A method and system for visual detection of hot-rolled strip deviation based on dynamic height compensation

Abstract

The present invention provides a method and system for visual detection of hot-rolled strip deviation based on dynamic height compensation. The method comprises: installing an industrial camera between two groups of frames and installing multiple ranging sensors at the same height, calibrating the camera's internal parameter matrix and external parameter matrix, and simultaneously recording the vertical distance from each ranging sensor to the calibration plane; acquiring the vertical distance from the ranging sensor to the strip surface in real time during the detection phase, calculating the height offset and its mean; acquiring the strip image in real time during the detection phase, and performing edge detection on the left and right edges of the strip image; calculating the coordinates of points on the left and right edges of the strip in a world coordinate system using the camera's internal parameter matrix and external parameter matrix based on the pixel coordinates on the left and right edges of the strip and the mean of the height offset; performing outlier filtering and mean calculation on the left and right edge points of the strip in the world coordinate system, determining the actual edge position, and calculating the deviation.

Description

Technical Field

[0001] The present invention belongs to the technical field of strip finishing control, and relates to a method and system for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation. Background Art

[0002] During hot-rolled strip production, strip deviation is a critical issue affecting rolling quality and equipment safety. Due to factors such as rolling tension fluctuations, temperature gradients, and roll wear, the strip often experiences lateral deviation while running between the mills. In severe cases, this can lead to edge scraping, steel accumulation, and even damage to roll bearings, resulting in unplanned line downtime. Therefore, real-time, high-precision detection of strip deviation is crucial for improving rolling stability and product quality.

[0003] Chinese invention patent application CN116274419A, "A Method for Detecting Hot-Rolled Strip Deviation Considering the Effect of Loops," relies on the looper angle and strip thickness to determine the world coordinate system transformation relationship during strip rotation, thereby eliminating the looper's influence on strip deviation detection. However, in actual production, this method is limited by mechanical hysteresis and cannot respond to transient strip fluctuations in real time. Furthermore, due to factors such as thermal expansion and vibration, the strip cannot accurately reflect changes in strip height, resulting in errors in the coordinate transformation model.

[0004] Chinese invention patent CN102319743B, "Laser scanning detection method and correction system for strip deviation and floating amount," uses the shape of a linear structured light beam to determine the edge of the strip and calculates the vertical floating amount of the strip based on the displacement of the laser linear beam in the length direction of the strip. However, the patent does not consider the impact of floating amount on deviation measurement. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of the present invention is to provide a method and system for visual detection of hot-rolled strip deviation based on dynamic height compensation.

[0006] The present invention provides a method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation, comprising:

[0007] Step 1: Install an industrial camera between two sets of racks and install multiple ranging sensors at the same height. Calibrate the camera's intrinsic parameter matrix K and extrinsic parameter matrix T = [R|t], and establish the initial mapping relationship between the pixel coordinate system and the world coordinate system. At the same time, record the vertical distance a from each ranging sensor to the calibration plane. g ;

[0008] Step 2: During the detection phase, the vertical distance b between the distance measuring sensor and the strip surface is obtained in real time. g , calculate the height offset Δh g =a g -b gand its mean ΔH;

[0009] Step 3: In the detection phase, an industrial camera is used to capture the strip image in real time and perform edge detection on the left and right edges of the strip image;

[0010] Step 4: Based on the pixel coordinates of the points on the left and right edges of the strip and ΔH, the coordinates of the points on the left and right edges of the strip in the world coordinate system are calculated using the camera's intrinsic and extrinsic matrix.

[0011] Step 5: Filter outliers and calculate the mean of the left and right edge points of the strip in the world coordinate system to determine the actual edge position and calculate the deviation amount.

[0012] Furthermore, the step 1 is specifically as follows:

[0013] Step 1.1: Install an industrial camera between the two sets of racks. The edge of the strip must be within the camera's field of view, and the strip must move vertically within the camera's field of view.

[0014] Step 1.2: Install multiple distance sensors at the same height between the two sets of racks. The distance sensors form a predetermined angle with the optical axis of the industrial camera to ensure that the strip inspection area is covered.

[0015] Step 1.3: Use a checkerboard calibration plate. When the rolling mill is stopped, lower the looper to a horizontal position. Place the checkerboard calibration plate on the upper surface of the looper so that its plane is parallel to the rolling centerline and coincides with the plane of the looper's initial position. Define this plane as the calibration plane, that is, the plane of the world coordinate system Z = 0, with the positive direction of the Z axis pointing vertically downward. Record the vertical distance a from each ranging sensor to the calibration plane. g ;

[0016] Step 1.4: Use an industrial camera to collect multiple sets of calibration plate images in different poses, and calculate the camera's intrinsic parameter matrix using Zhang Zhengyou's calibration method:

[0017]

[0018] And the external parameter matrix, including the rotation matrix R and the translation vector t:

[0019]

[0020] Among them, f x 、f y are the focal lengths of the camera measured in pixels in the x and y directions, u0 and v0 are the coordinates of the principal point of the image in pixels, R is a 3×3 rotation matrix, and t is a 3×1 translation vector.

[0021] Furthermore, the step 2 is specifically as follows:

[0022] Step 2.1: During the detection phase, obtain the vertical distance b from the distance measuring sensor to the strip surface in real time. g , calculate the height offset Δh g =a g -b g ;

[0023] Step 2.2: Calculate the mean height offset ΔH:

[0024]

[0025] Where G is the total number of ranging sensors.

[0026] Furthermore, the edge detection in step 3 adopts a sub-pixel edge extraction algorithm to obtain the pixel coordinates of the left and right edges of the steel strip.

[0027] Furthermore, the step 4 is specifically as follows:

[0028] Step 4.1: Calculate the normalized imaging plane coordinates of the left edge point:

[0029]

[0030] Among them, u 1i , v 1i is the coordinate of the i-th point on the left edge in the pixel coordinate system; x ni ,y ni is the coordinate of the i-th point on the left edge transformed into the normalized imaging plane coordinate system;

[0031] Point x in the normalized imaging plane coordinate system ni ,y ni satisfy:

[0032]

[0033] Among them, X c ,Y c ,Z c is the coordinate of the edge point converted to the camera coordinate system;

[0034] The external parameter relationship from the world coordinate system to the camera coordinate system:

[0035]

[0036] Step 4.2: X c =x ni Z c and Y c =y ni Z c Substitute the external parameter equation to construct a linear equation system:

[0037]

[0038] Among them, Z w =-ΔH;X wi ,Y wi is the world coordinate of the i-th point on the left edge;

[0039] Step 4.3: Calculate the world coordinates of the left edge point:

[0040]

[0041] Step 4.4: Use the same method to calculate the normalized imaging plane coordinate system coordinate x of the jth point on the right edge nj ,y nj , construct a system of linear equations and solve the world coordinate X of the jth point on the right edge wj , Y wj .

[0042] Furthermore, the step 5 is specifically as follows:

[0043] Step 5.1: Calculate the mean and standard deviation of the world coordinates of the left and right edge points:

[0044]

[0045]

[0046] Where N is the total number of edge points on the left; M is the total number of edge points on the right; μ1 and σ1 are the mean and standard deviation of the world coordinates of the left edge points, respectively; μ2 and σ2 are the mean and standard deviation of the world coordinates of the right edge points, respectively;

[0047] Step 5.2: The left edge is retained to satisfy |X wi -μ1|<3σ1 coordinate point, the right edge retains the satisfying |X wj -μ2|<3σ2 coordinate points, the rest are considered as outliers and removed;

[0048] Step 5.3: Repeat steps 5.1-5.2 until any of the following conditions are met:

[0049] 1) The standard deviation of the remaining valid points on the left and the standard deviation of the remaining valid points on the right are both less than or equal to the corresponding set thresholds;

[0050] 2) The number of iterations reaches 3;

[0051] Step 5.4: Calculate the world coordinate average of the remaining valid points as the left and right edge reference positions. The formula is:

[0052]

[0053] Among them, n and m are the number of valid points on the left and right edges respectively;

[0054] Step 5.5: Calculate the deviation amount based on the left and right edge reference positions. The formula is:

[0055]

[0056] Furthermore, the distance measuring sensor is a non-contact distance measuring sensor, including a laser distance measuring sensor or an ultrasonic distance measuring sensor.

[0057] The present invention also provides a hot-rolled strip deviation visual detection system based on dynamic height compensation, which is used to implement the above-mentioned hot-rolled strip deviation visual detection method based on dynamic height compensation. The system includes: an image acquisition module, a distance detection module, a data processing module and a protective structure;

[0058] The image acquisition module is an industrial camera installed between the two sets of racks. The industrial camera covers the strip steel area between the racks, ensuring that the edge of the strip steel is within the camera's field of view. The field of view range needs to be verified during the calibration phase to ensure that only the strip steel area directly below the ranging sensor is covered;

[0059] The distance detection module includes multiple distance measuring sensors, which are used to measure the vertical distance from the sensor to the calibration plane during the calibration phase and to measure the vertical distance from the sensor to multiple measuring points on the strip surface during the detection phase;

[0060] The data processing module is connected to the image acquisition module and the distance detection module respectively; the data processing module has a built-in calibration parameter storage unit, an edge extraction unit, a first calculation unit, an outlier filtering unit and a second calculation unit;

[0061] The edge extraction unit extracts the pixel coordinates of the left and right edges of the steel strip based on the collected steel strip image; the first calculation unit calculates the height offset and its average value based on the vertical distance from the sensor to the calibration plane and the vertical distance from the sensor to multiple measurement points on the steel strip surface, and solves the actual coordinates of the left and right edge points in the world coordinate system based on the pixel coordinates of the left and right edge points and the average value of the height offset; the outlier filtering unit is used to filter the actual coordinates of the left and right edge points in the world coordinate system and eliminate invalid points; the second calculation unit is used to calculate the average world coordinates of the remaining valid points as the left and right edge reference positions, and calculate the deviation amount based on the left and right edge reference positions;

[0062] The protective structure is used to wrap the image acquisition module and the distance detection module, and has cooling and dust-proof functions.

[0063] Furthermore, the protective structure includes a water-cooling interlayer and an air curtain isolation device to adapt to the high temperature environment of the rolling site.

[0064] Furthermore, the data processing module communicates with the first-level system of the rolling mill via industrial Ethernet, and outputs the deviation value to the HMI interface and control unit of the first-level system of the rolling mill.

[0065] A method for visual detection of hot-rolled strip deviation based on dynamic height compensation has the following features:

[0066] Beneficial effects:

[0067] This method uses a laser rangefinder or ultrasonic sensor to collect real-time data on the height variation of the strip surface in the rolling mill. Combined with pre-calibrated camera internal and external parameters, it dynamically corrects the transformation from the pixel coordinate system to the world coordinate system, eliminating errors in strip height fluctuations that can affect runout measurement. This method solves the problem of inaccurate detection caused by strip height fluctuations in traditional methods and is suitable for high-temperature, high-vibration rolling environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a flow chart of a method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to the present invention;

[0069] Figure 2 This is a deployment diagram of a hot-rolled strip deviation visual detection system based on dynamic height compensation according to the present invention. DETAILED DESCRIPTION

[0070] like Figure 1 As shown, the present invention provides a method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation, comprising:

[0071] Step 1: Install an industrial camera between two sets of racks and install multiple ranging sensors at the same height. Calibrate the camera's intrinsic parameter matrix K and extrinsic parameter matrix T = [R|t], and establish the initial mapping relationship between the pixel coordinate system and the world coordinate system. At the same time, record the vertical distance a from each ranging sensor to the calibration plane. g , specifically:

[0072] Step 1.1: Install an industrial camera between the two sets of racks. The edge of the steel strip needs to be within the camera's field of view, and the steel strip moves vertically in the camera's field of view.

[0073] Step 1.2: Install multiple distance sensors at the same height between the two sets of racks. The distance sensors form a predetermined angle with the optical axis of the industrial camera to ensure coverage of the strip inspection area. The distance sensors are non-contact distance sensors, including laser distance sensors or ultrasonic distance sensors.

[0074] Step 1.3: Use a checkerboard calibration plate. When the rolling mill is stopped, lower the looper to a horizontal position. Place the checkerboard calibration plate on the upper surface of the looper so that its plane is parallel to the rolling centerline and coincides with the plane of the looper's initial position. Define this plane as the calibration plane, that is, the plane of the world coordinate system Z = 0, with the positive direction of the Z axis pointing vertically downward. Record the vertical distance a from each ranging sensor to the calibration plane. g .

[0075] In the world coordinate system, the X-axis is located in the calibration plane and perpendicular to the rolling centerline, with its positive direction pointing to the right of the strip (as viewed along the strip's direction of motion). The Y-axis is located in the calibration plane and parallel to the rolling centerline, with its positive direction opposite the strip's direction of motion. The Z-axis is perpendicular to the calibration plane, with its positive direction pointing vertically downward. The calibration plane is the Z=0 plane; the rolling centerline corresponds to X=0 and Z=0.

[0076] Step 1.4: Use an industrial camera to collect multiple sets of calibration plate images in different poses, and calculate the camera's intrinsic parameter matrix using Zhang Zhengyou's calibration method:

[0077]

[0078] And the external parameter matrix, including the rotation matrix R and the translation vector t:

[0079]

[0080] Among them, f x 、f y are the focal lengths of the camera measured in pixels in the x and y directions, u0 and v0 are the coordinates of the principal point of the image in pixels, R is a 3×3 rotation matrix, and t is a 3×1 translation vector.

[0081] Step 2: During the detection phase, the vertical distance b between the distance measuring sensor and the strip surface is obtained in real time. g , calculate the height offset Δh g =a g -b g and its mean ΔH, specifically:

[0082] Step 2.1: During the detection phase, obtain the vertical distance b from the distance measuring sensor to the strip surface in real time. g , calculate the height offset Δh g =a g -b g .

[0083] Step 2.2: Calculate the mean height offset ΔH:

[0084]

[0085] Where G is the total number of ranging sensors.

[0086] Step 3: In the detection phase, an industrial camera is used to capture the strip image in real time and perform edge detection on the left and right edges of the strip image;

[0087] In specific implementation, edge detection uses a sub-pixel edge extraction algorithm to obtain the pixel coordinates of the left and right edges of the strip.

[0088] Step 4: Based on the pixel coordinates of the points on the left and right edges of the strip and ΔH, the actual coordinates of the points on the left and right edges of the strip in the world coordinate system are calculated using the camera's intrinsic and extrinsic matrix. Specifically:

[0089] Step 4.1: Calculate the normalized imaging plane coordinates of the left edge point:

[0090]

[0091] Among them, u 1i , v 1i is the coordinate of the i-th point on the left edge in the pixel coordinate system; x ni ,y ni is the coordinate of the i-th point on the left edge transformed into the normalized imaging plane coordinate system.

[0092] Point x in the normalized imaging plane coordinate system ni ,y ni satisfy:

[0093]

[0094] Among them, X c ,Y c ,Z c It is the coordinate of the edge point converted to the camera coordinate system.

[0095] The external parameter relationship from the world coordinate system to the camera coordinate system:

[0096]

[0097] Step 4.2: X c =x ni Z c and Y c =y ni Z c Substitute the external parameter equation to construct a linear equation system:

[0098]

[0099] Among them, Z w =-ΔH;X wi ,Y wi is the world coordinate of the i-th point on the left edge.

[0100] Step 4.3: Calculate the world coordinates of the left edge point:

[0101]

[0102] Step 4.4: Use the same method to calculate the normalized imaging plane coordinate system coordinate x of the jth point on the right edge nj ,y nj , construct a system of linear equations and solve the world coordinate X of the jth point on the right edge wj , Y wj .

[0103] Step 5: Filter outliers and calculate the mean of the left and right edge points of the strip in the world coordinate system to determine the actual edge position and calculate the deviation amount. Specifically:

[0104] Step 5.1: Calculate the mean and standard deviation of the world coordinates of the left and right edge points:

[0105]

[0106] Where N is the total number of edge points on the left; M is the total number of edge points on the right; μ1 and σ1 are the mean and standard deviation of the world coordinates of the left edge points, respectively; μ2 and σ2 are the mean and standard deviation of the world coordinates of the right edge points, respectively.

[0107] Step 5.2: The left edge is retained to satisfy |X wi -μ1|<3σ1 coordinate point, the right edge retains the satisfying |X wj The coordinate points with -μ2|<3σ2 are selected, and the rest are considered as outliers and removed.

[0108] Step 5.3: Repeat steps 5.1-5.2 until any of the following conditions are met:

[0109] 1) The standard deviation of the remaining valid points on the left and the standard deviation of the remaining valid points on the right are both less than or equal to the corresponding set thresholds;

[0110] 2) The number of iterations reaches 3;

[0111] Step 5.4: Calculate the world coordinate average of the remaining valid points as the left and right edge reference positions. The formula is:

[0112]

[0113]

[0114] Among them, n and m are the number of valid points on the left and right edges respectively.

[0115] Step 5.5: Calculate the deviation amount based on the left and right edge reference positions. The formula is:

[0116]

[0117] Among them, D is the strip deviation.

[0118] like Figure 2 The figure shows a deployment diagram of a hot-rolled strip deviation visual detection system based on dynamic height compensation. The detection system includes: an image acquisition module, a distance detection module, a data processing module and a protective structure.

[0119] The image acquisition module is an industrial camera installed between two groups of racks. The industrial camera covers the strip steel area between the racks to ensure that the edge of the strip steel is within the field of view of the camera. The field of view range needs to be verified during the calibration phase to ensure that only the strip steel area directly below the ranging sensor is covered.

[0120] The distance detection module includes multiple distance measuring sensors, which are used to measure the vertical distance from the sensor to the calibration plane during the calibration phase and to measure the vertical distance from the sensor to multiple measuring points on the strip surface during the detection phase.

[0121] The data processing module is connected to the image acquisition module and the distance detection module respectively; the data processing module has built-in calibration parameter storage unit, edge extraction unit, first calculation unit, outlier filtering unit and second calculation unit.

[0122] The edge extraction unit extracts the pixel coordinates of the left and right edges of the steel strip based on the collected steel strip image; the first calculation unit calculates the height offset and its average value based on the vertical distance from the sensor to the calibration plane and the vertical distance from the sensor to multiple measurement points on the steel strip surface, and solves the actual coordinates of the left and right edge points in the world coordinate system based on the pixel coordinates of the left and right edge points and the average value of the height offset; the outlier filtering unit is used to filter the actual coordinates of the left and right edge points in the world coordinate system and eliminate invalid points; the second calculation unit is used to calculate the average world coordinates of the remaining valid points as the left and right edge reference positions, and calculate the deviation amount based on the left and right edge reference positions;

[0123] The protective structure is used to enclose the image acquisition module and the distance detection module, and has cooling and dust-proof functions. In specific implementation, the protective structure includes a water-cooling interlayer and an air curtain isolation device to adapt to the high-temperature environment of the rolling site.

[0124] The data processing module communicates with the first-level system of the rolling mill via industrial Ethernet, and outputs the deviation value to the HMI interface and control unit of the first-level system of the rolling mill.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the concept of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation, characterized in that: include: Step 1: Install an industrial camera between two sets of racks and install multiple ranging sensors at the same height. Calibrate the camera's intrinsic parameter matrix K and extrinsic parameter matrix T = [R|t], and establish the initial mapping relationship between the pixel coordinate system and the world coordinate system. At the same time, record the vertical distance a from each ranging sensor to the calibration plane. g ; Step 2: During the detection phase, the vertical distance b between the distance measuring sensor and the strip surface is obtained in real time. g , calculate the height offset Δh g =a g -b g and its mean ΔH; Step 3: In the detection phase, an industrial camera is used to capture the strip image in real time and perform edge detection on the left and right edges of the strip image; Step 4: Based on the pixel coordinates of the points on the left and right edges of the strip and ΔH, the coordinates of the points on the left and right edges of the strip in the world coordinate system are calculated using the camera's intrinsic and extrinsic matrix. Step 5: Filter outliers and calculate the mean of the left and right edge points of the strip in the world coordinate system to determine the actual edge position and calculate the deviation amount.

2. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 1, characterized in that: The step 1 is specifically as follows: Step 1.1: Install an industrial camera between the two sets of racks. The edge of the strip must be within the camera's field of view, and the strip must move vertically within the camera's field of view. Step 1.2: Install multiple distance sensors at the same height between the two sets of racks. The distance sensors form a predetermined angle with the optical axis of the industrial camera to ensure that the strip inspection area is covered. Step 1.3: Use a checkerboard calibration plate. When the rolling mill is stopped, lower the looper to a horizontal position. Place the checkerboard calibration plate on the upper surface of the looper so that its plane is parallel to the rolling center line and coincides with the plane of the looper's initial position. Define this plane as the calibration plane, that is, the plane of the world coordinate system Z = 0, with the positive direction of the Z axis pointing vertically downward. Record the vertical distance a from each ranging sensor to the calibration plane. g ; Step 1.4: Use an industrial camera to collect multiple sets of calibration plate images in different poses, and calculate the camera's intrinsic parameter matrix using Zhang Zhengyou's calibration method: And the external parameter matrix, including the rotation matrix R and the translation vector t: Among them, f x 、f y are the focal lengths of the camera measured in pixels in the x and y directions, u0 and v0 are the coordinates of the principal point of the image in pixels, R is a 3×3 rotation matrix, and t is a 3×1 translation vector.

3. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 1, characterized in that: The step 2 is specifically as follows: Step 2.1: During the detection phase, obtain the vertical distance b from the distance measuring sensor to the strip surface in real time. g , calculate the height offset Δh g =a g -b g ; Step 2.2: Calculate the mean height offset ΔH: Where G is the total number of ranging sensors.

4. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 1, characterized in that: The edge detection in step 3 adopts a sub-pixel edge extraction algorithm to obtain the pixel coordinates of the left and right edges of the strip.

5. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 1, characterized in that: The step 4 is specifically as follows: Step 4.1: Calculate the normalized imaging plane coordinates of the left edge point: Among them, u 1i , v 1i is the coordinate of the i-th point on the left edge in the pixel coordinate system; x ni ,y ni is the coordinate of the i-th point on the left edge transformed into the normalized imaging plane coordinate system; Point x in the normalized imaging plane coordinate system ni ,y ni satisfy: Among them, X c ,Y c ,Z c is the coordinate of the edge point converted to the camera coordinate system; The external parameter relationship from the world coordinate system to the camera coordinate system: Step 4.2: X c =x ni Z c and Y c =y ni Z c Substitute the external parameter equation to construct a linear equation system: Among them, Z w =-ΔH;X wi ,Y wi is the world coordinate of the i-th point on the left edge; Step 4.3: Calculate the world coordinates of the left edge point: Step 4.4: Use the same method to calculate the normalized imaging plane coordinate system coordinate x of the jth point on the right edge nj ,y nj , construct a system of linear equations and solve the world coordinate X of the jth point on the right edge wj , Y wj .

6. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 5, characterized in that: The step 5 is specifically as follows: Step 5.1: Calculate the mean and standard deviation of the world coordinates of the left and right edge points: Where N is the total number of edge points on the left; M is the total number of edge points on the right; μ1 and σ1 are the mean and standard deviation of the world coordinates of the left edge points, respectively; μ2 and σ2 are the mean and standard deviation of the world coordinates of the right edge points, respectively; Step 5.2: The left edge is retained to satisfy |X wi -μ1|<3σ1 coordinate point, the right edge retains the satisfying |X wj -μ2|<3σ2 coordinate points, the rest are considered as outliers and removed; Step 5.3: Repeat steps 5.1-5.2 until any of the following conditions are met: 1) The standard deviation of the remaining valid points on the left and the standard deviation of the remaining valid points on the right are both less than or equal to the corresponding set thresholds; 2) The number of iterations reaches 3; Step 5.4: Calculate the world coordinate average of the remaining valid points as the left and right edge reference positions. The formula is: Among them, n and m are the number of valid points on the left and right edges respectively; Step 5.5: Calculate the deviation amount based on the left and right edge reference positions. The formula is:

7. The method for visually detecting deviation of hot-rolled strip steel based on dynamic height compensation according to claim 1, characterized in that: The distance measuring sensor is a non-contact distance measuring sensor, including a laser distance measuring sensor or an ultrasonic distance measuring sensor.

8. A hot-rolled strip deviation visual detection system based on dynamic height compensation, used to implement the hot-rolled strip deviation visual detection method based on dynamic height compensation according to any one of claims 1 to 7, characterized in that: include: Image acquisition module, distance detection module, data processing module and protective structure; The image acquisition module is an industrial camera installed between the two sets of racks. The industrial camera covers the strip steel area between the racks, ensuring that the edge of the strip steel is within the camera's field of view. The field of view range needs to be verified during the calibration phase to ensure that only the strip steel area directly below the ranging sensor is covered; The distance detection module includes multiple distance measuring sensors, which are used to measure the vertical distance from the sensor to the calibration plane during the calibration phase and to measure the vertical distance from the sensor to multiple measuring points on the strip surface during the detection phase; The data processing module is connected to the image acquisition module and the distance detection module respectively; the data processing module has a built-in calibration parameter storage unit, an edge extraction unit, a first calculation unit, an outlier filtering unit and a second calculation unit; The edge extraction unit extracts the pixel coordinates of the left and right edges of the steel strip based on the collected steel strip image; the first calculation unit calculates the height offset and its average value based on the vertical distance from the sensor to the calibration plane and the vertical distance from the sensor to multiple measurement points on the steel strip surface, and solves the actual coordinates of the left and right edge points in the world coordinate system based on the pixel coordinates of the left and right edge points and the average value of the height offset; the outlier filtering unit is used to filter the actual coordinates of the left and right edge points in the world coordinate system to eliminate invalid points; The second calculation unit is used to calculate the world coordinate average of the remaining valid points as the left and right edge reference positions, and calculate the deviation amount based on the left and right edge reference positions; The protective structure is used to wrap the image acquisition module and the distance detection module, and has cooling and dust-proof functions.

9. The hot-rolled strip deviation visual detection system based on dynamic height compensation according to claim 8, characterized in that: The protective structure includes a water-cooling interlayer and an air curtain isolation device, which can adapt to the high temperature environment of the rolling site.

10. The hot-rolled strip deviation visual detection system based on dynamic height compensation according to claim 8, characterized in that: The data processing module communicates with the first-level system of the rolling mill via industrial Ethernet, and outputs the deviation value to the HMI interface and control unit of the first-level system of the rolling mill.

Images