Strip steel outlet detection method and related equipment

Abstract

The invention discloses a strip steel outlet detection method and related equipment, and relates to the technical field of industrial automatic detection, and the method comprises the steps: obtaining image data of a target detection point in a strip steel outlet area; target detection is conducted on the image data based on a strip steel state detection model, the abnormal state of the strip steel is determined, and the abnormal state comprises the steel stacking state, the steel clamping state or the strip steel arching state; performing motion artifact compensation verification on the abnormal state based on the roller way vibration amplitude data to obtain a verification result; and generating a control instruction based on the verification result and the abnormal state to adjust the operation action of the production line. According to the method, manual monitoring is replaced by automatic detection, the omission ratio and the misjudgment rate are reduced, response and interlocking control of the abnormal state are achieved, non-planned shutdown of a production line is effectively avoided, and continuity and high efficiency of the production process are guaranteed.

Description

Technical Field

[0001] The present application relates to the field of industrial automation detection technology, and in particular to a strip steel outlet detection method and related equipment. Background Art

[0002] During the continuous production of strip steel, abnormalities such as steel jamming, steel piling, and strip arching are prone to occur at the strip exit. Traditional detection methods rely primarily on manual monitoring of industrial television screens or single-sensor technology, which suffer from high missed detection rates and delayed response times. Manual monitoring is limited by operator visual fatigue and subjective judgment errors, making it difficult to maintain accurate and reliable operation around the clock, which can easily lead to involuntary production line slowdowns or shutdowns. Therefore, a strip steel exit detection method is urgently needed to address these technical issues. Summary of the Invention

[0003] The Summary of the Invention introduces a series of simplified concepts that will be further described in the Detailed Description of the Invention. The Summary of the Invention of this application is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0004] In a first aspect, the present application provides a strip steel outlet detection method, comprising:

[0005] Acquire image data of target detection points in the strip exit area;

[0006] Performing target detection on the image data based on the strip state detection model to determine abnormal states of the strip, where abnormal states include steel piling, steel jamming, or strip arching.

[0007] Based on the vibration amplitude data of the roller table, motion artifact compensation verification is performed on the abnormal state to obtain verification results;

[0008] Based on the verification results and abnormal status, control instructions are generated to adjust the production line operation.

[0009] In some embodiments, acquiring image data of a target detection point in a strip exit area includes:

[0010] Based on the preset acquisition frequency of the target industrial camera, the target industrial camera is controlled to perform image acquisition operations to generate raw image data of the target detection point;

[0011] Based on the preset image enhancement algorithm, the original image data of the target detection point is denoised and the contrast is adjusted to generate the image data of the target detection point in the strip exit area.

[0012] In some embodiments, the strip state detection model includes a feature pyramid network module and an attention mechanism module. Target detection is performed on image data based on the strip state detection model to determine abnormal states of the strip, including:

[0013] Based on image data, the global and local features of the strip are extracted through the feature pyramid network module;

[0014] Based on the attention mechanism module, the weights of global features and local features are assigned to determine the attention-enhanced features of the strip;

[0015] Perform bounding box regression on the attention-enhanced features based on dynamic anchor box parameters to generate potential abnormal regions of the strip and the confidence of the potential abnormal regions;

[0016] Based on the comparison result of the confidence level of the potential abnormal area and the preset confidence threshold, the abnormal state of the strip is determined.

[0017] In some embodiments, extracting global and local features of the steel strip based on image data using a feature pyramid network module includes:

[0018] Based on image data, the global semantic features of the strip are extracted through the top-down information flow path of the feature pyramid network module;

[0019] Based on image data, local detail features of the strip are extracted through the bottom-up information flow path of the feature pyramid network module;

[0020] Based on the horizontal convolution connection structure of the feature pyramid network module, multi-level feature fusion is performed on global semantic features and local detail features to generate a multi-scale fusion feature map;

[0021] Based on the multi-scale fusion feature map, the global and local features of the strip are determined.

[0022] In some embodiments, the roller vibration amplitude data is obtained based on a roller vibration sensor in a strip exit area. Based on the roller vibration amplitude data, motion artifact compensation verification is performed on the abnormal state to obtain a verification result, including:

[0023] Determine the vibration frequency and amplitude parameters based on the roller vibration amplitude data;

[0024] Calculate the motion artifact offset of the strip based on the vibration frequency parameter and the amplitude parameter;

[0025] Based on the motion artifact offset, the coordinates of the bounding box of the potential abnormal area corresponding to the abnormal state are compensated for the spatial position to generate the corrected coordinates of the abnormal area;

[0026] Calculating the confidence fluctuation value of the abnormal state based on the difference between the abnormal area coordinates before correction and the abnormal area coordinates after correction;

[0027] Based on a comparison result between the confidence fluctuation value and a preset confidence fluctuation threshold, a verification result of the abnormal state is determined, wherein the verification result is a valid determination or an invalid determination of the abnormal state.

[0028] In some embodiments, based on the verification results and the abnormal status, generating control instructions to adjust the production line operation includes:

[0029] When the verification result is a valid determination, based on the target type of the abnormal state, a target instruction type corresponding to the target type is determined;

[0030] Based on the target instruction type, control instructions are generated and sent to the production line control system through the industrial communication protocol to adjust the production line operation.

[0031] In some implementations, determining a target instruction type corresponding to the target type based on the target type of the abnormal state includes:

[0032] When the abnormal state is the strip arching state, the target instruction type is determined to be a production line speed reduction instruction or a production line stop instruction;

[0033] When the abnormal state is a steel pile state or a steel stuck state, the target instruction type is determined to be a production line stop instruction.

[0034] In a second aspect, the present application proposes a strip steel outlet detection device, comprising:

[0035] A strip steel exit image acquisition unit, used to acquire image data of target detection points in the strip steel exit area;

[0036] The strip abnormal state determination unit performs target detection on the image data based on the strip state detection model to determine the abnormal state of the strip, which includes a steel pile state, a steel stuck state, or a steel arch state.

[0037] The strip abnormal state correction unit verifies the abnormal state through motion artifact compensation based on the roller vibration amplitude data and obtains the verification results;

[0038] The strip production line adjustment unit generates control instructions to adjust the production line operation based on the verification results and abnormal status.

[0039] In a third aspect, an electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the strip exit detection method of any one of the first aspects when executing the computer program stored in the memory.

[0040] In a fourth aspect, the present application proposes a computer-readable storage medium having a computer program stored thereon, which implements the strip exit detection method of any one of the first aspects when the computer program is executed by a processor.

[0041] In summary, this application uses multiple cameras to collaboratively collect dynamic image data, combines a deep learning model to detect the state of the strip, and verifies the motion artifact compensation of the detection results based on the roller vibration amplitude data, thereby improving the accuracy and reliability of abnormality recognition. Through dynamic analysis of vibration frequency and amplitude parameters, spatial position compensation correction is performed on the artifact interference caused by equipment vibration in image detection, effectively distinguishing real abnormalities from vibration noise, and reducing the misjudgment rate. The verified abnormal state is linked to the production line control system through an industrial communication protocol, and control instructions are generated according to the abnormality type to achieve the positioning of the abnormal state and automatic interlocking control. This application reduces manual dependence, avoids the expansion of production line failures due to response delays, ensures the continuous and efficient operation of the production line, and improves the overall efficiency of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present description. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0043] Figure 1 A schematic flow chart of a strip steel outlet detection method provided in an embodiment of the present application;

[0044] Figure 2 A schematic diagram of a strip steel outlet detection interface provided in an embodiment of the present application;

[0045] Figure 3 A schematic structural diagram of a strip steel outlet detection device provided in an embodiment of the present application;

[0046] Figure 4 A schematic structural diagram of an electronic device for strip steel exit detection provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the embodiments described are only part of the embodiments of the present application, not all of the embodiments.

[0048] See also Figure 1 , which is a flow chart of a strip steel outlet detection method provided in an embodiment of the present application, which may specifically include:

[0049] S110, acquiring image data of target detection points in the strip exit area;

[0050] For example, during the strip exit inspection process, multiple high-resolution industrial cameras are first required to capture dynamic images of key inspection points in the exit area. The target inspection points are pre-set locations along the strip's path where steel jams, piles, and arching are likely to occur. The industrial cameras, triggered by PLC or sensor signals, capture high-definition images of the strip's surface and edges in real time. A multi-camera collaborative triggering mechanism switches the working cameras in sections based on the strip's running speed, ensuring dynamic coverage of the inspection area and avoiding missed inspections due to blind spots. The collected image data is transmitted to the server via Gigabit Industrial Ethernet or optical fiber, providing high-quality output for subsequent algorithm processing.

[0051] The image acquisition process is dynamically optimized based on the actual operating conditions of the strip. Based on a preset acquisition frequency, the industrial camera periodically captures images and applies pre-processing, such as denoising and contrast adjustment, to the raw data using a preset image enhancement algorithm to improve image clarity and feature recognizability. The system also adjusts the camera trigger interval in real time based on the strip's operating speed, dynamically balancing acquisition frequency with data processing load to ensure timely image acquisition and resource utilization, providing a stable and reliable data foundation for anomaly detection.

[0052] S120, performing target detection on the image data based on the strip steel state detection model to determine an abnormal state of the strip steel, wherein the abnormal state includes a steel pile state, a steel stuck state, or a steel arched state;

[0053] For example, the strip state detection model uses a deep learning algorithm to perform multi-level feature analysis and fusion on the output image data to achieve efficient recognition of abnormal states. The model is based on the feature pyramid network (FPN) and the path aggregation network (PAN), and extracts the global semantic features and local detail features of the strip through top-down and bottom-up information flow paths, respectively. The global semantic features are used to capture the overall contour information of the strip (such as the stacking shape of the steel pile and the bending trend of the arch), while the local detail features focus on the subtle changes in the surface texture and edge geometry of the strip (such as local deformation of the stuck steel and surface scratches). By fusing multi-scale features through a lateral connection structure, the model can simultaneously take into account the overall characterization of large-scale anomalies and the fine positioning of small-scale defects, thereby improving the comprehensiveness and robustness of detection.

[0054] The model introduces an attention mechanism module, dynamically assigning weights to the fused feature maps, enhancing the response strength of key abnormal areas and suppressing background noise interference. Combining dynamic anchor box clustering with loss function optimization techniques, the model further optimizes the positioning accuracy and confidence calculation of the target detection box. By analyzing the matching relationship between the confidence level of the detection results and the preset threshold, the system quickly determines whether abnormal conditions such as steel piles, steel jams, or arches exist, and outputs the judgment results to the control module. Through algorithm pruning and hardware acceleration technologies, detection latency is ensured to be less than 50ms, meeting the real-time requirements of high-speed production scenarios.

[0055] S130, based on the roller vibration amplitude data, performing motion artifact compensation verification on the abnormal state to obtain a verification result;

[0056] For example, during strip exit inspection, roller vibration can cause motion artifacts during image acquisition, interfering with accurate determination of abnormal conditions. By capturing vibration amplitude data using vibration sensors deployed on the rollers, the system can quantify vibration frequency and amplitude parameters. This, combined with the strip's running speed, dynamically analyzes the impact of vibration on the location of abnormal areas in the image, providing data support for subsequent artifact compensation.

[0057] After calculating the motion artifact offset based on vibration parameters, the coordinates of the bounding box of the detected potential abnormal area are spatially compensated to eliminate coordinate deviations caused by vibration. Furthermore, by comparing the confidence fluctuations of the abnormal area before and after correction, the reliability of the detection results is verified, distinguishing true abnormalities from artifacts caused by vibration interference. This verification mechanism effectively reduces the false positive rate, ensures more accurate determination of abnormal conditions, and provides a reliable basis for the generation of subsequent control instructions.

[0058] S140. Based on the verification result and the abnormal state, generate a control instruction to adjust the operation of the production line.

[0059] For example, when the verification result indicates a valid anomaly, a pre-set control strategy is dynamically applied based on the type of anomaly. For example, if the anomaly is strip arching with a vertical height exceeding a threshold, a line speed reduction command is generated to alleviate stress concentration. If the anomaly is a serious one, such as steel piling or jamming, an emergency shutdown command is directly triggered to prevent equipment damage. Control commands are transmitted in real time to the production line control system via an industrial communication protocol, driving the actuators to adjust operating parameters or stop equipment, ensuring rapid response and precise intervention to anomalies.

[0060] If the verification result is an invalid exception (i.e., the detection is distorted due to vibration interference), the current exception judgment will be ignored, and the image data re-acquisition process will be triggered for secondary verification to avoid the impact of incorrect operation on the stability of the production line. At the same time, the verification results and processing suggestions are fed back in real time through the HMI interface to assist operators in making quick decisions. This application improves the accuracy and reliability of control command generation by integrating abnormality detection and vibration compensation verification mechanisms, thereby ensuring the continuity and safety of the production process.

[0061] The generated control instructions are transmitted in real time to the production line control system via industrial communication protocols, driving the actuators to adjust production line operating parameters or stop equipment. This process achieves automated interlocking control between abnormal conditions and production line operations, eliminating the need for manual decision-making, shortening response time, effectively preventing the escalation of faults caused by delayed processing, and ensuring the safety and continuity of the production line.

[0062] In summary, the embodiment of the present application uses multiple cameras to dynamically collect image data from the strip exit area, and combines an improved deep learning model to perform real-time analysis of the strip state. It can accurately identify abnormal conditions such as steel piling, steel jamming, and strip arching, and verify the motion artifact compensation of the detection results based on the roller vibration amplitude data. According to the verified results, control instructions are generated to adjust the production line operation. The multi-level feature fusion technology based on the feature pyramid network and the attention mechanism module effectively extracts the global contour information of the strip and the detailed features of the local surface texture and edge geometry, thereby improving the comprehensiveness and accuracy of anomaly detection. The dynamic anchor frame clustering and confidence threshold matching mechanism further optimizes the target positioning accuracy. Combined with model pruning and hardware acceleration technology, the detection delay is reduced to less than 50 milliseconds, meeting the real-time requirements of high-speed production scenarios. Further eliminate the detection deviation caused by equipment vibration, verify the validity of the abnormal state by matching the confidence fluctuation value with the preset threshold, effectively distinguish between real anomalies and vibration artifact interference, and reduce the misjudgment rate. Verified abnormal results are linked to the production line control system via industrial communication protocols, triggering differentiated control strategies based on the type of abnormality, achieving automated interlocking control and significantly reducing reliance on manual intervention and response delays. This application not only reduces the risk of misjudgment and missed detection, but also effectively avoids equipment damage and production interruptions by quickly locating abnormalities and proactively adjusting the production process, ensuring continuous and efficient operation of the production line. It also provides technical support for increasing production line speed and improving overall equipment efficiency (OEE).

[0063] In some examples, acquiring image data of a target detection point in a strip exit area includes:

[0064] Based on the preset acquisition frequency of the target industrial camera, the target industrial camera is controlled to perform image acquisition operations to generate raw image data of the target detection point;

[0065] Based on the preset image enhancement algorithm, the original image data of the target detection point is denoised and the contrast is adjusted to generate the image data of the target detection point in the strip exit area.

[0066] Exemplarily, image data acquisition of the strip exit area is achieved through multiple target industrial cameras deployed at the exit section of the production line. The target industrial camera performs image acquisition operations according to a preset acquisition frequency. The preset acquisition frequency is determined based on the adjusted image acquisition frequency, and the image acquisition trigger interval is calculated based on the running speed of the strip. The image acquisition frequency of the target industrial camera is dynamically adjusted in combination with the preset response time parameters of the target industrial camera (for example, the time it takes for the camera to receive a trigger signal and complete a single image acquisition) to ensure that the acquisition interval matches the moving speed of the strip. For example, when the strip runs at high speed, the trigger interval is shortened proportionally to avoid incomplete coverage of the detection area due to strip displacement; when the strip runs at low speed, the trigger interval is extended accordingly to avoid overloading of camera resources. Please refer to Figure 2The figure below is a schematic diagram of a strip exit inspection interface provided by an embodiment of the present application. Six target industrial cameras can be set; these cameras are deployed at preset inspection points in the strip exit area. Each inspection point corresponds to at least one high-resolution industrial camera and is equipped with an auxiliary light source (such as an LED linear light source) to eliminate the effects of uneven ambient lighting. A periodic trigger signal is generated based on the adjusted image acquisition frequency, driving the image sensor to capture raw image data of the strip exit area and transmit it to the server in real time via a high-speed transmission interface (such as optical fiber or Gigabit Ethernet).

[0067] The original image data needs to be preprocessed to improve the accuracy of subsequent target detection. First, a denoising algorithm (such as non-local mean denoising or Gaussian filtering) is used to eliminate the interference of dust, reflection or sensor noise in the industrial environment on the image quality. Secondly, contrast enhancement technology (such as histogram equalization or adaptive gamma correction) is used to optimize the detailed features of the strip edges and abnormal areas in the image to ensure the target recognizability in low-contrast scenes. For example, for the arched area of ​​the strip, the difference between the arch height and the background is highlighted by enhancing the local contrast; for the piled or stuck steel area, the texture features are enhanced to distinguish between abnormal accumulation and normal strip surface. The preprocessed image data retains the geometric shape and abnormal features of the strip, providing a high signal-to-noise ratio output for the subsequent deep learning model. The parameters of the image enhancement algorithm (such as the filter kernel size and the contrast gain coefficient) are set according to the actual situation to ensure that it adapts to the detection requirements of different lighting conditions and strip materials.

[0068] In some examples, the method for determining the preset acquisition frequency includes:

[0069] Based on the running speed of the strip steel, the image acquisition trigger interval of the target industrial camera is determined;

[0070] Dynamically adjust the image acquisition frequency of the target industrial camera based on the image acquisition trigger interval and the preset response time of the target industrial camera;

[0071] The adjusted image acquisition frequency is used as the preset acquisition frequency.

[0072] Exemplarily, the image acquisition trigger interval = preset reference distance / strip running speed, where the preset reference distance is determined by the field of view coverage and detection accuracy requirements of the target industrial camera. For example, if the camera's horizontal field of view coverage is 500mm, and it is necessary to ensure that the strip displacement in two adjacent frames does not exceed 50mm to avoid missed detection, the preset reference distance is set to 50mm. When the strip runs at a speed of 2m / s, the trigger interval is calculated to be 25ms. By real-time monitoring of the strip running speed (for example, obtained through an encoder or speed sensor), the trigger interval parameters are dynamically updated to ensure seamless coverage of the detection area at different speeds.

[0073] The preset response time of the target industrial camera refers to the total time it takes from receiving the trigger signal to completing a single image acquisition and transmission (for example, 10ms). The dynamic adjustment logic process is as follows: if the calculated trigger interval is greater than the preset response time, the image acquisition trigger interval is directly used as the trigger interval. At this time, the camera has sufficient time to complete a single acquisition. To ensure that the camera has sufficient time to complete a single acquisition, the adjusted image acquisition frequency is 1 / image acquisition trigger interval; if the trigger interval is less than or equal to the preset response time, the trigger interval is adjusted to an integer multiple k of the preset response time (take 1.5 times, that is, 15ms) to avoid image loss or blurring due to the trigger frequency exceeding the camera processing capability, so as to ensure that the camera completes the current acquisition before triggering the next operation. The adjusted image acquisition frequency is 1 / (k×preset response time).

[0074] In some examples, the strip state detection model includes a feature pyramid network module and an attention mechanism module. The strip state detection model performs target detection on image data and determines abnormal states of the strip, including:

[0075] Based on image data, the global and local features of the strip are extracted through the feature pyramid network module;

[0076] Based on the attention mechanism module, the weights of global features and local features are assigned to determine the attention-enhanced features of the strip;

[0077] Perform bounding box regression on the attention-enhanced features based on dynamic anchor box parameters to generate potential abnormal regions of the strip and the confidence of the potential abnormal regions;

[0078] Based on the comparison result of the confidence level of the potential abnormal area and the preset confidence threshold, the abnormal state of the strip is determined.

[0079] For example, based on the output image data of the strip exit area, the Feature Pyramid Network (FPN) module extracts the global semantic features of the strip through a top-down information flow path. The top-down path outputs high-dimensional semantic information from the deep convolutional layer, which is gradually upsampled and fused with the shallow feature map to generate a multi-scale feature map. The global semantic features represent the overall contour information of the strip, such as the stacking shape of the pile or the bending trend of the arch. At the same time, the Path Aggregation Network (PAN) module extracts the local detail features of the strip through a bottom-up information flow path. The bottom-up path transfers the detail information of the underlying high-resolution feature map (such as the surface texture and edge geometry of the strip) to the higher layers layer by layer, enhancing the model's ability to capture minor defects. Through the synergistic effect of the bidirectional information flow of FPN and PAN, and using the horizontal convolution connection structure to perform multi-level feature fusion of global semantic features and local detail features, a multi-scale fused feature map is generated, thereby determining the global and local features of the strip, taking into account the expression of both the overall morphology of the strip and the local abnormal features.

[0080] The attention mechanism module includes a channel attention submodule and a spatial attention submodule. First, the channel attention submodule calculates the weight coefficient of each feature channel in the global feature, specifically including: performing global average pooling and maximum pooling operations on the global feature map to generate a channel description vector; performing nonlinear transformation on the channel description vector through the fully connected layer and activation function, and outputting the weight coefficient of each channel. Secondly, the spatial attention submodule calculates the weight coefficient of each spatial position in the local feature, specifically including: performing mean pooling and maximum pooling on the local feature map along the channel dimension to generate a spatial attention map; performing feature mapping on the spatial attention map through the convolution layer, and outputting the weight coefficient of each spatial position. The channel weight coefficient and the spatial weight coefficient are multiplied element by element to generate an attention-enhanced feature. The attention-enhanced feature strengthens the response strength of the abnormal area by suppressing background noise and redundant information.

[0081] The dynamic anchor box parameters are generated by statistically analyzing the shape and size of the abnormal regions in the training dataset using the K-means clustering algorithm, ensuring that the anchor box size closely matches the geometric shape of the stuck steel, piled steel, and arched areas. Based on the attention-enhanced features, the dynamic anchor box parameters are adjusted through a bounding box regression network, and the complete intersection-over-union loss function (CIoULoss) is combined to optimize the position and size prediction accuracy of the bounding box. The CIoU loss function comprehensively considers the overlapping area, center point distance, and aspect ratio differences of the bounding boxes to ensure the spatial match between the candidate abnormal regions after regression and the actual abnormal regions. After regression, the potential abnormal regions of the strip steel and their confidence levels are generated. The confidence level represents the probability that the region is in an abnormal state and is calculated by normalizing the output of the classification branch using the Softmax function.

[0082] The confidence level of the potential abnormal area is compared with the preset confidence threshold to determine the abnormal state of the strip. The preset confidence threshold is determined based on the performance optimization of the model on the validation set. For example, the threshold for the steel pile state is 0.92, the steel jam state is 0.90, and the strip arch state is 0.88. If the confidence level of a potential abnormal area exceeds the corresponding threshold, it is determined to be the corresponding abnormal type; otherwise, it is considered to be in a normal state. For multiple candidate abnormal areas in the same image, the Soft-NMS algorithm is used to suppress redundant detection frames, and the detection result with the highest confidence is retained as the final judgment basis. The judgment result is output to the control module, triggering the production line adjustment instruction corresponding to the abnormal type, such as generating an emergency stop instruction when the confidence level of steel pile or steel jam exceeds the threshold.

[0083] For example, in this example, based on the requirements of strip steel outlet inspection, a dataset of 100,000 strip steel images was constructed, 5% of which were manually annotated negative samples. These negative samples were generated by simulating abnormal scenarios such as steel jams and steel piles to enhance the model's generalization to rare abnormal conditions. The model training used a pre-trained YOLOv5 architecture as the base network. Model parameters were initialized using transfer learning techniques, and FPN and PAN were introduced to achieve multi-scale feature fusion. FPN extracts global semantic features of the strip steel through top-down information flow, while PAN enhances the transfer of local detail features through bottom-up information flow. Combined with a lateral convolutional connection structure, global and local features are fused layer by layer to generate a multi-scale fused feature map. Furthermore, the model integrates channel and spatial attention mechanisms to dynamically assign feature weights to enhance the response strength of abnormal areas and suppress background noise interference. The Adam optimizer was used during training, with an initial learning rate of 0.001, which was decayed by 0.1 every 50 training cycles to balance convergence speed and training stability. Model performance evaluation uses mean average precision (mAP) greater than or equal to 95% and an F1 score greater than or equal to 92% as acceptance criteria. mAP comprehensively considers detection accuracy at different confidence thresholds, while the F1 score reflects the model's balance between recall and precision for anomalies. Through these optimization strategies, the model retains the efficient detection capabilities of YOLOv5 while improving the accuracy of localizing and classifying anomalies such as stuck steel, piled steel, and strip arching, meeting the high robustness and real-time requirements of industrial scenarios.

[0084] In some instances, roller vibration amplitude data is acquired based on roller vibration sensors in the strip exit area. Based on the roller vibration amplitude data, motion artifact compensation verification is performed for abnormal conditions, and verification results include:

[0085] Determine the vibration frequency and amplitude parameters based on the roller vibration amplitude data;

[0086] Calculate the motion artifact offset of the strip based on the vibration frequency parameter and the amplitude parameter;

[0087] Based on the motion artifact offset, the coordinates of the bounding box of the potential abnormal area corresponding to the abnormal state are compensated for the spatial position to generate the corrected coordinates of the abnormal area;

[0088] Calculating the confidence fluctuation value of the abnormal state based on the difference between the abnormal area coordinates before correction and the abnormal area coordinates after correction;

[0089] Based on a comparison result between the confidence fluctuation value and a preset confidence fluctuation threshold, a verification result of the abnormal state is determined, wherein the verification result is a valid determination or an invalid determination of the abnormal state.

[0090] For example, roller vibration amplitude data is collected in real time by vibration sensors deployed on the rollers in the strip exit area. The sensors record the roller vibration waveform data at a preset sampling frequency. Based on this vibration waveform data, frequency domain analysis is used to extract vibration frequency parameters. These parameters represent the frequency range of the roller's periodic vibration, for example, 5 Hz to 20 Hz. Simultaneously, the amplitude parameters are determined by calculating the maximum peak and root mean square values ​​of the vibration signal. These amplitude parameters represent the amplitude range of the vibration displacement, for example, 0.1 mm to 2 mm. These parameters provide a quantitative basis for the subsequent calculation of motion artifact offsets.

[0091] The calculation of motion artifact offset requires considering the correlation between the strip's operating speed and vibration parameters. First, based on the acquired strip's operating speed and vibration frequency parameters, the displacement change of the strip within a single vibration cycle is calculated. The displacement change is the ratio of the strip's operating speed to the vibration frequency. Second, based on the combined effect of the amplitude parameters and the displacement change, the integrated motion artifact offset of the strip in the image is determined. The offset is the square root of the sum of the squares of the horizontal displacement change and the amplitude, and the square root of the sum of the squares of the longitudinal displacement change and the amplitude. These calculations quantify the offset effect of vibration on the positioning of abnormal areas in the strip image.

[0092] Based on the calculated motion artifact offset, the bounding box coordinates of the potential anomaly area output by the strip condition detection model are corrected. Specifically, the horizontal offset is subtracted from the horizontal component of the original bounding box center coordinates, and the longitudinal offset is subtracted from the longitudinal component to generate the corrected coordinates of the anomaly area center. This coordinate correction eliminates interference with the positioning of the anomaly area caused by image artifacts caused by roller vibration, thereby improving the geometric accuracy of the detection results.

[0093] The confidence fluctuation value is used to assess the consistency of detection results before and after vibration compensation. The confidence fluctuation value is calculated by taking the absolute difference between the confidence of the abnormal region before and after correction. If the correction results in a significant decrease in confidence, it indicates that the original detection result may be contaminated by vibration artifacts. If the confidence fluctuation is small, the abnormal state is considered reliable. Furthermore, the change in the overlapping area of ​​the bounding boxes before and after correction is used for auxiliary verification to further quantify the stability of the detection results.

[0094] The calculated confidence fluctuation value is compared with a preset confidence fluctuation threshold. If the confidence fluctuation value is less than or equal to the threshold, the verification result is considered valid, indicating that the confidence level of the abnormal state is stable after vibration compensation and the detection result is reliable. If the confidence fluctuation value is greater than the threshold, the verification result is considered invalid, indicating that the original detection result is severely affected by vibration artifacts, triggering image recapture or manual review. The threshold is determined based on the statistical distribution of true anomalies and artifact interference in historical data, ensuring that detection sensitivity is maintained while reducing the false positive rate. The verification results are transmitted to the control module via an industrial communication protocol, providing a reliable basis for subsequent production line control.

[0095] In some instances, based on the verification results and abnormal conditions, control instructions are generated to adjust production line operations, including:

[0096] When the verification result is a valid determination, based on the target type of the abnormal state, a target instruction type corresponding to the target type is determined;

[0097] Based on the target instruction type, control instructions are generated and sent to the production line control system through the industrial communication protocol to adjust the production line operation.

[0098] Exemplarily, when the verification result is a valid judgment, the corresponding target instruction type is determined based on the mapping relationship between the target type of the abnormal state (including steel piling state, steel jamming state or strip arching state) and the preset control strategy. Specifically, if the abnormal state is a strip arching state, the vertical height parameter of the arching area is calculated based on the bounding box coordinates of the potential abnormal area, and compared with the first preset height threshold (for example, 100mm) and the second preset height threshold (for example, 50mm). When the vertical height parameter is greater than or equal to the first preset height threshold, it is determined to be a high-risk arching and an emergency shutdown instruction for the production line is generated; when the vertical height parameter is less than the first preset height threshold but greater than or equal to the second preset height threshold, it is determined to be a potential risk and a production line speed reduction instruction is generated; if the abnormal state is a steel piling state or a steel jamming state, it is determined to be an emergency abnormality and a production line shutdown instruction is generated; the above judgment logic is implemented through a preset rule base to ensure strict correspondence between the abnormal type and the instruction type, so that manual intervention can be performed while maintaining production. The first preset height threshold and the second preset height threshold are set according to historical data and process safety specifications to ensure the adaptability and safety of the control strategy.

[0099] When the verification result is invalid, an abnormal review instruction is generated, triggering the image data re-collection and manual review process. At the same time, the review prompt information and abnormal area coordinates are output on the human-machine interface (HMI) to assist operators in quickly locating the problem.

[0100] Based on the target command type, standardized control commands are generated via industrial communication protocols (such as OPC). These control commands contain a command code, target device address, and execution parameters (such as the stop code "EMG_STOP" or the speed reduction parameter "SPEED_30%"). Once generated, these commands are sent to the production line control system (L1 system) via a real-time communication interface (such as Industrial Ethernet) to adjust production line operations. When a line stop command is generated, the control command includes a stop signal and the location of the abnormality. This triggers the L1 system to immediately shut down the roller table and activate the mechanical brakes, ensuring that the strip stops within the set time and simultaneously triggering an audible and visual alarm. When a line speed reduction command is generated, the control command includes a target speed reduction ratio (e.g., to 40% of the current speed) and a duration parameter. This triggers the L1 system to adjust the roller table speed according to a preset gradient and activate an early warning prompt. The generation and transmission of control commands follows a pre-set communication cycle, ensuring that response latency remains below the production line's safety threshold. Simultaneously, the execution status and abnormality details of the control command are updated in real time to the HMI, including the abnormality type, command type, execution timestamp, and recommended actions, providing decision support for operators. Through the above steps, real-time closed-loop control of abnormal conditions and production line operations can be achieved, improving response efficiency and operational safety.

[0101] See also Figure 3 , is a schematic structural diagram of a strip steel outlet detection device provided in an embodiment of the present application, comprising:

[0102] The strip steel outlet image acquisition unit 21 acquires image data of target detection points in the strip steel outlet area;

[0103] The strip abnormal state determination unit 22 performs target detection on the image data based on the strip state detection model to determine the abnormal state of the strip;

[0104] The strip abnormal state correction unit 23 performs motion artifact compensation verification on the abnormal state based on the roller vibration amplitude data to obtain a verification result;

[0105] The strip steel production line adjustment unit 24 generates control instructions to adjust the production line operation based on the verification results and abnormal status.

[0106] See also Figure 4 An embodiment of the present application also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any method for strip exit detection are implemented.

[0107] Since the electronic device introduced in this embodiment is the equipment used to implement a strip steel exit detection device in the embodiment of this application, based on the method introduced in the embodiment of this application, technical personnel in this field can understand the specific implementation of the electronic device of this embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiment of this application will not be introduced in detail here. As long as the equipment used by technical personnel in this field to implement the method in the embodiment of this application falls within the scope of protection of this application.

[0108] During the specific implementation process, when the computer program 311 is executed by the processor, any implementation method of the embodiments corresponding to the first aspect can be implemented.

[0109] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0110] Those skilled in the art will appreciate that the embodiments of the present application may provide methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0111] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0112] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0113] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0114] The present application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device executes Figure 1 The process of a strip steel exit detection method in the corresponding embodiment.

[0115] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium can be a magnetic medium, an optical medium or a semiconductor medium, etc.

[0116] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0118] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0119] In addition, the functional units in the various embodiments of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The above-mentioned integrated units may be implemented in the form of hardware and / or software functional units.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disk.

[0121] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

[0122] Although the preferred embodiments of this specification have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of this specification.

[0123] Obviously, those skilled in the art may make various changes and modifications to this specification without departing from the spirit and scope of this specification. Thus, if such changes and modifications fall within the scope of the claims of this specification and their equivalents, this specification is intended to include such changes and modifications.

Claims

1. A strip steel outlet detection method, characterized in that: include: Acquire image data of target detection points in the strip exit area; performing target detection on the image data based on a strip steel state detection model to determine an abnormal state of the strip steel, wherein the abnormal state includes a steel pile state, a steel stuck state, or a steel arched state; Based on the roller vibration amplitude data, motion artifact compensation verification is performed on the abnormal state to obtain a verification result; Based on the verification result and the abnormal state, a control instruction is generated to adjust the operation of the production line.

2. The method according to claim 1, characterized in that The acquiring of image data of target detection points in the strip steel exit area includes: Based on the preset acquisition frequency of the target industrial camera, the target industrial camera is controlled to perform image acquisition operations to generate raw image data of the target detection point; Based on a preset image enhancement algorithm, the original image data of the target detection point is denoised and the contrast is adjusted to generate image data of the target detection point in the strip exit area.

3. The method according to claim 1, characterized in that The strip state detection model includes a feature pyramid network module and an attention mechanism module. The strip state detection model is used to perform target detection on the image data to determine the abnormal state of the strip, including: Based on the image data, extracting global features and local features of the steel strip through the feature pyramid network module; weighting the global features and the local features based on the attention mechanism module to determine an attention-enhanced feature of the steel strip; Performing bounding box regression on the attention enhancement feature based on dynamic anchor box parameters to generate a potential abnormal area of ​​the steel strip and a confidence score of the potential abnormal area; Based on the comparison result of the confidence of the potential abnormal area and the preset confidence threshold, the abnormal state of the steel strip is determined.

4. The method according to claim 3, characterized in that The extracting global features and local features of the steel strip based on the image data by the feature pyramid network module includes: Based on the image data, extracting global semantic features of the steel strip through a top-down information flow path of the feature pyramid network module; Based on the image data, extracting local detail features of the steel strip through a bottom-up information flow path of the feature pyramid network module; Based on the horizontal convolution connection structure of the feature pyramid network module, multi-level feature fusion is performed on the global semantic features and the local detail features to generate a multi-scale fusion feature map; Based on the multi-scale fusion feature map, the global features and local features of the steel strip are determined.

5. The method according to claim 3, characterized in that The roller vibration amplitude data is obtained based on the roller vibration sensor in the strip exit area. The abnormal state is verified by motion artifact compensation based on the roller vibration amplitude data to obtain a verification result, including: Determining vibration frequency parameters and amplitude parameters based on the roller vibration amplitude data; Calculating a motion artifact offset of the steel strip based on the vibration frequency parameter and the amplitude parameter; Based on the motion artifact offset, performing spatial position compensation correction on the bounding box coordinates of the potential abnormal area corresponding to the abnormal state to generate corrected abnormal area coordinates; Calculating a confidence fluctuation value of the abnormal state based on a difference between the abnormal region coordinates before correction and the abnormal region coordinates after correction; Based on a comparison result between the confidence fluctuation value and a preset confidence fluctuation threshold, a verification result of the abnormal state is determined, wherein the verification result is a valid determination or an invalid determination of the abnormal state.

6. The method according to claim 5, characterized in that The generating of a control instruction based on the verification result and the abnormal state to adjust the operation of the production line includes: When the verification result is a valid determination, determining a target instruction type corresponding to the target type based on the target type of the abnormal state; Based on the target instruction type, the control instruction is generated, and the control instruction is sent to the production line control system through an industrial communication protocol to adjust the production line operation action.

7. The method according to claim 6, characterized in that The determining, based on the target type of the abnormal state, a target instruction type corresponding to the target type, includes: When the abnormal state is the strip arching state, determining the target instruction type to be a production line speed reduction instruction or a production line shutdown instruction; When the abnormal state is the steel piling state or the steel stuck state, the target instruction type is determined to be the production line shutdown instruction.

8. A strip steel outlet detection device, characterized in that: include: A strip steel exit image acquisition unit, used to acquire image data of target detection points in the strip steel exit area; a steel strip abnormal state determining unit, which performs target detection on the image data based on a steel strip state detection model to determine an abnormal state of the steel strip, wherein the abnormal state includes a steel pile state, a steel stuck state, or a steel strip arch state; The strip abnormal state correction unit performs motion artifact compensation verification on the abnormal state based on the roller vibration amplitude data to obtain a verification result; The strip steel production line adjustment unit generates a control instruction to adjust the production line operation based on the verification result and the abnormal state.

9. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the strip exit detection method as claimed in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the strip steel exit detection method according to any one of claims 1 to 7 is implemented.

Images