Method and system for detecting misregistration of a top roll

Abstract

The application provides a detection method and system for a winding centering deviation, comprising the following steps: S1, constructing and training a centering deviation detection model; arranging an illuminating device and a camera device to capture images at intervals; S2, inputting the images into the centering deviation detection model, calculating the center coordinates of an ellipse in the images, and then calculating a pixel deviation value; obtaining an actual deviation distance according to the pixel deviation value; S3, drawing a trajectory of the ellipse on the images; S4, judging whether the actual deviation distance is greater than or equal to a threshold value; if the result is yes, storing current data and issuing a warning signal; and if the result is no, storing the current data. The application can detect the centering height of a steel coil with high precision through visual recognition, avoid the occurrence of the winding phenomenon, and reduce the labor intensity of an operator.

Description

Technical Field

[0001] This invention belongs to the field of steel coil inner ring inspection technology, specifically, it relates to a method and system for detecting upper coil alignment deviation. More specifically, it is a method for detecting upper coil alignment deviation under oblique viewing conditions. Background Technology

[0002] In cold-rolled sheet metal production, processes such as continuous rolling, annealing, coating, and finishing mostly require coiling and uncoiling operations. An abnormality in the center position of the inner coil during production can cause problems when the coil is loaded onto the production line. This is usually due to issues with the inspection equipment, operational errors, or collapse of the inner coil, preventing the coil from being properly inserted into the mandrel and disrupting the normal operation of the production line.

[0003] To address these issues, a series of measures are typically implemented on the production line to ensure the quality and standards of the inner coil of the steel coil. This includes rigorous quality inspection of the inner coil to ensure its dimensions and shape meet production requirements. Simultaneously, operators receive training to ensure they can correctly and safely handle the alignment height of the inner coil during production, avoiding abnormalities caused by operational errors.

[0004] Currently, in actual production, during coiling, the coil is first lifted to the corresponding height by the coiling trolley according to the coil size. Then, the height of the centering grating is checked by the detection equipment. After confirmation that it is correct, the coiling trolley moves the coil to the reel for the coiling operation. However, the centering grating frequently malfunctions, increasing on-site maintenance, and the detection error is relatively large, easily leading to coil flipping during coiling and causing property damage. This problem urgently needs to be solved.

[0005] Patent document CN104475463A discloses a steel coil height alignment device and method. This device is used for height alignment between the steel coil and the uncoiling machine mandrel during coiling. It includes a coiling trolley with a saddle for placing the steel coil, a lifting device for adjusting the coil height, a lifting encoder for recording the lifting height of the coil by the lifting device, and a limiter for defining the mechanical zero point. A column is mounted on each side of the trolley's moving track. One column has a positioning light source for emitting a concentrated beam of light, and the other column has a light receiving device for receiving the light emitted by the positioning light source. The positioning light source and the light receiving device are at the same height, and their line of connection is perpendicular to the moving track. The steel coil height alignment method using this device aligns the coil axially, directly measuring the height difference between the upper and lower edges of the inner diameter to determine the coil's center. This solution uses light source transmission and reception for height detection, altering the original production configuration, increasing equipment maintenance costs, and cannot avoid the problem of alignment grating failure.

[0006] Patent document CN107344684A discloses a touch-type winding positioning device and its winding centering method. The device includes two guide supports, with a hydraulic cylinder fixed between the two guide supports. Two guide rods are arranged parallel to each other on both sides of the hydraulic cylinder. The two guide rods pass through guide sleeves embedded in the two guide supports. The cylinder rod of the hydraulic cylinder passes through one of the guide supports and is fixedly connected to one side of a movable support together with the two guide rods. The upper surface of the movable support is provided with a rotating shaft and a small shaft. A positioning swing arm is mounted on the rotating shaft. The positioning swing arm is divided into a long arm and a short arm with the rotating shaft as the dividing point. The short arm is connected to the small shaft by a spring, and a sensing plate is provided at the end of the short arm. A proximity switch is provided on the movable support at the corresponding position of the sensing plate. Winding centering can be achieved by performing two consecutive operations using this device. This solution uses a mechanical swing arm and a sensor plate for height detection, which changes the original production configuration. However, it suffers from high maintenance costs for the mechanical swing arm and sensor plate, and requires two operations to complete, increasing time costs and reducing work efficiency. Summary of the Invention

[0007] In view of the deficiencies in the prior art, the purpose of this invention is to provide a method and system for detecting the centering deviation of the upper winding.

[0008] A method for detecting centering deviation of an upper roll according to the present invention includes: an illumination device and a camera device; the illumination device is used to illuminate the object to be detected, highlighting the outline and surface features of the object; the camera device is used to acquire a front view of the object to be detected.

[0009] Step S1: Construct and train the centering deviation detection model; set up lighting and camera equipment, and acquire images at intervals;

[0010] Step S2: Input the image into the deviation detection model, calculate the center coordinates of the ellipse in the image, and then calculate the pixel deviation value; based on the pixel deviation value, obtain the actual offset distance;

[0011] Step S3: Based on the data obtained in step S2, derive the trajectory of the ellipse and draw it on the image;

[0012] Step S4: Determine whether the actual offset distance is greater than or equal to the threshold; if the result is yes, store the current data and issue a warning signal; if the result is no, store the current data.

[0013] Preferably, step S1 includes:

[0014] Step S1.1: Load dataset function;

[0015] Step S1.2: Adjust the parameters based on the dataset function;

[0016] Step S1.3: Based on the parameters, construct and train the alignment deviation detection model;

[0017] In step S1.1, the data is preprocessed;

[0018] The preprocessing includes: image labeling, image resizing, and data augmentation;

[0019] The dataset function is torch.utils.data.DataLoader;

[0020] In step S1.2, the parameters include: the dimension of the input data, the learning rate, the number of training iterations, and the batch size.

[0021] In step S1.3, the loss function of the deviation detection model is cross-entropy; the mathematical expression of the loss function is:

[0022]

[0023] in, For loss function, For real labels, These are the model's predicted values. This represents the number of samples.

[0024] Preferably, in step S1, after the steel coil reaches the saddle position, an image is captured every 0.5 seconds;

[0025] Step S2 includes:

[0026] Step S2.1: Convert the image output by the deviation detection model into a binary image;

[0027] Step S2.2: Construct a planar coordinate system. Take the midpoint between the upper and lower vertices of the inner circle of the binary graph as the y-axis coordinate point of the circle center. Take the midpoint between the left and right vertices of the inner circle of the binary graph as the x-axis coordinate point of the circle center to obtain the coordinates of the circle center.

[0028] Step S2.3: Based on the two stated center points, determine the equation of the line containing the standard center point;

[0029] Step S2.4: Substitute the x-coordinate of the standard circle center into the equation of the line containing the standard circle center to obtain the y-coordinate corresponding to the x-coordinate on the line; subtract the y-coordinate of the standard circle center from the y-coordinate of the standard circle center to obtain the pixel deviation value;

[0030] Step S2.5: Based on the conversion ratio between pixel values ​​and spatial values, convert the pixel deviation value into the corresponding actual offset distance;

[0031] In step S2.3, the mathematical expression of the equation of the line containing the center of the circle is:

[0032]

[0033] in, The y-coordinate of the standard circle's center; Represents the y-axis coordinate of the second circle; Represents the y-coordinate of the first circle; Represents the x-axis coordinate of the first circle; Represents the x-axis coordinate of the second circle; The x-coordinate of the standard circle's center;

[0034] In step S2.5, the conversion ratio between pixel values ​​and spatial values ​​is expressed mathematically as follows:

[0035] K=L÷N

[0036] Where K represents the conversion ratio between pixel value and spatial value, L represents the actual width of the steel coil, and N represents the number of pixels in the width.

[0037] Preferably, in step S3, the ellipse is fitted using the cv2.fitEllipse function provided by OpenCV to obtain the corresponding ellipse parameters; the ellipse is then drawn onto the original image based on the ellipse parameters, i.e., the image, to obtain the trajectory of the ellipse.

[0038] The ellipse parameters include: the coordinates of the ellipse's center point, the lengths of its major and minor axes, and its tilt angle;

[0039] In step S3, the ellipse parameters are calculated according to the ellipse regression equation, and the trajectory of the ellipse is drawn on the original image.

[0040] The mathematical expression for the elliptic regression calculation equation is:

[0041] + =1

[0042] Where k represents the ordinate of the center of the inner circle, a represents the length between the top and bottom vertices of the inner circle, and b represents the length between the left and right vertices of the inner circle.

[0043] Preferably, in step S4, it is determined whether the actual offset distance exceeds a threshold. If the result is yes, an alarm is issued; if the result is no, the image is saved.

[0044] The threshold is set based on the trajectory of the ellipse;

[0045] The threshold is 15.

[0046] A detection system for upper roll alignment deviation provided by the present invention includes: an illumination device and a camera device; the illumination device is used to illuminate the object to be detected, highlighting the outline and surface features of the object to be detected; the camera device is used to acquire a front view of the object to be detected.

[0047] Module M1: Construct and train the centering deviation detection model; arrange lighting and camera equipment, and acquire images at intervals;

[0048] Module M2: Input the image into the deviation detection model, determine the center coordinates of the ellipse in the image, and then calculate the pixel deviation value; based on the pixel deviation value, obtain the actual offset distance;

[0049] Module M3: Based on the data obtained from module M2, derive the trajectory of the ellipse and draw it on the image;

[0050] Module M4: Determines whether the actual offset distance is greater than or equal to the threshold; if the result is yes, it stores the current data and issues a warning signal; if the result is no, it stores the current data.

[0051] Preferably, module M1 includes:

[0052] Module M1.1: Function for loading datasets;

[0053] Module M1.2: Adjusts parameters based on the dataset function;

[0054] Module M1.3: Based on parameters, construct and train a training centering deviation detection model;

[0055] In module M1.1, the data is preprocessed;

[0056] The preprocessing includes: image labeling, image resizing, and data augmentation;

[0057] The dataset function is torch.utils.data.DataLoader;

[0058] In module M1.2, the parameters include: the dimension of the input data, the learning rate, the number of training iterations, and the batch size.

[0059] In module M1.3, the loss function of the deviation detection model is cross-entropy; the mathematical expression of the loss function is:

[0060]

[0061] in, For loss function, For real labels, These are the model's predicted values. This represents the number of samples.

[0062] Preferably, in module M1, after the steel coil arrives at the saddle position, an image is captured every 0.5 seconds;

[0063] The module M2 includes:

[0064] Module M2.1: Converts the image output by the deviation detection model into a binary image;

[0065] Module M2.2: Construct a planar coordinate system, take the midpoint between the upper and lower vertices of the inner circle of the binary graph as the y-axis coordinate point of the circle center, and take the midpoint between the left and right vertices of the inner circle of the binary graph as the x-axis coordinate point of the circle center to obtain the coordinates of the circle center;

[0066] Module M2.3: Based on the two stated center points, determine the equation of the line containing the standard center point;

[0067] Module M2.4: Substitute the x-coordinate of the standard circle center into the equation of the line containing the standard circle center to obtain the y-coordinate on the line corresponding to the x-coordinate; subtract the y-coordinate of the standard circle center from the y-coordinate of the standard circle center to obtain the pixel deviation value;

[0068] Module M2.5: Based on the conversion ratio between pixel values ​​and spatial values, convert the pixel deviation value into the corresponding actual offset distance.

[0069] In module M2.3, the mathematical expression for the equation of the line containing the center of the circle is:

[0070]

[0071] in, The y-coordinate of the standard circle's center; Represents the y-axis coordinate of the second circle; Represents the y-coordinate of the first circle; Represents the x-axis coordinate of the first circle; Represents the x-axis coordinate of the second circle; The x-coordinate of the standard circle's center;

[0072] In module M2.5, the conversion ratio between pixel values ​​and spatial values ​​is expressed mathematically as follows:

[0073] K=L÷N

[0074] Where K represents the conversion ratio between pixel value and spatial value, L represents the actual width of the steel coil, and N represents the number of pixels in the width.

[0075] Preferably, in module M3, the cv2.fitEllipse function provided by OpenCV is used to fit an ellipse and obtain the corresponding ellipse parameters; the ellipse is then drawn onto the original image based on the ellipse parameters, i.e., the image, to obtain the trajectory of the ellipse.

[0076] The ellipse parameters include: the coordinates of the ellipse's center point, the lengths of its major and minor axes, and its tilt angle;

[0077] In module M3, the ellipse parameters are calculated based on the elliptic regression equation, and the trajectory of the ellipse is drawn on the original image.

[0078] The mathematical expression for the elliptic regression calculation equation is:

[0079] + =1

[0080] Where k represents the ordinate of the center of the inner circle, a represents the length between the top and bottom vertices of the inner circle, and b represents the length between the left and right vertices of the inner circle.

[0081] Preferably, in module M4, it is determined whether the actual offset distance exceeds a threshold. If the result is yes, an alarm is issued; if the result is no, the image is saved.

[0082] The threshold is set based on the trajectory of the ellipse;

[0083] The threshold is 15.

[0084] Compared with the prior art, the present invention has the following beneficial effects:

[0085] 1. This invention uses visual recognition to perform high-precision detection of the centering height of steel coils, avoiding the occurrence of coil flipping and reducing the labor intensity of operators;

[0086] 2. This invention can replace the centering grating, reducing equipment configuration and production maintenance.

[0087] 3. This invention has strong applicability and can be used in all cold rolling coiling processes, thus reducing development costs. Attached Figure Description

[0088] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0089] Figure 1 This is a schematic diagram of the segmentation network structure provided by the present invention;

[0090] Figure 2 A schematic diagram of an ellipse used in regression calculations provided by this invention;

[0091] Figure 3 A schematic diagram for measuring the inner diameter width of a steel coil provided by the present invention;

[0092] Figure 4 A schematic diagram of the centering extension line of the steel coil core provided by the present invention. Detailed Implementation

[0093] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0094] This solution includes: an inner ring segmentation detection model for steel coils.

[0095] This embodiment is based on the deep learning U2Net model, with the addition of the CBAM (Convolutional Block Attention Module) to the model structure. It also collects data from specific scenarios, manually annotates the inner ring of the steel coil, and uses the annotated images to perform supervised optimization training on the model to adjust the model parameters and optimize the detection performance of the model in specific application scenarios.

[0096] Ellipse regression calculation. Based on the inner circle of a correctly wound steel coil, its inner circle is approximated as an ellipse. The major and minor axes, foci, and ellipse regression equation of the inner circle are calculated. Combined with the results of the segmentation detection model, the trajectory of the inner circle edge curve is calculated using the equation. In cases of insufficient light or poor detection effect of the segmentation model, this method can significantly improve the accuracy of edge drawing.

[0097] A method for converting pixel values ​​to spatial values. By constructing a non-linear mapping function between pixel length and actual length in an image through methods such as mapping pixel correspondence in actual space, pixel length can be converted into actual length.

[0098] Centering extension line. In the same coiling application scenario, the center of the inner circle of the correctly coiled steel coil in the acquired image is statistically analyzed to obtain a set of coordinate points. The line segment obtained by connecting these points is extended appropriately to form the centering extension line of the steel coil. Subsequent deviation comparisons are based on this line for deviation calculation.

[0099] Deviation alarm interlocking system. After detecting the inner circle in the image, the center point of the inner circle is calculated, and its ordinate is compared with the ordinate of the point with the same abscissa on the centering extension line to obtain the pixel deviation value. Then, according to the conversion function between pixel value and spatial value, it is converted into a deviation value in actual space, which is compared with the set threshold to determine whether an alarm needs to be triggered, and the deviation value is sent to the control system for interlocking.

[0100] The embodiment uses the Hunan Hongwang six-strand rolling mill coiling unit as an example. According to the present invention, a method for detecting coiling alignment deviation includes:

[0101] Step 1: Manually label the inner ring of the steel coil in the collected images pixel by pixel;

[0102] The inner circle segmentation and detection algorithm was optimized and trained using labeled data; the trained algorithm was then deployed on a server equipped with a graphics card.

[0103] The steps for optimizing and training the segmentation detection algorithm include:

[0104] The segmentation detection algorithm is an AI segmentation algorithm based on deep learning.

[0105] Step 1.1: Preprocess the labeled data;

[0106] The preprocessing includes: image resizing, flipping, and other data enhancement methods;

[0107] Step 1.2: Install training environments such as Python and PyTorch on the training server equipped with a graphics card;

[0108] Step 1.3: Write the model training script;

[0109] Specifically, step 1.3 includes:

[0110] Step 1.3.1: Load the dataset function; specifically, the function is torch.utils.data.DataLoader;

[0111] Step 1.3.2: Based on the dataset function, adjust the parameters; the parameters include: batch size, which is set to 8 in this model; dimension of the input data, which is set to 4 in this model; learning rate, which is set to 1×e-5 in this model; and the number of training iterations, which is set to 100 in this model.

[0112] Step 1.3.3: The loss function of the model is cross-entropy, and the mathematical formula is:

[0113]

[0114] in, For real labels, These are the model's predicted values. This represents the number of samples.

[0115] Step 1.4: Perform inference on the test set using the trained model to test the training effect of the model;

[0116] Step 2: Place an industrial camera and light source diagonally above the upper roll position;

[0117] The industrial camera includes: a Dalsa C4030 industrial camera;

[0118] The industrial camera is horizontally mounted diagonally in front of the inspection station, about 5 meters away from the object being inspected. It is used to acquire a front view of the object being inspected. Its high resolution and high frame rate can clearly capture the details and rapidly changing features of the object, providing high-quality raw data for subsequent image processing and analysis.

[0119] The light source includes: a Yaming 4000W circular LED light source;

[0120] The light source, mounted above the camera, illuminates the object being detected horizontally, providing uniform and stable lighting to the object, reducing shadows and reflections, thereby highlighting the outline and surface features of the object, enhancing image contrast and clarity, and helping to improve the accuracy and reliability of the detection.

[0121] Step 3: Once the steel coil reaches the saddle position, automatically capture images every 0.5 seconds;

[0122] Specifically, by connecting an industrial camera to the acquisition server, calling the camera's SDK, and setting acquisition parameters such as frequency and exposure, the acquired images are transmitted to the server via optical fiber for subsequent retrieval and storage. The camera itself only stores camera parameters and does not store images.

[0123] Step 4: Input the image captured by the industrial camera into the segmentation and detection algorithm described above, and then obtain the center coordinates of the inner circle based on the prediction result image output by the algorithm.

[0124] The steps and process for obtaining the coordinates of the center of a circle include:

[0125] Step 4.1: Convert the segmentation result image into a binary image with a threshold of 0.5, where the background region value is 0 and the inner region value is 1.

[0126] Step 4.2: Take the midpoint between the top and bottom points of the inner circle of the binary graph as the y-axis coordinate of the circle's center; take the midpoint between the left and right points of the inner circle, i.e., the midpoint between the left and right endpoints of the inner circle, as the x-axis coordinate of the circle's center, thus obtaining the coordinates of circle center 1, i.e., the first circle's center. , );

[0127] Step 4.3: Select another captured image to obtain the coordinates of the center point 2, which is the second center point of the circle. , );

[0128] Step 4.4: Based on the coordinates of the two center points, obtain the equation of the line containing the standard center of the circle:

[0129]

[0130] Step 5: Obtain the center coordinates C of the image collected in actual production according to the above steps, and calculate the pixel deviation value using the standard straight line equation; the center coordinates C are the standard center coordinates.

[0131] The specific steps to obtain the pixel deviation value include:

[0132] Step 5.1: Substitute the x-coordinate of the circle's center into the equation formula of the line containing the standard circle's center to calculate the y-coordinate corresponding to the x-coordinate on the line;

[0133] Step 5.2: Subtract the calculated ordinate value of the circle center from the ordinate value of the circle center C to obtain the pixel deviation value;

[0134] Step 6: Use the pixel value to spatial value conversion function to convert the pixel deviation value obtained in the previous step into the corresponding actual offset distance;

[0135] Step 6.1: Select a clear image. Based on the actual width of the steel coil and the number of pixels in the image, calculate the conversion ratio between pixel values ​​and spatial values ​​by dividing the actual width by the number of pixels in the image. The mathematical expression is:

[0136] K=L÷N

[0137] Where K represents the conversion ratio between pixel value and spatial value; the value of K varies depending on different winding machines and production environments; L represents the actual width of the steel coil; and N represents the number of pixels in the width.

[0138] Step 6.2: Multiply the pixel deviation value obtained in the previous step by the conversion ratio to obtain the corresponding actual pixel offset distance; this is used to determine whether it exceeds the specified error range, and if it does, an alarm signal will be issued.

[0139] Step 7: Call the elliptic regression calculation equation. The length between the uppermost and lowermost points of the inner circle is calculated based on the algorithm output image obtained in Step 4.1. The trajectory of the ellipse is then drawn on the original image and saved to the storage server.

[0140] The equation for elliptic regression is expressed mathematically as follows:

[0141] + =1

[0142] Where (h,k) represents the center of the inner circle, a represents the length between the top and bottom points of the inner circle, and b represents the length between the leftmost and rightmost points of the inner circle.

[0143] Step 7.1: Use the cv2.findContours function from the OpenCV package in Python to extract contour data from the input segmentation result image;

[0144] Step 7.2: Use the cv2.fitEllipse function provided by OpenCV to fit the ellipse and obtain the returned ellipse parameters;

[0145] The ellipse parameters include: the coordinates of the ellipse's center point, the lengths of its major and minor axes, and its tilt angle.

[0146] In other words, the cv2.fitEllipse function returns a tuple containing the key parameters of the ellipse, including: the coordinates of the ellipse's center point, the lengths of its major and minor axes, and the tilt angle.

[0147] Step 7.3: Based on the obtained ellipse parameters, use the cv2.ellipse function provided by OpenCV to draw an ellipse onto the original image. Then, display the result on the front-end interface. Compared to the output obtained directly from the segmentation model, its edges are clearer and more complete, making it relatively more aesthetically pleasing.

[0148] Step 8: Compare the actual offset distance with the threshold to determine whether a warning signal needs to be issued;

[0149] Specifically, if the offset distance is greater than or equal to the threshold, the result will be stored in the database and a warning signal will be issued;

[0150] If the offset distance is less than the threshold, no alarm signal will be sent, and the result will only be stored in the database.

[0151] Through tracking and observing the test results and communicating with the client, a threshold of 15 was set.

[0152] Step 9: Continue to repeat step 3 to achieve 24-hour uninterrupted centering deviation detection.

[0153] The present invention also provides a system for detecting upper winding centering deviation. The system for detecting upper winding centering deviation can be implemented by executing the process steps of the method for detecting upper winding centering deviation. That is, those skilled in the art can understand the method for detecting upper winding centering deviation as a preferred embodiment of the system for detecting upper winding centering deviation.

[0154] A detection system for upper roll alignment deviation provided by the present invention includes: an illumination device and a camera device; the illumination device is used to illuminate the object to be detected, highlighting the outline and surface features of the object to be detected; the camera device is used to acquire a front view of the object to be detected.

[0155] Module M1: Construct and train the centering deviation detection model; arrange lighting and camera equipment, and acquire images at intervals;

[0156] Module M2: Input the image into the deviation detection model, determine the center coordinates of the ellipse in the image, and then calculate the pixel deviation value; based on the pixel deviation value, obtain the actual offset distance;

[0157] Module M3: Based on the data obtained from module M2, derive the trajectory of the ellipse and draw it on the image;

[0158] Module M4: Determines whether the actual offset distance is greater than or equal to the threshold; if the result is yes, it stores the current data and issues a warning signal; if the result is no, it stores the current data.

[0159] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0160] In the description of this application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0161] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method of detecting a winding-on centering deviation, characterized in that Comprise: Step S1: build and train the centering deviation detection model, interval acquisition image; Step S2: the image is input into the centering deviation detection model, the center coordinates of the ellipse in the image are calculated, and then the pixel deviation value is calculated; according to the pixel deviation value, the actual offset distance is obtained; Step S3: according to the data obtained in step S2, the trajectory of the ellipse is obtained, and is drawn on the image; Step S4: determine whether the actual offset distance is greater than or equal to the threshold value; The result is yes, store the current data, and issue a warning signal; The result is no, then store the current data; In the step S1, comprising: Step S1.1: load the dataset function; Step S1.2: adjust the parameters based on the dataset function; Step S1.3: build and train the centering deviation detection model based on the parameters; In the step S1.1, pre-process the data; The pre-processing includes: labeling the image, adjusting the image size and data enhancement; The dataset function is torch.utils.data.DataLoader; In the step S1.2, the parameters include: the dimension of the input data, the learning rate, the number of training iterations and the batch size value; In the step S1.3, the loss function of the centering deviation detection model is cross entropy; the mathematical expression of the loss function is: wherein, is a loss function, is a true label, is a model prediction value, is a number of samples; In the step S1, arrange the lighting device and the camera device to collect images at intervals; the lighting device is used to irradiate the detected object, highlight the outline and surface features of the detected object; the camera device is used to collect the front view of the detected object; In the step S1, after the steel coil reaches the saddle position, images are collected every 0.5 seconds; In the step S2, comprising: Step S2.1: convert the image output by the centering deviation detection model into a binary image; Step S2.2: build a plane coordinate system, take the middle value of the upper vertex and the lower vertex of the inner circle of the binary image as the y-axis coordinate point of the center, and take the middle value of the left vertex and the right vertex of the inner circle of the binary image as the x-axis coordinate point of the center, to obtain the center coordinates; Step S2.3: according to two said center, the straight line equation of the standard center is obtained; Step S2.4: substitute the horizontal coordinate of the standard center into the straight line equation of the standard center to obtain the vertical coordinate corresponding to the horizontal coordinate on the straight line; subtract the vertical coordinate of the standard center from the vertical coordinate to obtain the pixel deviation value; Step S2.5: according to the conversion ratio of pixel value and space value, the pixel deviation value is converted into the corresponding actual offset distance; In the step S2.3, the mathematical expression of the straight line equation of the center is: wherein represents the y coordinate of the center of the standard circle; represents the y axis coordinate of the second circle; represents the y axis coordinate of the first circle; represents the x axis coordinate of the first circle; represents the x axis coordinate of the second circle; represents the x coordinate of the center of the standard circle; In the step S2.5, the mathematical expression of the conversion ratio of pixel value and space value is: K=L÷N Wherein, K represents the conversion ratio of pixel value and space value, L represents the actual width of the steel coil; N represents the number of width pixels; In the step S3, an ellipse is fitted by a cv2.fitEllipse function provided by OpenCV to obtain corresponding ellipse parameters; and an ellipse is drawn to the original image according to the ellipse parameters; The ellipse parameters include: an ellipse center point coordinate, a major and minor axis length, and an inclination angle; In the step S3, ellipse parameters are calculated according to an ellipse regression calculation equation, and a trajectory of the ellipse is drawn on the original image; The ellipse regression calculation equation has a mathematical expression as follows: Wherein, k represents a longitudinal coordinate of a center of an inner circle, h represents a horizontal coordinate of the center of the inner circle, a represents a length between upper and lower vertices of the inner circle, and b represents a length between left and right vertices of the inner circle. In the step S4, the threshold value is set based on the trajectory of the ellipse; and the threshold value is 15.

2. A system for detecting a winding on-center deviation, characterized by Comprise: Module M1: construct and train a centering deviation detection model, and collect images at intervals; Module M2: input the images into the centering deviation detection model, calculate a center point coordinate of an ellipse in the images, and further calculate a pixel deviation value; and obtain an actual offset distance according to the pixel deviation value; Module M3: obtain the trajectory of the ellipse according to data obtained by the module M2, and draw the trajectory on the images; Module M4: judge whether the actual offset distance is greater than or equal to a threshold value; If the result is yes, store current data and issue a warning signal; If the result is no, store the current data; In the module M1, comprise: Module M1.1: load a data set function; Module M1.2: adjust parameters based on the data set function; Module M1.3: construct and train a centering deviation detection model based on the parameters; In the module M1.1, pre-process data; The pre-processing comprises: labeling images, adjusting image sizes, and data enhancement; The data set function is torch.utils.data.DataLoader; In the module M1.2, the parameters comprise: a dimension of input data, a learning rate, a training iteration number, and a batch size value; In the module M1.3, a loss function of the centering deviation detection model is cross entropy; and the loss function has a mathematical expression as follows: wherein, is a loss function, is a true label, is a model prediction, is a number of samples; In the module M1, arrange a lighting device and a camera device to collect images at intervals; the lighting device is used to irradiate a detected object to highlight a contour and surface features of the detected object; and the camera device is used to collect a front view of the detected object; In the module M1, collect images every 0.5 seconds after a steel coil reaches a saddle position; In the module M2, comprise: Module M2.1: convert images output by the centering deviation detection model into binary images; Module M2.2: construct a plane coordinate system, take a middle value of upper and lower vertices of an inner circle of the binary images as a y-axis coordinate point of a center, and take a middle value of left and right vertices of the inner circle as an x-axis coordinate point of the center to obtain a center coordinate; Module M2.3: calculate a straight line equation of a standard center according to two center points; Module M2.4: substituting the horizontal coordinate of the standard circle center into the equation of the straight line where the standard circle center is located to obtain the vertical coordinate of the straight line corresponding to the horizontal coordinate; subtracting the vertical coordinate from the vertical coordinate of the standard circle center to obtain a pixel deviation value; Module M2.5: converting the pixel deviation value into a corresponding actual offset distance according to a conversion ratio of pixel value and spatial value; In the module M2.3, the mathematical expression of the equation of the straight line where the circle center is located is: wherein represents the y coordinate of the center of the standard circle; represents the y axis coordinate of the second circle; represents the y axis coordinate of the first circle; represents the x axis coordinate of the first circle; represents the x axis coordinate of the second circle; represents the x coordinate of the center of the standard circle; In the module M2.5, the conversion ratio of the pixel value and the spatial value has a mathematical expression: K = L ÷ N Wherein, K represents the conversion ratio of the pixel value and the spatial value, L represents the actual width of the steel coil; N represents the number of width pixels; In the module M3, an ellipse is fitted by using the cv2.fitEllipse function provided by OpenCV to obtain corresponding ellipse parameters; and the ellipse is drawn on the original image according to the ellipse parameters; The ellipse parameters include: the coordinates of the center of the ellipse, the lengths of the major and minor axes, and the tilt angle; In the module M3, the ellipse parameters are calculated according to an ellipse regression calculation equation, and the trajectory of the ellipse is drawn on the original image; The mathematical expression of the ellipse regression calculation equation is: Wherein, k represents the vertical coordinate of the inner circle center, h represents the horizontal coordinate of the inner circle center, a represents the length between the upper and lower vertices of the inner circle, and b represents the length between the left and right vertices of the inner circle; In the module M4, the threshold is set based on the trajectory of the ellipse; and the threshold is 15.

Images