Detection device for rapidly measuring deviation of strip steel by adopting AI visual identification technology

Abstract

The invention discloses a detection device for rapidly measuring strip steel deviation by adopting an AI visual identification technology, and relates to the technical field of strip steel processing, the detection device comprises a device support, a linear array camera and a visual identification system, the visual identification system comprises an image acquisition unit, an image analysis unit, an environment compensation unit, a deviation prediction unit and a result output unit; real-time image data of a strip steel edge area are collected in real time through a linear array camera, the real-time image data are analyzed through a deep learning model, strip steel edge pixel-level coordinates and image abnormal feature points are output, then distance data and inclination angle data are obtained, fusion correction is carried out through output of a Kalman filter and an AI processing module, and the accuracy of strip steel edge detection is improved. And the correction unit is used for obtaining corrected image data, establishing a strip steel deviation characteristic model based on a convolutional neural network, inputting the corrected image data into the strip steel deviation characteristic model, extracting strip steel deviation characteristics, calculating a strip steel deviation probability and sending the strip steel deviation probability to the result output unit.

Description

Technical Field

[0001] The present invention relates to the technical field of strip steel processing, and in particular to a detection device for quickly measuring the deviation of strip steel by using AI visual recognition technology. Background Art

[0002] Strip steel is a kind of steel plate that is thinner, narrower and has a very long length. It is usually supplied in coils. Compared with steel plates, strip steel has the advantages of high dimensional accuracy, good surface quality and convenient use. Strip steel is widely used in the production of welded steel pipes, cold-formed steel billets, and the manufacture of products such as bicycle frames, rims, clamps, washers, spring sheets, cable armors, saw blades, knife blades, and packing iron sheets. The production of strip steel is a complex process involving multiple processes, mainly including raw material processing, smelting, rolling, quenching, tempering and annealing, etc.; During the hot rolling production process, it is very easy for the center of the strip steel to deviate from the center of the roller table, which is commonly referred to as deviation in industrial production. Many domestic and foreign scholars have conducted relevant research on the influencing factors causing deviation and the deviation correction methods. For example: strip steel shape defects, the influence of roller geometry, the axial movement of conveyor rollers, and the process parameters of the unit. Most hot rolling production lines are equipped with side guide plate devices in front of the coiler to clamp the strip steel and align it with the pinch rolls of the coiler for deviation correction, but it can only correct the strip steel entering the side guide plate part and cannot completely avoid the deviation phenomenon. When the hot rolled strip steel deviates before coiling, it mainly causes strip steel surface defects and hole defects formed on the surface of the dropped strip steel; Traditional detection methods rely on photoelectric sensors or manual inspections, and have problems such as insufficient accuracy, slow response speed, and being easily affected by the environment (such as dust and low light). Traditional machine vision methods (such as edge detection and object detection) have insufficient robustness under complex working conditions and are prone to false alarms or missed detections; In view of the above technical defects, a solution is now proposed. Summary of the Invention

[0003] The purpose of the present invention is to: establish a strip steel deviation feature model based on a convolutional neural network, input the corrected image data into the strip steel deviation feature model, extract the strip steel deviation features, and calculate the strip steel deviation probability. During the strip steel conveying process, obtain the strip steel image through a line array camera, and realize strip steel deviation prediction through high-precision image processing and visual recognition, so as to be able to intervene in time before the strip steel deviates and fundamentally solve the production losses caused by strip steel deviation.

[0004] In order to achieve the above purpose, the present invention adopts the following technical solution: a detection device for quickly measuring the deviation of strip steel by using AI visual recognition technology, including a device support, a line array camera and a visual recognition system. The device support is fixedly arranged on the top surface of the frame, and an angle adjustment mechanism is installed on the outer surface of the device support. Two line array cameras are respectively connected to the outer surface of the angle adjustment mechanism; The visual recognition system includes an image acquisition unit, an image analysis unit, an environment compensation unit, a deviation prediction unit, and a result output unit; The image acquisition unit collects real-time image data of the strip edge area through a linear array camera in real time. Its baseline distance is dynamically matched with the strip width, and image spatio-temporal alignment is achieved through high-frequency pulse synchronous control; The image analysis unit includes an AI processing module and an image analysis module. The AI processing module is used to integrate an edge computing unit and has a built-in deep learning model. The deep learning model is implemented based on a fusion architecture of a residual network and an attention mechanism. The input is the real-time image data output by the image acquisition unit, and the strip edge pixel-level coordinates and image abnormal feature points are output and sent to the image analysis module; The image analysis module is used to obtain a preset threshold of the number of feature points. If the number of image abnormal feature points is greater than or equal to the threshold of the number of feature points, an image correction signal is generated and sent to the environment compensation unit; The environment compensation unit includes a laser ranging unit and an image correction module. The laser ranging unit is used to monitor the distance between the strip and the linear array camera and the inclination angle of the strip conveying roller table in real time, and the obtained distance data and inclination angle data are integrated into a data set and sent to the image correction module; The image correction module is used to obtain and process the image correction signal, fuse and correct the data set with the output of the AI processing module through a Kalman filter, eliminate the influence of strip vibration and mechanical installation errors, and send the corrected image data to the deviation prediction unit; The deviation prediction unit is used to obtain the historical production data of the hot-rolled strip and perform preprocessing, construct a hot-rolled strip image data set as a training sample, establish a strip deviation feature model based on a convolutional neural network, input the corrected image data into the strip deviation feature model, extract the strip deviation features, and calculate the strip deviation probability and send it to the result output unit.

[0005] Furthermore, the device bracket includes self-aligning ball bearings and vertical support shafts. The two vertical support shafts are symmetrically arranged along the strip width direction on the top surface of the adjacent rack of the conveying path. The two self-aligning ball bearings are respectively connected to the top surface of the vertical support shafts. A horizontal support shaft is jointly installed between the two self-aligning ball bearings, and an angle adjustment mechanism is installed on the outer surface of the horizontal support shaft.

[0006] Furthermore, the angle adjustment mechanism includes a fixed seat and a polarization filter assembly. The fixed seat is fixed on the end surface of the horizontal support shaft, and two driving motors are fixed inside the fixed seat. The two linear array cameras are respectively fixed on the outer surface of the output ends of the driving motors.

[0007] Furthermore, the polarization filter assembly includes a pressure sensor and an elastic sensing block, and dynamic adjustment of the polarization angle is achieved through an electric rotating mechanism to eliminate reflective interference from the surface of the steel strip. The outer surface of the driving motor is sleeved with a gear ring, and the outer surface of the fixed seat is provided with a long sliding groove. A movable rack is movably connected to the inner wall of the long sliding groove, and the movable rack is meshed with the gear ring. The pressure sensor is fixed to the end surface of the movable rack, and the elastic sensing block is fixed to the outer surface of the fixed seat, and the pressure sensor and the elastic sensing block are in contact with each other.

[0008] Furthermore, the image acquisition unit also includes a high-frequency pulse LED array and a light source controller; The light emission frequency of the high-frequency pulse LED array is synchronized with the line frequency of the linear array camera; The light source controller has a built-in closed-loop feedback circuit to adjust the output power in real time according to the ambient light intensity to ensure that the image signal-to-noise ratio is less than or equal to 40dB.

[0009] Furthermore, the specific process of outputting the pixel-level coordinates of the strip edge and the image abnormal feature points is as follows: S101. Build a deep learning model based on the improved YOLOv5 architecture and embed a deformable convolutional layer in the backbone network to adapt to the geometric deformation of the strip edge. S102, introducing a direction classification branch in the detection head part, and synchronously outputting the deviation direction determination result; S103, the loss function is designed as a weighted combination of the deviation regression loss and the direction classification loss, and the weight coefficient is automatically determined by Bayesian optimization; S104, synthetic data enhancement technology is used in the training of the deep learning model, including simulation scenes of dynamic lighting, dust occlusion and changes in strip surface texture; S105. Input the real-time image data into the deep learning model, extract the edge feature points of the steel strip, construct the three-dimensional point cloud data, obtain the pixel-level coordinates of the edge of the steel strip, and perform feature comparison based on the standard steel strip image to obtain the image abnormal feature points.

[0010] Furthermore, the specific process of outputting the strip deviation probability is as follows: S201, obtaining historical production data of hot-rolled steel strips, wherein the historical production data includes standard images of steel strips, conveying paths of steel strips, and straight lines of steel strip edges, integrating historical operation data of road rollers as training samples, and splitting the training samples into a training set and a test set in a ratio of 8:2; S202, constructing a strip steel bias feature model based on a convolutional neural network, downloading a weight file and loading it onto a corresponding network to initialize migration network parameters; S203. Modify the last fully connected layer of the network, keep the input unchanged, set the output as the straight line of the strip edge, initialize the weights of the last layer, use the gradient descent algorithm for learning, and adopt fixed-step decay to optimize the training parameters. Retrain the entire network to obtain the strip deviation feature model; S204. During the training process, randomly and non-repeatedly extract small batches of training samples from the training set for training. After all the training samples are drawn, it is regarded as one training cycle. Iterate for a certain number of cycles to complete the training and obtain the strip deviation feature model; S205. Input the corrected image data into the strip deviation feature model to output the straight line of the strip edge. Use the Hough transform line detection algorithm to extract the potential roller paths therein, and then predict the deviation warning and deviation amount based on the position coordinates of the strip edge straight line screened between the standard strip image and the roller path to obtain the strip deviation probability.

[0011] Further, the result output unit is used to obtain a preset strip deviation judgment interval, and make a result judgment based on the strip deviation probability. If the strip deviation probability exceeds the defined range of the strip deviation judgment interval, a strip deviation signal is generated and sent to the industrial control system for visual display and voice reminder.

[0012] Further, it also includes a control interface unit, which is used to provide a standardized communication protocol interface, output the deviation detection result to the industrial control system at a refresh rate greater than or equal to 1 kHz, support the adaptive adjustment function of PID control parameters, and dynamically adjust the response threshold of the deviation correction actuator according to the deviation amount.

[0013] Further, the environment compensation unit also includes a temperature compensation module and a vibration compensation module, where: The temperature compensation module uses a thermistor array to monitor the temperature field distribution in the camera installation area in real time, and corrects the thermal expansion and contraction errors of the optical elements through the look-up table method; The vibration compensation module uses an accelerometer to collect the vibration spectrum of the mechanical structure, and suppresses the interference signals in the resonance frequency band through a notch filter.

[0014] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are: The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology collects real-time image data of the edge area of the strip steel through a line array camera, analyzes the real-time image data through a deep learning model, outputs the pixel-level coordinates of the strip steel edge and the abnormal feature points of the image, and then obtains the distance data and inclination data. The Kalman filter is used to fuse and correct the output of the AI processing module to obtain the corrected image data. A strip steel deviation feature model is established based on a convolutional neural network. The corrected image data is input into the strip steel deviation feature model to extract the strip steel deviation features, calculate the strip steel deviation probability and send it to the result output unit. During the strip steel conveying process, the strip steel image is obtained through the line array camera, and the strip steel deviation prediction is realized through high-precision image processing and vision recognition, which can intervene in time before the strip steel deviates and fundamentally solve the production loss caused by the strip steel deviation. Brief Description of the Drawings

[0015] Figure 1 Shows the overall external structure schematic diagram of the present invention; Figure 2 Shows the overall external structure schematic diagram of another angle of the present invention; Figure 3 Shows the structure schematic diagram of the vision recognition system of the present invention; Legend: 1. Line array camera; 2. Vertical support shaft; 3. Self-aligning ball bearing; 4. Horizontal support shaft; 5. Fixed seat; 6. Driving motor; 7. Gear ring; 8. Sliding long groove; 9. Moving rack; 10. Pressure sensor; 11. Elastic induction block. Detailed Description of the Embodiments

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0017] Embodiment 1:

[0018] As Figure 1 - Figure 2 shown, a detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology includes a device bracket, a line array camera 1 and a vision recognition system. The device bracket is fixedly arranged on the top surface of the frame, and an angle adjustment mechanism is installed on the outer surface of the device bracket. Two line array cameras 1 are respectively connected to the outer surface of the angle adjustment mechanism; The device support includes a spherical roller bearing 3 and a vertical support shaft 2. The two vertical support shafts are symmetrically arranged along the width direction of the strip steel on the top surface of the adjacent frames of the conveying path. The two spherical roller bearings 3 are respectively connected to the top surface of the vertical support shaft 2. A horizontal support shaft 4 is commonly installed between the two spherical roller bearings 3. An angle adjustment mechanism is installed on the outer surface of the horizontal support shaft 4.

[0019] The angle adjustment mechanism includes a fixed seat 5 and a polarization filter component. The fixed seat 5 is fixed on the end surface of the horizontal support shaft 4. Two driving motors 6 are fixedly installed inside the fixed seat 5. The two linear array cameras 1 are respectively fixed on the outer surface of the output end of the driving motors 6.

[0020] The polarization filter component includes a pressure sensor 10 and an elastic induction block 11. The dynamic adjustment of the polarization angle is realized through an electric rotation mechanism to eliminate the reflection interference on the strip steel surface. A gear ring 7 is sleeved on the outer surface of the driving motor 6. A sliding long groove 8 is opened on the outer surface of the fixed seat 5. A moving rack 9 is movably connected to the inner wall of the sliding long groove 8. The moving rack 9 meshes with the gear ring 7. The pressure sensor 10 is fixed on the end surface of the moving rack 9. The elastic induction block 11 is fixed on the outer surface of the fixed seat 5. The pressure sensor 10 is in contact with the elastic induction block 11.

[0021] Working principle: The linear array camera 1 is vertically fixed on the frame of the strip steel transportation path through the vertical support shaft 2. The real-time image data of the strip steel edge area is collected in real time by the linear array camera 1. During the detection process, the horizontal state of the linear array camera 1 is maintained through the spherical roller bearing 3. The linear array camera 1 is driven to rotate by the driving motor 6 to adjust the shooting angle. During the rotation, the gear ring 7 drives the moving rack 9 to move along the sliding long groove 8, and then presses the elastic induction block 11. The pressure change is detected by the pressure sensor 10, and then the adjustment angle of the linear array camera 1 is reflected, which is convenient for accurate adjustment.

[0022] Embodiment 2:

[0023] As Figure 3 shown, a detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology includes a device support, a linear array camera 1 and a vision recognition system. The vision recognition system includes an image acquisition unit, an image analysis unit, an environment compensation unit, a deviation prediction unit and a result output unit; The image acquisition unit collects the real-time image data of the strip steel edge area in real time through the linear array camera 1. Its baseline distance is dynamically matched with the strip steel width, and the image spatio-temporal alignment is realized through high-frequency pulse synchronous control; The image acquisition unit further includes a high-frequency pulse LED array and a light source controller; The emission frequency of the high-frequency pulse LED array is synchronized with the line frequency of the linear array camera 1; The light source controller has a built-in closed-loop feedback circuit that adjusts the output power in real time according to the ambient light intensity to ensure that the image signal-to-noise ratio is less than or equal to 40 dB.

[0024] The image analysis unit includes an AI processing module and an image analysis module. The AI processing module is used to integrate an edge computing unit and has a built-in deep learning model. The deep learning model is implemented with a fusion architecture of a residual network and an attention mechanism. The input is the real-time image data output by the image acquisition unit, and it outputs the strip edge pixel-level coordinates and image abnormal feature points and sends them to the image analysis module; The specific process of outputting the strip edge pixel-level coordinates and image abnormal feature points is as follows: S101. Build a deep learning model based on the improved YOLOv5 architecture, and embed a deformable convolutional layer in the backbone network to adapt to the geometric deformation of the strip edge; S102. Introduce a direction classification branch in the detection head part to synchronously output the deviation direction determination result; S103. The loss function is designed as a weighted combination of the deviation amount regression loss and the direction classification loss, and the weight coefficient is automatically determined by Bayesian optimization; S104. When training the deep learning model, use the synthetic data augmentation technology, including simulated scenarios with dynamic lighting, dust occlusion, and strip surface texture changes; S105. Input the real-time image data into the deep learning model, extract the strip edge feature points, construct three-dimensional point cloud data, obtain the strip edge pixel-level coordinates, and obtain the image abnormal feature points based on the feature comparison with the standard strip image.

[0025] The image analysis module is used to obtain the preset threshold of the number of feature points. If the number of image abnormal feature points is greater than or equal to the threshold of the number of feature points, it generates an image correction signal and sends the image correction signal to the environment compensation unit; The environment compensation unit includes a laser ranging unit and an image correction module. The laser ranging unit is used to monitor the distance between the strip and the line array camera and the inclination angle of the strip conveyor roller table in real time, and integrates the obtained distance data and inclination angle data into a data set and sends it to the image correction module; The environment compensation unit also includes a temperature compensation module and a vibration compensation module, where: The temperature compensation module uses a thermistor array to monitor the temperature field distribution in the camera installation area in real time, and corrects the thermal expansion and contraction errors of the optical components by the look-up table method; The vibration compensation module uses an accelerometer to collect the vibration spectrum of the mechanical structure and suppresses the interference signals in the resonance frequency band through a notch filter.

[0026] The image correction module is used to obtain and process the image correction signal, fuse and correct the data set through the Kalman filter and the output of the AI processing module, eliminate the influence of strip vibration and mechanical installation error, and send the corrected image data to the deviation prediction unit; The deviation prediction unit is used to obtain the historical production data of the hot-rolled strip steel and preprocess it, construct an image data set of the hot-rolled strip steel as a training sample, establish a strip steel deviation feature model based on a convolutional neural network, input the corrected image data into the strip steel deviation feature model, extract the strip steel deviation feature, calculate the strip steel deviation probability and send it to the result output unit; The specific process of outputting the strip steel deviation probability is as follows: S201. Obtain the historical production data of the hot-rolled strip steel. The historical production data includes the standard image of the strip steel, the strip steel conveying path and the strip steel edge straight line. Integrate the historical operation data of the roller press as a training sample, and split the training sample into a training set and a test set according to a ratio of 8:2; S202. Construct a strip steel deviation feature model based on a convolutional neural network, download the weight file and load it onto the corresponding network to initialize the transfer network parameters; S203. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the strip steel edge straight line, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use a fixed step size decay to optimize the training parameters, and retrain the entire network to obtain the strip steel deviation feature model; S204. During the training process, randomly and non-repeatedly extract small batches of training samples from the training set for training. After finishing all the training samples, it is a training cycle. Iterate to a certain number of cycles to complete the training and obtain the strip steel deviation feature model; S205. Input the corrected image data into the strip steel deviation feature model to output the strip steel edge straight line. Use the Hough transform line detection algorithm to extract the potential roller paths, and then screen out the position coordinates of the strip steel edge straight line according to the position between the strip steel standard image and the roller paths to perform deviation warning and deviation amount prediction, and obtain the strip steel deviation probability.

[0027] The result output unit is used to obtain the preset strip steel deviation judgment interval, and perform result judgment according to the strip steel deviation probability. If the strip steel deviation probability exceeds the defined range of the strip steel deviation judgment interval, a strip steel deviation signal is generated and sent to the industrial control system for visual display and voice reminder.

[0028] It also includes a control interface unit, which is used to provide a standardized communication protocol interface, output the deviation detection result to the industrial control system at a refresh rate greater than or equal to 1 kHz, support the PID control parameter adaptive adjustment function, and dynamically adjust the response threshold of the deviation correction actuator according to the deviation amount.

[0029] In the present invention, real-time image data of the strip edge region is collected in real time by a line array camera 1, the real-time image data is analyzed by a deep learning model, the strip edge pixel-level coordinates and image abnormal feature points are output, and then distance data and inclination data are obtained. The output of the Kalman filter and the AI processing module are fused and corrected to obtain corrected image data. A strip deviation feature model is established based on a convolutional neural network. The corrected image data is input into the strip deviation feature model to extract strip deviation features, calculate the strip deviation probability and send it to the result output unit. During the strip conveying process, a strip image is obtained by a line array camera, and strip deviation prediction is realized through high-precision image processing and visual recognition, which can intervene in time before the strip deviates and fundamentally solve the production loss caused by strip deviation.

[0030] The setting of the interval and the size of the threshold is for the convenience of comparison. Regarding the size of the threshold, it depends on the amount of sample data and the base quantity set by those skilled in the art for each group of sample data; as long as the proportional relationship between the parameter and the quantized value is not affected.

[0031] In the two embodiments provided in the present application, it should be understood that the disclosed devices and systems can be implemented in other ways; for example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed; another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the devices or modules can be in an electrical, mechanical or other form; The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology, comprising a device bracket, a line array camera (1) and a vision recognition system, characterized in that, The device bracket is fixedly arranged on the top surface of the frame. An angle adjustment mechanism is installed on the outer surface of the device bracket, and two linear array cameras (1) are respectively connected to the outer surface of the angle adjustment mechanism; The visual recognition system includes an image acquisition unit, an image analysis unit, an environmental compensation unit, a deviation prediction unit, and a result output unit; The image acquisition unit uses the linear array camera (1) to collect real-time image data of the strip edge area in real time; The image analysis unit includes an AI processing module and an image analysis module. The AI processing module is used to integrate an edge computing unit and internally build a deep learning model. The deep learning model is implemented based on a fusion architecture of a residual network and an attention mechanism. The input is the real-time image data output by the image acquisition unit, and the strip edge pixel-level coordinates and image abnormal feature points are output and sent to the image analysis module; The image analysis module is used to obtain a preset threshold of the number of feature points. If the number of image abnormal feature points is greater than or equal to the threshold of the number of feature points, an image correction signal is generated and sent to the environmental compensation unit; The environmental compensation unit includes a laser ranging unit and an image correction module. The laser ranging unit is used to monitor the distance between the strip and the linear array camera (1) and the inclination angle of the strip conveying roller table in real time, and the distance data and inclination angle data are integrated into a data set and sent to the image correction module; The image correction module is used to obtain and process the image correction signal, fuse and correct the data set with the output of the AI processing module through a Kalman filter, and send the corrected image data to the deviation prediction unit; The deviation prediction unit is used to obtain the historical production data of the hot-rolled strip and perform preprocessing, construct a hot-rolled strip image data set as a training sample, establish a strip deviation feature model based on a convolutional neural network, input the corrected image data into the strip deviation feature model, extract the strip deviation features, and calculate the strip deviation probability and send it to the result output unit.

2. The detection device for quickly measuring the deviation of the strip steel by using the AI vision recognition technology according to claim 1, wherein, The device bracket includes a spherical roller bearing (3) and a vertical support shaft (2). The two vertical support shafts (2) are symmetrically arranged along the strip width direction on the top surface of the frame adjacent to the conveying path. The two spherical roller bearings (3) are respectively connected to the top surface of the vertical support shaft (2). A horizontal support shaft (4) is commonly installed between the two spherical roller bearings (3), and an angle adjustment mechanism is installed on the outer surface of the horizontal support shaft (4).

3. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, wherein, The angle adjustment mechanism includes a fixed seat (5) and a polarization filter assembly. The fixed seat (5) is fixedly arranged on the end surface of the horizontal support shaft (4). Two driving motors (6) are fixedly arranged inside the fixed seat (5), and the two linear array cameras (1) are respectively fixedly arranged on the outer surface of the output end of the driving motors (6).

4. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 3, characterized in that, The polarization filter assembly includes a pressure sensor (10) and an elastic sensing block (11). A gear ring (7) is sleeved on the outer surface of the drive motor (6). A sliding long groove (8) is formed on the outer surface of the fixed seat (5). A moving rack (9) is movably connected to the inner wall of the sliding long groove (8). The moving rack (9) meshes with the gear ring (7). The pressure sensor (10) is fixed on the end surface of the moving rack (9). The elastic sensing block (11) is fixed on the outer surface of the fixed seat (5). The pressure sensor (10) is in contact with the elastic sensing block (11).

5. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, characterized in that, The image acquisition unit further includes a high-frequency pulse LED array and a light source controller; The emission frequency of the high-frequency pulse LED array is synchronized with the line frequency of the line array camera (1); The light source controller has a built-in closed-loop feedback loop, and adjusts the output power in real time according to the ambient light intensity to ensure that the image signal-to-noise ratio is less than or equal to 40 dB.

6. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, wherein, The specific process of outputting the pixel-level coordinates of the strip edge and the image abnormal feature points is as follows: S101. Construct a deep learning model based on the improved YOLOv5 architecture, and embed a deformable convolutional layer in the backbone network to adapt to the geometric deformation of the strip edge; S102. Introduce a direction classification branch in the detection head part to synchronously output the deviation direction determination result; S103. The loss function is designed as a weighted combination of the deviation amount regression loss and the direction classification loss, and the weight coefficient is automatically determined by Bayesian optimization; S104. When training the deep learning model, a synthetic data augmentation technology is adopted, including simulated scenarios of dynamic illumination, dust occlusion, and strip surface texture changes; S105. Input the real-time image data into the deep learning model, extract the strip edge feature points, construct three-dimensional point cloud data, obtain the pixel-level coordinates of the strip edge, and obtain the image abnormal feature points based on the comparison of features with the standard strip image.

7. The detection device for quickly measuring the deviation of the strip steel by using the AI vision recognition technology according to claim 1, wherein, The specific process of outputting the strip deviation probability is as follows: S201. Obtain the historical production data of the hot-rolled strip. The historical production data includes the standard strip image, the strip conveying path, and the strip edge straight line. Integrate the historical operation data of the roller as the training samples, and split the training samples into a training set and a test set according to the ratio of 8:2; S202. Construct a strip deviation feature model based on a convolutional neural network, download the weight file and load it onto the corresponding network to initialize the transfer network parameters; S203. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the strip edge straight line, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use a fixed step size decay to optimize the training parameters, and retrain the entire network to obtain the strip deviation feature model; S204. During the training process, randomly and non-repeatedly extract small batches of training samples from the training set for training. After extracting all the training samples, it is a training cycle. Iterate to a certain number of cycles to complete the training and obtain the strip deviation feature model; S205. Input the corrected image data into the strip deviation feature model to output the strip edge straight line. Detect the potential roller path therein through the Hough transform straight line detection algorithm, and then perform deviation warning and deviation amount prediction based on the position coordinates of the strip edge straight line selected according to the position between the strip standard image and the roller path, so as to obtain the strip deviation probability.

8. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, characterized in that, The result output unit is used to obtain a preset strip deviation judgment interval and make a result judgment according to the strip deviation probability. If the strip deviation probability exceeds the limit range of the strip deviation judgment interval, a strip deviation signal is generated and sent to the industrial control system for visual display and voice reminder.

9. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, characterized in that, It further includes a control interface unit, which is used to provide a standardized communication protocol interface, output the deviation detection result to the industrial control system at a refresh rate greater than or equal to 1 kHz, support the adaptive adjustment function of PID control parameters, and dynamically adjust the response threshold of the deviation correction actuator according to the deviation amount.

10. The detection device for quickly measuring the deviation of strip steel by using AI vision recognition technology according to claim 1, characterized in that, The environment compensation unit further includes a temperature compensation module and a vibration compensation module, where: The temperature compensation module uses a thermistor array to monitor the temperature field distribution in the camera installation area in real time, and corrects the thermal expansion and contraction error of the optical element through the look-up table method; The vibration compensation module uses an accelerometer to collect the vibration spectrum of the mechanical structure, and suppresses the interference signal in the resonance frequency band through a notch filter.

Images