Copper foil shearing on-line quality detection and feedback control system based on machine vision

Abstract

The invention discloses a copper foil shearing on-line quality detection and feedback control system based on machine vision, and relates to the technical field of copper foil quality detection, and the system comprises a data collection module which is used for arranging a plurality of industrial cameras at key positions of a copper foil shearing machine, collecting shearing edge and surface images in real time, and sending the images to a control module; a tension sensor, a speed sensor and a temperature sensor are synchronously arranged; the compensation correction module is used for organizing the surface images into a time sequence and performing multi-scale wavelet decomposition on each frame of image in the time sequence to obtain detail components; the index design module is used for calculating feature indexes of each time sequence and each detail component; the defect classification module is used for obtaining defect classification fine judgment according to the real-time characteristic indexes; and the feedback control module is used for feeding back the detection result to the shearing machine controller in real time and adjusting the pressure and the feeding speed of the cutter. According to the invention, the problem that a fixed threshold value is easy to generate false alarm and missing alarm under the condition of working condition fluctuation is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of copper foil quality detection, and in particular to a copper foil shearing online quality detection and feedback control system based on machine vision. Background Art

[0002] The quality of copper foil shearing directly impacts the performance and yield of subsequent key components, such as solar cells and circuit boards. Traditional manual visual inspection or simple sensor monitoring struggles to detect minute defects promptly and accurately. Online inspection systems often focus on overall dimensions, overlooking edge burrs, microcracks, and surface texture anomalies. Furthermore, environmental changes and fluctuating operating conditions lead to high false positive and false negative rates, making real-time closed-loop feedback control difficult.

[0003] Traditional inspection methods rely on manual visual inspection, which is highly subjective and inefficient. They can only be sampled and inspected when the machine is stopped, making continuous online monitoring impossible. Using tension or torque sensors alone to monitor shear force changes makes it difficult to distinguish between tiny burrs and normal fluctuations, leading to missed detections. Furthermore, many industrial vision solutions only measure copper foil width and cut straightness, making them insensitive to detailed defects caused by microscopic edge burrs and high-frequency vibrations. Algorithms designed to detect large-area defects (such as scratches and stains) often fail to capture tiny cracks and burrs at the moment of shearing. Most systems use fixed thresholds, and detection performance degrades significantly when operating conditions (such as speed, tension, and temperature) change. Furthermore, interference from ambient lighting fluctuations and camera shake makes it difficult to distinguish between real defects and artifacts, resulting in a high false alarm rate.

[0004] To this end, the present invention provides an online quality detection and feedback control system for copper foil shearing based on machine vision. Summary of the Invention

[0005] The purpose of the present invention is to provide an online quality detection and feedback control system for copper foil shearing based on machine vision to solve the existing problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a machine vision-based copper foil shearing online quality detection and feedback control system, comprising: The data acquisition module is used to deploy multiple industrial cameras at key locations on the copper foil shearing machine to capture images of the shearing edge and surface in real time, and is also equipped with tension sensors, speed sensors, and temperature sensors. The compensation and correction module is used to organize the surface image into a time series and perform multi-scale wavelet decomposition on each frame image in the time series to obtain detail components; Indicator design module, used to calculate the characteristic indicators of each time series and each detail component; Defect classification module, which obtains detailed defect classification judgment based on real-time feature indicators; The feedback control module is used to feed back the detection results to the shearing machine controller in real time to adjust the cutter pressure and feed speed.

[0007] A further improvement of the present invention is that the image acquisition module includes N industrial cameras installed along both sides of the assembly line before and after the shearing machine outlet, and performs illumination balancing, geometric distortion correction and camera calibration compensation calibration on the original image; the camera is equipped with a ring LED light source and a polarizing filter, and the focal length, principal point coordinates, and tangential distortion coefficient are calculated through a checkerboard calibration plate, and the transformation matrix between the camera coordinate system and the shearing machine coordinate system is calculated; the image is converted to Lab color space, contrast-limited adaptive histogram equalization is performed on the brightness channel, the copper foil shearing area is located based on template matching, and an ROI of size H×M pixels is intercepted to obtain the incision edge and surface area.

[0008] A further improvement of the present invention is that the compensation correction module includes a trust calculation unit and a correction execution unit; The trust calculation unit includes the following steps: taking sensor nodes whose distance from a target sensor node in a production line layout is less than a set distance threshold as adjacent sensor nodes; obtaining quality parameters of each adjacent sensor node of the target sensor node; obtaining historical data of the target sensor node and its adjacent sensor nodes; calculating the alarm accuracy of the target sensor node and its adjacent sensor nodes based on the historical data; and calculating the trust of each adjacent sensor node based on the alarm accuracy.

[0009] A further improvement of the present invention is that the trust calculation unit also includes false alarm data that is detected as abnormal in statistical historical data but is normal after manual re-inspection, and the alarm accuracy is obtained according to the proportion of the difference between the total data volume and the false alarm data in the total data volume; the Pearson correlation coefficient of the data of the adjacent nodes and the target node is calculated to obtain the trust correlation between the target sensor and the adjacent sensor.

[0010] A further improvement of the present invention is that the correction execution unit calculates the compensation amount of the target detection sensor based on the trust of each adjacent sensor; and based on the compensation amount, the quality parameter of the target sensor is compensated and corrected through weighted fusion correction of adjacent node data.

[0011] A further improvement of the present invention is that the compensation correction module also includes arranging continuously collected image features, including a copper foil edge coordinate sequence and a defect area sequence, and process parameters, including a shear force time series, according to timestamps to form a sliding window with a length of N points; performing a three-layer Daubechies wavelet decomposition on the copper foil edge coordinate sequence to obtain an approximate component and a detail component; and decomposing the shear force time series to extract high-frequency components to identify impact force anomalies at the moment of shearing.

[0012] A further improvement of the present invention is that the indicator design module includes an edge jitter index for measuring the degree of chaos of the edge coordinate sequence, calculates the defect distribution correlation dimension through the GP algorithm, calculates the texture information entropy through the entropy value of the image grayscale distribution, and obtains the shear force mutation coefficient through the ratio of the standard deviation of the detail component to the mean of the approximate component.

[0013] A further improvement of the present invention is that the defect classification module includes a preliminary defect detection unit and a fine classification unit; the preliminary defect detection unit performs preliminary anomaly detection based on time series, and trains a threshold prediction model based on production conditions and historical image features to dynamically generate discrimination thresholds for each feature for the current working conditions; specifically, by collecting visual features and process parameters of defect-free copper foil, a historical sample database is constructed; an indicator threshold model is established, based on a support vector regression model, with process parameters and environmental parameters as input data, the process parameters include shear speed, tension and blade gap, and the environmental parameters include temperature and humidity, and the maximum threshold of all indicators in the indicator design module is output as the discrimination threshold; if any index is greater than the sum of the current output maximum threshold and 5% redundancy, a preliminary anomaly warning is triggered.

[0014] A further improvement of the present invention is that the fine classification unit performs fine defect judgment based on the component sequence; specifically, the shear speed, tension, and blade gap are encoded into a 10-dimensional process feature vector, the indicators of each wavelet component are combined into a 15-dimensional component feature vector, the copper foil defect dataset is used as the training dataset, and a ResNet-18 convolutional neural network is adopted. The W×H pixel incision image and the 25-dimensional feature vector are input, and the defect category and confidence level are output.

[0015] A further improvement of the present invention is that the feedback control module is used to trigger the shearing machine control strategy when both the preliminary detection and the supplementary detection are determined to be abnormal: the cutter pressure and feed speed are automatically adjusted through the PLC system until any index is less than or equal to the sum of the maximum threshold of the current output and the 5% redundancy, and at the beginning of each shift, the camera field of view and light source brightness are automatically calibrated using standard copper foil.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention firstly overcomes monocular blind spots and single-point failures through multi-camera fusion and trust evaluation, and dynamically weights the correlation and historical accuracy between adjacent cameras to achieve more reliable edge and surface defect capture. The nonlinear texture index is used to quantify the chaos and complex fluctuations in the shearing process, which can better reflect the dynamic instability than simple statistical features, thus capturing shear anomalies in advance. The dynamic threshold prediction model inputs working condition parameters such as shear velocity, tension, temperature and humidity into the regression model to generate the optimal discrimination threshold for each feature in real time, avoiding the performance degradation of fixed thresholds under fluctuating working conditions; A two-stage detection architecture of threshold filtering and fine classification is adopted to quickly filter the vast majority of qualified frames first, and then perform in-depth classification of potentially defective frames, taking into account both real-time performance and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a framework diagram of the machine vision-based copper foil shearing online quality detection and feedback control system of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0019] The term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. Additionally, the character " / " generally indicates an "or" relationship between the related objects.

[0020] Example 1 Figure 1 The framework diagram of the machine vision-based copper foil shearing online quality detection and feedback control system disclosed in this embodiment is shown, including: The data acquisition module is used to deploy multiple industrial cameras at key locations on the copper foil shearing machine to capture images of the shearing edge and surface in real time, and is also equipped with tension sensors, speed sensors, and temperature sensors. The compensation and correction module is used to organize the surface image into a time series and perform multi-scale wavelet decomposition on each frame image in the time series to obtain detail components; Indicator design module, used to calculate the characteristic indicators of each time series and each detail component; Defect classification module, which obtains detailed defect classification judgment based on real-time feature indicators; The feedback control module is used to feed back the detection results to the shearing machine controller in real time to adjust the cutter pressure and feed speed.

[0021] The image acquisition module includes installing N industrial cameras on both sides of the assembly line before and after the shearing machine outlet, and performing illumination balance, geometric distortion correction and camera calibration compensation calibration on the original image; the camera is equipped with a ring LED light source and a polarization filter, and the focal length, principal point coordinates, tangential distortion coefficient, and the conversion matrix between the camera coordinate system and the shearing machine coordinate system are calculated through a checkerboard calibration plate; multiple industrial cameras are installed on both sides of the assembly line before and after the shearing machine outlet to ensure that the full width of the copper foil is covered, and imaging errors are eliminated through calibration; the image will be preprocessed, including: Dark field correction is performed: a completely black image is captured with the light source turned off, and CMOS dark current noise is subtracted. White field correction is performed: an image of a standard white ceramic plate is captured, and the gain coefficient of each pixel is calculated to compensate for non-uniform illumination. The image is converted to Lab color space, and contrast-limited adaptive histogram equalization is performed on the luminance channel (L channel). The copper foil cutout area is located based on template matching, a fixed-size ROI is captured, and the cutout edge and surface area are focused. The copper foil cutout area is located based on template matching, and an ROI of size H×M pixels is captured to obtain the cutout edge and surface area.

[0022] The compensation correction module includes a trust calculation unit and a correction execution unit; The trust calculation unit includes the following steps: taking sensor nodes whose distance from a target sensor node in a production line layout is less than a set distance threshold as adjacent sensor nodes; obtaining quality parameters of each adjacent sensor node of the target sensor node; obtaining historical data of the target sensor node and its adjacent sensor nodes; calculating the alarm accuracy of the target sensor node and its adjacent sensor nodes based on the historical data; and calculating the trust of each adjacent sensor node based on the alarm accuracy.

[0023] The trust calculation unit also includes false alarm data detected as abnormal in statistical historical data but normal after manual re-inspection, and obtains the alarm accuracy according to the proportion of the difference between the total data volume and the false alarm data in the total data volume; calculates the Pearson correlation coefficient of the data of the adjacent nodes and the target node to obtain the trust association between the target sensor and the adjacent sensor.

[0024] The correction execution unit calculates the compensation amount for the target detection sensor based on the confidence level of each adjacent sensor. Based on this compensation amount, the target sensor's quality parameters are compensated and corrected through weighted fusion correction of adjacent node data. Weight = confidence level / total confidence level. For example, if the image from the entry camera has blurred edges due to light source fluctuations, the stable edge data from the exit camera is used for interpolation and repair.

[0025] The compensation and correction module also includes arranging continuously collected image features, including the copper foil edge coordinate sequence and the defect area sequence, and process parameters, including the shear force time series, according to timestamps to form a sliding window of length N points, which is dynamically adjusted according to the shear speed. For example, when the shear speed is 2m / s, the window corresponds to a copper foil length of 2.5-5m; performing a three-layer Daubechies wavelet decomposition on the copper foil edge coordinate sequence to obtain an approximate component that reflects the low-frequency trend of the overall direction of the copper foil (such as the long-term drift of the edge offset) and a detail component that reflects the edge jitter of high-frequency fluctuations (such as edge burrs caused by shear vibration); decomposing the shear force time series, extracting the high-frequency component to identify the impact force anomaly at the moment of shearing.

[0026] The indicator design module includes obtaining an edge jitter index through Lyapunov to measure the degree of chaos of the edge coordinate sequence. An index greater than 0.1 is considered shear instability (calculated based on the Wolf algorithm). The defect distribution correlation dimension is calculated through the GP algorithm to describe the complexity of the spatial distribution of surface defects. Calculated through the GP algorithm, a dimension greater than 2.5 indicates abnormal density of defects. The texture information entropy is calculated through the entropy value of the image grayscale distribution. An entropy value less than 3.0 indicates abnormal surface texture (such as gloss uniformity destruction caused by overpressure). The shear force mutation coefficient is obtained by the ratio of the standard deviation of the detail component to the mean of the approximate component. For example, when the edge jitter Lyapunov is greater than 0.3, it indicates that the shear force fluctuation exceeds the limit.

[0027] The defect classification module includes a preliminary defect detection unit and a refined classification unit. The preliminary defect detection unit performs preliminary anomaly detection based on time series and trains a threshold prediction model based on production conditions and historical image features, dynamically generating discrimination thresholds for each feature for the current condition. Specifically, the module collects visual features and process parameters of defect-free copper foil to build a historical sample database. An indicator threshold model is established based on a support vector regression model, taking process parameters (including shear speed, tension, and blade gap) and environmental parameters (including temperature and humidity) as input data. The maximum threshold of all indicators in the indicator design module is output as the discrimination threshold. If any index exceeds the sum of the current maximum threshold and a 5% redundancy, a preliminary anomaly warning is triggered. For example, for every 1°C increase in temperature, the Lyapunov exponent threshold for edge jitter increases by 0.01. Visual features of defect-free copper foil include edge straightness ≤ 0.1mm and cut burr height ≤ 5μm. Five defect process parameters include tension 100±5N and shear speed 5±0.2m / s.

[0028] The fine classification unit performs fine defect judgment based on the component sequence; specifically, it encodes the shear speed, tension, and blade gap into a 10-dimensional process feature vector, combines the indicators of each wavelet component into a 15-dimensional component feature vector (3 components × 5 indicators), uses the copper foil defect dataset as the training dataset, adopts the ResNet-18 convolutional neural network, inputs the W × H pixel incision image and the 25-dimensional feature vector, and outputs the defect category and confidence.

[0029] The feedback control module is used to trigger the shearing machine control strategy when both the initial inspection and the supplementary inspection are determined to be abnormal: the cutter pressure and feed speed are automatically adjusted through the PLC system (each adjustment step is 0.05mm) until any index is less than or equal to the sum of the maximum threshold of the current output and the 5% redundancy. At the beginning of each shift, the camera field of view and light source brightness are automatically calibrated using standard copper foil.

[0030] The setting values ​​of the thresholds and weights may be set by default according to the present invention, or may be set by those skilled in the art.

[0031] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0032] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, 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 processor, 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 flowcharts and / or block diagrams. 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.

[0033] 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.

[0034] 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.

[0035] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.

Claims

1. The online quality detection and feedback control system for copper foil shearing based on machine vision is characterized by: include: The data acquisition module is used to deploy multiple industrial cameras at key locations on the copper foil shearing machine to capture images of the shearing edge and surface in real time, and is also equipped with tension sensors, speed sensors, and temperature sensors. The compensation and correction module is used to organize the surface image into a time series and perform multi-scale wavelet decomposition on each frame image in the time series to obtain detail components; Indicator design module, used to calculate the characteristic indicators of each time series and each detail component; Defect classification module, which obtains detailed defect classification judgment based on real-time feature indicators; The feedback control module is used to feed back the detection results to the shearing machine controller in real time to adjust the cutter pressure and feed speed.

2. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 1, characterized in that: The image acquisition module includes installing N industrial cameras along both sides of the assembly line before and after the shearing machine outlet, and performing illumination balancing, geometric distortion correction, and camera calibration compensation calibration on the original image. The camera is equipped with a ring LED light source and a polarization filter, and the focal length, principal point coordinates, and tangential distortion coefficient are calculated using a checkerboard calibration plate. The transformation matrix between the camera coordinate system and the shearing machine coordinate system is calculated. The image is converted to Lab color space, and contrast-limited adaptive histogram equalization is performed on the brightness channel. The copper foil shearing area is located based on template matching, and an ROI of size H×M pixels is intercepted to obtain the cut edge and surface area.

3. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 1, characterized in that: The compensation correction module includes a trust calculation unit and a correction execution unit; The trust calculation unit includes taking sensor nodes whose distance from the target sensor node in the production line layout is less than a set distance threshold as adjacent sensor nodes; obtaining quality parameters of each adjacent sensor node of the target sensor node; Obtain historical data of the target sensor node and its adjacent sensor nodes; The alarm accuracy of the target sensor node and its adjacent sensor nodes is calculated based on the historical data; and the trust degree of each adjacent sensor node is calculated based on the alarm accuracy.

4. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 3, characterized in that: The confidence calculation unit further includes counting the false alarm data detected as abnormal in the historical data but found to be normal after manual re-inspection, and obtaining the alarm accuracy rate according to the proportion of the difference between the total data volume and the false alarm data in the total data volume; The Pearson correlation coefficient of the data of the adjacent nodes and the target node is calculated to obtain the trust correlation between the target sensor and the adjacent sensors.

5. The machine vision-based online quality detection and feedback control system for copper foil shearing according to claim 3 is characterized by: The correction execution unit calculates a compensation amount of the target detection sensor based on the trustworthiness of each adjacent sensor; Based on the compensation amount, the quality parameter of the target sensor is compensated and corrected by weighted fusion correction of adjacent node data.

6. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 3, characterized in that: The compensation and correction module also includes arranging continuously collected image features, including a copper foil edge coordinate sequence and a defect area sequence, and process parameters, including a shear force time series, according to timestamps to form a sliding window with a length of N points; performing a three-layer Daubechies wavelet decomposition on the copper foil edge coordinate sequence to obtain an approximate component and a detail component; and decomposing the shear force time series to extract high-frequency components to identify impact force anomalies at the moment of shearing.

7. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 1, characterized in that: The index design module includes an edge jitter index, which is used to measure the degree of chaos of the edge coordinate sequence, calculates the defect distribution correlation dimension through the GP algorithm, calculates the texture information entropy through the entropy value of the image grayscale distribution, and obtains the shear force mutation coefficient through the ratio of the standard deviation of the detail component to the mean of the approximate component.

8. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 1, characterized in that: The defect classification module includes a preliminary defect detection unit and a fine classification unit; The preliminary defect detection unit performs preliminary anomaly detection based on time series, and trains a threshold prediction model based on production conditions and historical image features to dynamically generate discrimination thresholds for each feature for the current condition; Specifically, the following steps are taken: by collecting the visual features and process parameters of defect-free copper foil, a historical sample database is constructed; an indicator threshold model is established based on a support vector regression model, with process parameters and environmental parameters as input data. The process parameters include shear speed, tension and blade gap, and the environmental parameters include temperature and humidity. The maximum threshold of all indicators in the indicator design module is output as the discrimination threshold; if any index is greater than the sum of the current output maximum threshold and 5% redundancy, a preliminary abnormality warning is triggered.

9. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 8, characterized in that: The fine classification unit performs fine defect judgment based on the component sequence; specifically, it encodes the shear speed, tension, and blade gap into a 10-dimensional process feature vector, combines the indicators of each wavelet component into a 15-dimensional component feature vector, uses the copper foil defect dataset as the training dataset, adopts the ResNet-18 convolutional neural network, inputs the W×H pixel incision image and the 25-dimensional feature vector, and outputs the defect category and confidence.

10. The machine vision-based copper foil shearing online quality detection and feedback control system according to claim 1, characterized in that: The feedback control module is used to trigger the shearing machine control strategy when both the preliminary inspection and the supplementary inspection are determined to be abnormal: the cutter pressure and feed speed are automatically adjusted through the PLC system until any index is less than or equal to the sum of the maximum threshold of the current output and the 5% redundancy. At the beginning of each shift, the camera field of view and light source brightness are automatically calibrated using standard copper foil.

Images