrical point cloud dataset, to obtain a fourth point cloud dataset; The fourth point cloud dataset and the cylindrical point cloud dataset are rotated according to the normal vector of the fitted plane so that the normal vector direction of the steel coil end surface is parallel to the x-axis direction of the scanning plane of the three-dimensional laser radar, thereby obtaining a calibrated fourth point cloud dataset; and the normal vector of the steel coil plane is made perpendicular to the scanning plane of the laser radar, thereby obtaining a calibrated cylindrical point cloud dataset; wherein the calibration matrix T is shown as follows: Among them, A 11 、B 11 、C 11 、D 11 are the four parameters of the plane equation of the steel coil end face; Through straight-through filtering, multiple radii on the coil end surface are obtained, and the features of the point cloud radius are extracted to determine whether the hot-rolled steel coil is a flat coil and complete the coil type identification of the hot-rolled steel coil, including: Step 1: Based on the calibrated cylindrical point cloud dataset, index the point with the largest and smallest y-axis values on the cylinder, record the coordinate values of the indexed point cloud B1 = (x1, y1, z1) and B2 = (x2, y2, z2), and calculate the center coordinates of a point on the axis of the cylinder, as shown in the following formula; Step 2: Obtain i points within a range of 1 cm above and below z0 on the currently calibrated fourth point cloud dataset through straight-through filtering, and set all z-axis values of the obtained point cloud to 0, while keeping the x and y coordinate values unchanged, to obtain the fifth point cloud dataset; Step 3: Index the fifth point cloud dataset and record the point C1 = (x3, y3, z3) with the smallest y coordinate among the points with y coordinate values greater than 0, and the point C2 = (x4, y4, z4) with the largest y coordinate among the points with y coordinate values less than 0. The value obtained by calculating y3-y4 is the inner diameter of the steel coil. If the calculated value of y3-y4 is less than 0.8D or greater than 1.2D, the current steel coil is determined to be a flat coil; where D is the inner radius of the standard coil. Step 4: Obtain a point cloud radius dataset with y coordinate values greater than 0 through straight-through filtering; Step 5: Rotate the calibrated fourth point cloud dataset by a preset angle to obtain a new calibrated fourth point cloud dataset, and re-execute steps 1 to 4 to obtain a new point cloud radius dataset. Step 6: Iterate step 5 six times to obtain seven point cloud radius datasets. Sorting the seven point cloud radius datasets from large to small according to the y-axis value and removing the points with smaller x-axis values in the point cloud data with the same y-axis value to obtain the sorted point cloud radius dataset. Step 7: Count the flat coil judgment results. If the flat coil is judged as flat coil more than twice, it is determined that the steel coil has a flat coil defect. Step 8: Extract features of the radius corresponding to each point cloud radius data set to preliminarily determine the coil shape of the steel coil, and count the judgment results of each radius to obtain the final judgment result of the coil shape.

2. The coil shape detection device based on the three-dimensional point cloud of the end face of the hot-rolled steel coil according to claim 1, characterized in that: The height of the mounting bracket is adjustable; The mounting bracket includes: a fixed support fixed part, a support movable part, a rotating base plate, an upper shell and a conical panel; wherein, the three-dimensional laser radar is placed in the middle position of the upper shell, and its scanning surface faces outward toward the conical panel, and the air cooler is placed at the tail position of the upper shell for dissipating heat for the three-dimensional laser radar. The installation height of the three-dimensional laser radar is 30 to 50 cm lower than the height of the center line of the steel coil, and the yoz scanning plane of the three-dimensional laser radar is perpendicular to the center line direction of the steel coil.

3. The coil shape detection device based on the three-dimensional point cloud of the end face of the hot-rolled steel coil according to claim 1, characterized in that: The feature extraction is performed on the radius corresponding to each point cloud radius dataset to preliminarily judge the coil shape. The judgment results of each radius are statistically analyzed to obtain the final coil shape judgment result, including: For each point cloud radius dataset, perform the following steps: Detect the distance in the y-axis direction between two adjacent point clouds in the current point cloud radius dataset. If the distance between adjacent points is greater than 5mm, it is judged as loose roll; Extract the point cloud of the inner 1 / 3 of the current point cloud radius dataset and calculate the maximum concave-convex value σ in the x-axis direction of the extracted point cloud max11 , the average value x of all points on the x-axis avg11 And the minimum value point in the x direction is away from x avg11 If the minimum value of the x direction point is away from x avg11 The distance is less than σ max11 / 2, it is judged to be a pyramidal roll; if the minimum value point in the x direction is away from x avg11 The distance is greater than σ max11 / 2, it is judged as an inner circle overflow volume; Extract the point cloud of the outer 1 / 3 of the current point cloud radius dataset and calculate the point x with the minimum x value in the extracted point cloud. min12 The point x with the maximum value in the x direction max12 The distance x between point clouds max12 -x min12 , if the calculated x max12 -x min12 If the value is greater than the overflow standard of 10cm, it is judged as an outer circle overflow roll; Extract the point cloud of the outer 1 / 3 of the current point cloud radius dataset, calculate the mean value of the x-direction value in the extracted point cloud and calculate the variance δ. When δ1<δ<δ2, it is recorded as the outer circle interlaced volume, and when δ>δ2, it is recorded as the interlaced volume, where δ1 is the judgment threshold between the outer circle interlaced volume and the normal volume, and δ2 is the judgment threshold between the outer circle interlaced volume and the interlaced volume. The judgment results corresponding to each radius are counted. If there are more than three radii that judge that the current steel coil has the same preset type of coil shape defect, it is finally judged that the current steel coil has the preset type of coil shape defect.

4. A coil shape detection method implemented by using the coil shape detection device based on the three-dimensional point cloud of the end face of a hot-rolled steel coil according to any one of claims 1 to 3, characterized in that: The method comprises: When the steel coil arrives at the designated position, a steel coil in-position signal is transmitted to the three-dimensional laser radar, which is then activated to scan the end face of the steel coil on the operating side to obtain a point cloud of the steel coil end face; The point cloud of the steel coil end surface obtained by the three-dimensional laser radar scanning is transmitted to the server, and the server extracts and identifies the features of the steel coil based on the point cloud of the steel coil end surface to realize coil shape detection.

Images