Intelligent detection system for carton production line

Abstract

The invention relates to the technical field of carton production detection, and discloses a carton production line intelligent detection system comprising an industrial camera module used for collecting carton appearance images; the dynamic speed sensor is used for monitoring the conveyor belt speed; the mechanical vibration sensor is used for detecting vibration interference of the conveyor belt; the image processing module is used for correcting and analyzing the acquired image; the control module is used for feeding back and controlling the detection result in real time; and the storage module is used for data storage and model training. Through dynamic multi-physical coupling modeling, the influence of optical distortion, speed fluctuation and vibration interference on size detection is comprehensively considered. Through a distributed optimization algorithm and a real-time feedback control mechanism, various errors in a dynamic environment are effectively compensated, the size detection precision reaches the micron level, a high-speed sensor and an industrial camera module are adopted for data acquisition, a detection algorithm is optimized through a distributed parallel computing technology, and the size detection precision is improved. And the response time of the system is controlled within milliseconds.

Description

Technical Field

[0001] The present invention relates to the technical field of carton production detection, and in particular to an intelligent detection system for a carton production line. Background Art

[0002] With the continuous development of industrial automation, cartons, as an important material for modern logistics and commodity packaging, have an increasing demand for production quality inspection. In carton production lines, size inspection is one of the key links to ensure carton quality. However, since carton production lines usually run at high speeds and have complex production environments, traditional inspection methods are difficult to meet the requirements of real-time and high precision.

[0003] Traditional carton size detection methods usually rely on manual measurement or static detection equipment. These methods have the following major defects:

[0004] Lack of real-time performance: Manual inspection and static inspection equipment cannot adapt to the dynamic production environment of high-speed conveyor belts and cannot complete the inspection task of a large number of cartons in a short period of time.

[0005] Limited detection accuracy: Static detection methods are usually unable to eliminate the impact of dynamic factors such as optical distortion, vibration interference and speed fluctuation on the detection results, resulting in large dimensional measurement errors.

[0006] Low degree of automation: Traditional methods require a lot of manual participation, which is not only inefficient but also easily affected by human subjective factors, making it difficult to meet the needs of modern industry for efficient and intelligent testing. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides an intelligent detection system for a carton production line, which solves the problem of insufficient size detection accuracy and real-time performance in a carton production line under a dynamic environment due to various interferences such as conveyor belt speed fluctuations, mechanical vibrations and optical distortion.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent detection system for a carton production line, comprising:

[0009] Industrial camera module, used to capture images of carton appearance;

[0010] Dynamic speed sensor to monitor conveyor belt speed;

[0011] Mechanical vibration sensors to detect conveyor belt vibration disturbances;

[0012] An image processing module, used for correcting and analyzing the collected images;

[0013] A control module, used for real-time feedback and control of the test results;

[0014] Storage module, used for data storage and model training.

[0015] Preferably, the industrial camera module adopts a pinhole camera model, the parameters of the model are obtained through calibration, and image distortion is corrected.

[0016] Preferably, the dynamic speed sensor is based on a speed fluctuation model, which includes a reference value, fluctuation amplitude, frequency and phase offset of the conveyor belt speed.

[0017] Preferably, the mechanical vibration sensor describes the vibration characteristics of the conveyor belt based on a single degree of freedom dynamic equation, and the dynamic equation includes the mass of the conveyor belt, the damping coefficient, the stiffness and the external interference force.

[0018] Preferably, the intelligent detection method for a carton production line comprises the following steps:

[0019] Capture carton images through industrial cameras to obtain initial data for size detection;

[0020] Monitor the speed fluctuation of the conveyor belt through a dynamic speed sensor and record the fluctuation data;

[0021] Detect vibration disturbance of the conveyor belt through mechanical vibration sensors;

[0022] Establishing a dynamic multi-physics coupling model based on the image data, velocity fluctuation data and vibration data;

[0023] Construct an optimization objective function and decompose it into multiple sub-problems through Lagrangian relaxation method for parallel solution;

[0024] Conduct real-time feedback control on the solution results and dynamically adjust the detection parameters to optimize the detection accuracy and efficiency.

[0025] Preferably, the dynamic multi-physics coupling model includes an optical distortion compensation model, a velocity fluctuation compensation model and a vibration interference compensation model.

[0026] Preferably, the optical distortion compensation model is established according to the intrinsic parameter matrix and distortion parameters of the pinhole camera, and the distortion parameters are solved by nonlinear optimization.

[0027] Preferably, the speed fluctuation compensation model calculates the error correction value of the speed fluctuation on the detection size by integration according to the influence of the conveyor belt speed fluctuation on the detection result.

[0028] Preferably, the vibration interference compensation model estimates the impact of conveyor belt vibration on the detection results in real time through Kalman filtering, and dynamically adjusts the detection parameters.

[0029] Preferably, the optimization objective function is decomposed into an optical distortion correction subproblem, a velocity fluctuation compensation subproblem and a vibration interference correction subproblem by Lagrangian relaxation method, and global optimization is achieved by updating the dual variables.

[0030] The present invention provides an intelligent detection system for a carton production line. It has the following beneficial effects:

[0031] 1. The present invention comprehensively considers the influence of optical distortion, speed fluctuation and vibration interference on size detection through dynamic multi-physics coupling modeling. Through distributed optimization algorithm and real-time feedback control mechanism, various errors in dynamic environment are effectively compensated, so that the accuracy of size detection reaches micron level. High-speed sensor and industrial camera module are used for data acquisition, and the detection algorithm is optimized through distributed parallel computing technology, so that the response time of the system is controlled within millisecond level. It ensures that the system can complete the detection and compensation of carton size in real time on the high-speed production line.

[0032] 2. The present invention uses dynamic speed sensors and mechanical vibration sensors to monitor the operating status of the production line in real time, and combines Kalman filtering to smooth the noise data, so that the system can adapt to complex dynamic environments such as conveyor belt speed fluctuations and vibration interference, and maintain the stability of detection.

[0033] 3. The storage module of the present invention supports classified storage and long-term management of data, and realizes batch analysis of historical data and dynamic optimization of model parameters in combination with edge computing devices. The system can not only improve the current detection performance, but also provide decision support for the continuous improvement of the production line by analyzing historical data.

[0034] 4. The present invention has abnormality detection and alarm functions. When speed fluctuations or vibrations exceed a safe range, the present invention can promptly trigger an alarm, prompting maintenance personnel to check the equipment status, thereby reducing downtime caused by equipment failure and ensuring stable operation of the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a system flow chart of the present invention;

[0036] Figure 2 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Embodiment 1:

[0039] Please see attached Figure 1 The embodiment of the present invention provides an intelligent detection system for a carton production line, comprising:

[0040] Industrial camera module, used to capture images of carton appearance;

[0041] The industrial camera module involved in this embodiment is an important component of the intelligent detection system of the carton production line. Its main function is to collect the appearance image of the carton and provide a data basis for subsequent size detection and error compensation. The industrial camera module works together with the dynamic speed sensor and the image processing module to collect and preprocess the carton images in a dynamic conveyor belt environment. In order to achieve high-precision detection, the industrial camera module establishes a mapping relationship between the image and the physical space through a pinhole camera model, and combines the optical distortion correction algorithm to eliminate detection errors. It should be noted that the working mode of this module is designed for the dynamic environment characteristics of the carton production line, and its operation is based on high-resolution image acquisition and rapid correction processing.

[0042] In this embodiment, the industrial camera module adopts a pinhole camera model, the basic principle of which is to map a three-dimensional point in space to an image plane through perspective projection. Specifically, the relationship between the spatial coordinate point (x, y, z)(x, y, z)(x, y, z) and the image plane coordinate (u, v)(u, v)(u, v) can be expressed in the following matrix form:

[0043] in:

[0044] f x ,f y f x ,f y f x ,f y is the focal length of the camera;

[0045] c x ,c y c x ,c y c x ,c y is the principal point offset, i.e. the position of the image center in the image coordinate system;

[0046] z is the depth value from the point to the camera optical axis.

[0047] As an option, in this embodiment, the intrinsic parameter matrix of the industrial camera module is obtained through a camera calibration process. The calibration process includes taking multiple calibration plate images, using the least squares method to fit the relationship between the image coordinates and the physical space coordinates of the calibration plate points, and thus determining parameters such as focal length and principal point offset. It should be noted that these parameters are stable in a fixed environment, but may be affected by vibration or optical distortion in a dynamic environment.

[0048] In order to solve the above problems, in some embodiments, the industrial camera module is combined with a distortion correction function, which is specifically processed using the following model:

[0049] r′=r(1+k1r 2 +k2r 4 +k3r 6 ),

[0050] in:

[0051] is the distance from the pixel to the center of the image; is the distance from the pixel to the center of the image;

[0052] k1, k2, k3 are radial distortion coefficients.

[0053] In one possible implementation, distortion correction is performed by the following steps:

[0054] The radial distortion parameters k1, k2, k3 are obtained using the calibration process.

[0055] Apply a back-projection algorithm to the acquired image coordinates (u,v)(u,v)(u,v) to map them back to the distortion-corrected coordinates (u′,v′)(u',v')(u′,v′);

[0056] The corrected image data is passed to the image processing module for further analysis.

[0057] For example, in the image acquisition of the industrial camera module, the acquisition rate is 120 frames per second and the resolution reaches 1920×1080 pixels. In order to cope with the speed fluctuation of the dynamic conveyor belt, the acquisition time of the camera in this embodiment is synchronously adjusted by the control module to ensure that the acquired image is consistent with the running state of the conveyor belt.

[0058] It should be noted that the light source system of the industrial camera module adopts ring LED lighting, which can provide a uniform lighting environment and reduce the impact of ambient light changes on image acquisition. Specifically, the intensity of the light source is adjusted in real time by the data feedback of the dynamic speed sensor and the mechanical vibration sensor to adapt to different production conditions. For example, when the conveyor belt speed is faster, the light source intensity automatically increases to compensate for underexposure.

[0059] As an extended implementation method, some embodiments also introduce a multi-camera array system to capture multi-view images of cartons. Specifically, multiple industrial cameras are arranged at fixed angles to capture images of different sides of the carton, and after coordinate alignment, complete three-dimensional reconstruction data is generated. The introduction of the multi-camera array further improves the system's dimensional detection accuracy, especially in the detection of the edge area of ​​the carton.

[0060] It is understandable that the specific implementation of the industrial camera module can be adjusted according to the specific needs of the production line. For example, in scenarios with low detection accuracy requirements, a lower-cost camera module is used, while in high-precision detection requirements, a camera device with higher resolution and larger dynamic range is selected. The core technical features of the above technical solution, including the pinhole camera model, optical distortion correction, dynamic synchronous acquisition, etc., provide reliable image data support for the present invention.

[0061] In this embodiment, the industrial camera module realizes the efficient operation of the intelligent detection system of the carton production line by cooperating with other modules and combining the real-time feedback mechanism in a dynamic environment. The specific collaboration process includes:

[0062] Synchronize with the dynamic speed sensor to adjust the acquisition time point;

[0063] Interact with the image processing module to complete image distortion correction and size measurement;

[0064] Collaborate with the control module to dynamically optimize light intensity and camera parameters.

[0065] Through the above design, the present invention can achieve high-precision detection of carton dimensions in a complex dynamic production environment, fully demonstrating the key role of the industrial camera module in the entire system.

[0066] Dynamic speed sensor to monitor conveyor belt speed;

[0067] The dynamic speed sensor plays a key role in the intelligent detection system of the carton production line of the present invention. It is mainly used to monitor the running speed of the conveyor belt in real time, obtain the speed fluctuation information of the conveyor belt, and transmit the data to the image processing module and the control module. The introduction of the dynamic speed sensor enables the system to dynamically adjust the image acquisition time and other parameters according to the change of the conveyor belt speed, thereby effectively compensating for the size detection error caused by speed fluctuation. It should be noted that the dynamic speed sensor provides accurate dynamic compensation data for the detection system through the synergy with the industrial camera module and the image processing module.

[0068] In this embodiment, the dynamic speed sensor uses a laser speed meter as the main detection device. Its principle is based on the Doppler effect. By emitting a laser beam and receiving a reflected signal, the speed of the conveyor belt is calculated. The speed measurement formula is:

[0069]

[0070] in:

[0071] v is the real-time speed of the conveyor belt;

[0072] f s is the received signal frequency offset;

[0073] λ is the laser wavelength.

[0074] Specifically, the laser speed meter is installed on the side of the conveyor belt, emitting laser light facing the surface of the conveyor belt, and calculating the speed through the built-in signal processing module after receiving the reflected signal. The system samples the speed value at a time interval of 1ms to ensure accurate monitoring of the high-speed conveyor belt.

[0075] As an option, in this embodiment, the output data of the dynamic speed sensor is combined with the conveyor belt speed fluctuation model to further analyze the impact of speed fluctuation on size detection. The speed fluctuation model describes the dynamic change of the conveyor belt speed, and its form is:

[0076] v(t)=v0+∈sin(ωt+φ),

[0077] in:

[0078] v0 is the base speed of the conveyor belt;

[0079] ∈sin is the velocity fluctuation amplitude;

[0080] ωt is the fluctuation frequency;

[0081] φ is the phase shift.

[0082] It should be noted that the speed measurement result of the dynamic speed sensor is transmitted to the image processing module to calculate the dimensional measurement error caused by speed fluctuation. The error is compensated by the following formula:

[0083]

[0084] in:

[0085] Δx speed is the error caused by speed fluctuation;

[0086] t1,t2t1,t2t1,t2 is the time range of image acquisition.

[0087] In one possible implementation, the data output of the dynamic speed sensor is transmitted to the control module via a high-speed data interface (such as CAN bus or Ethernet). Based on the speed data, the control module dynamically adjusts the trigger timing of the camera to ensure that the cartons corresponding to each frame of image acquisition are in the same physical position. This dynamic synchronization mechanism significantly improves the accuracy and robustness of detection.

[0088] For example, in some embodiments, the dynamic speed sensor also includes a multi-point speed measurement function, that is, speed data is collected at multiple locations of the conveyor belt at the same time to analyze the speed consistency of different locations. Specifically, the multi-point speed measurement results can help identify abnormal conditions in the operation of the conveyor belt, such as a section of the conveyor belt that is too fast or too slow, which may be caused by mechanical failure or uneven load.

[0089] It should be noted that, in order to cope with complex dynamic environments, the dynamic speed sensor of this embodiment is combined with a Kalman filter algorithm to perform real-time smoothing of noise data. The update formula of the Kalman filter is:

[0090] x k|k =x k|k-1 +K k (z k -Hx k|k-1 ),

[0091] K k =P k|k-1 H T (HP k|k-1 H T +R) -1

[0092] in:

[0093] x k|k is the state estimate after filtering; z k is the actual measured value; K k is the Kalman gain; H is the measurement matrix; P k|k-1 is the prediction error covariance; R is the measurement noise covariance.

[0094] It can be understood that the speed data processed by Kalman filtering is smoother, which can significantly reduce the impact of environmental noise on speed measurement accuracy, thereby improving the effect of error compensation.

[0095] In another possible implementation, the output data of the dynamic speed sensor can also be used for predictive maintenance. For example, when it is detected that the frequency of conveyor belt speed fluctuations increases abnormally or the fluctuation amplitude continues to increase, the system can trigger an alarm signal to prompt maintenance personnel to check the operating status of the conveyor belt or the wear of mechanical parts.

[0096] It should be emphasized that the dynamic speed sensor in the present invention is not limited to the laser speed measurement method. In some embodiments, other speed measurement technologies may also be used, such as:

[0097] Adopt photoelectric encoder to calculate the conveyor belt speed by recording the encoder pulse signal;

[0098] Video speed measurement technology is used to estimate the speed by analyzing the image feature displacement of the conveyor belt surface.

[0099] The selection of the above-mentioned different speed measurement technologies can be flexibly adjusted according to the specific production environment and cost requirements.

[0100] In this embodiment, the dynamic speed sensor provides reliable dynamic compensation data support for high-precision detection of carton size by working in conjunction with the industrial camera module, image processing module and control module. The design of the dynamic speed sensor fully considers the complex operating environment of the conveyor belt, and through advanced algorithms and data processing technology, it realizes accurate monitoring of dynamic changes.

[0101] Mechanical vibration sensors to detect conveyor belt vibration disturbances;

[0102] The mechanical vibration sensor is mainly used in the intelligent detection system of the carton production line of the present invention to monitor the mechanical vibration generated during the operation of the conveyor belt, and transmit the vibration interference signal to the image processing module and the control module. By acquiring the vibration characteristic data of the conveyor belt in real time, the system can analyze the impact of vibration on the carton size detection, and then eliminate the error caused by vibration interference through dynamic compensation methods. It should be noted that the mechanical vibration sensor cooperates with the dynamic speed sensor and the image processing module to form the basic data source for dynamic multi-physics field modeling, providing support for high-precision detection of the system.

[0103] In this embodiment, the mechanical vibration sensor is designed based on the acceleration sensor and is installed at the key parts of the conveyor belt (such as the conveyor belt support shaft or the tensioning mechanism) to monitor the vibration characteristics of the conveyor belt in real time. Specifically, the sensor measures the vibration acceleration and calculates the displacement and frequency of the vibration in combination with the dynamic equation. The vibration characteristics follow the single degree of freedom dynamic model, which is described as:

[0104]

[0105] Where: m is the equivalent mass of the conveyor belt; c is the damping coefficient; k is the stiffness of the conveyor belt; F(t) is the external interference force; x are the acceleration, velocity and displacement of vibration respectively.

[0106] It should be noted that in this embodiment, the acceleration signal is integrated to calculate the vibration velocity and displacement to obtain vibration characteristic parameters, including amplitude, frequency, vibration mode, etc.

[0107] As an option, the data output frequency of the mechanical vibration sensor is set to 1kHz to ensure that high-frequency vibration signals can be captured. In addition, the sensor data is transmitted to the image processing module via a data interface (such as CAN bus or Ethernet) for further analysis of the impact of vibration on carton detection.

[0108] Specifically, during the vibration compensation process, the vibration displacement Δx vibration It is used to correct the error of size detection, and its calculation formula is:

[0109] Δx vibration =Asin(2πft+θ),

[0110] Where: A is the vibration amplitude; f is the vibration frequency; θ is the initial phase.

[0111] In one possible implementation, the mechanical vibration sensor is combined with a Kalman filter algorithm to smooth vibration data in real time and reduce the interference of environmental noise. The state prediction and update formula of the Kalman filter is as follows:

[0112] x k|k =x k|k-1 +K k (z k -Hx k|k-1 ),

[0113] K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ,

[0114] Where: x k|k is the state estimate after filtering; z k is the acceleration measurement value at the current moment; K k is the Kalman gain; P k|k-1 is the prediction error covariance; H is the measurement matrix; R is the measurement noise covariance.

[0115] Through Kalman filtering processing, the system can extract the main features from the vibration signal while suppressing noise interference, providing accurate vibration parameters for the dynamic compensation model.

[0116] For example, in some embodiments, the mechanical vibration sensor is also used to diagnose the status of the conveyor belt. For example, when it is detected that the vibration frequency increases abnormally or the vibration amplitude continues to increase, the system can determine that the conveyor belt may have a mechanical fault, such as looseness, bearing wear, or uneven load. At this time, the system triggers an alarm signal to prompt the maintenance personnel to check the operating status of the conveyor belt.

[0117] It should be noted that, in order to adapt to different production environments, the mechanical vibration sensor in this embodiment can select different acceleration ranges and resolutions according to specific needs. For example, in a high-speed conveyor belt operation environment, a sensor with a larger range is preferred, while in a low-speed operation environment, a high-resolution sensor can be selected to capture tiny vibrations.

[0118] In another possible implementation, the data from the mechanical vibration sensor can also be used to improve the design of the conveyor belt. For example, through the long-term collection of vibration data, the vibration mode of the conveyor belt can be analyzed, and the support structure and tensioning device can be optimized to reduce the vibration intensity and fundamentally reduce the impact of vibration interference on dimensional detection.

[0119] It is understandable that the function of the mechanical vibration sensor is not limited to data acquisition, but also includes the coordination with other modules. Specifically:

[0120] The data from the mechanical vibration sensor is combined with the data from the dynamic speed sensor to comprehensively analyze the running stability of the conveyor belt;

[0121] After the data is transmitted to the image processing module, it participates in the dynamic compensation calculation of dimensional error together with the optical distortion correction data;

[0122] Vibration data is fed back to the actuator through the control module to adjust the conveyor belt tension to reduce the vibration amplitude.

[0123] In this embodiment, the mechanical vibration sensor provides reliable technical support for the dynamic compensation of the intelligent detection system of the carton production line through accurate vibration monitoring and real-time data processing. The sensor design fully considers the complex operating environment and actual application requirements, and its implementation method has high adaptability and practicality, laying the foundation for the high-precision operation of the system.

[0124] An image processing module, used for correcting and analyzing the collected images;

[0125] The image processing module is one of the core functional modules in the intelligent detection system of the carton production line of the present invention. Its main function is to process and analyze the image data collected by the industrial camera module. Through a series of algorithm processing such as optical distortion correction, multi-physics field modeling, size detection and error compensation, the image processing module provides key technical support for high-precision detection of carton dimensions. It should be noted that the image processing module completes the detection task in real time by working in collaboration with the industrial camera module, dynamic speed sensor, mechanical vibration sensor and control module, while ensuring a balance between detection accuracy and efficiency.

[0126] In this embodiment, the image processing module adopts an optical distortion correction method based on a pinhole camera model to eliminate the influence of image distortion on size detection. Specifically, the pinhole camera model describes the relationship between a point in a three-dimensional space and a point in an image plane, and its mathematical expression is:

[0127]

[0128] Where: f x ,f y is the focal length of the camera; c x , c y is the position of the principal point on the image plane; (x, y, z) is the point in physical space; (u, v) is the point on the image plane.

[0129] As an option, the image processing module corrects the acquired image data, including compensation of radial distortion. The radial distortion is described by the following function:

[0130] r′=r(1+k1r 2 +k2r 4 +k3r 6 ),

[0131] in: is the distance from the pixel to the center of the image; k1, k2, k3 are radial distortion coefficients.

[0132] In a possible implementation, the image processing module obtains the camera's intrinsic matrix and distortion parameters through a calibration process, and uses a back-projection method to correct the distorted image coordinate point (u, v) to a non-distorted coordinate point (u′, v′). It should be noted that this process significantly improves the accuracy of size detection.

[0133] Specifically, the image processing module also combines dynamic multi-physics modeling to comprehensively compensate for the effects of optical distortion, velocity fluctuation and vibration interference on dimensional measurement. In this embodiment, the error of the dynamic multi-physics model is described as:

[0134] Δx=Δx optical +Δx speed+Δx vibration ,

[0135] Where: Δx optical is the error caused by optical distortion; Δx speed is the error caused by speed fluctuation; Δx vibration The error is caused by mechanical vibration.

[0136] Exemplarily, in some embodiments, the image processing module calculates the speed fluctuation compensation value based on the speed fluctuation data by the following formula:

[0137]

[0138] in:

[0139] Δv(t) represents the deviation of the conveyor belt speed relative to the reference speed; t1, t2 are the time ranges for image acquisition.

[0140] In order to further eliminate vibration interference, the image processing module uses the vibration characteristic data provided by the vibration sensor to calculate the vibration compensation value through the following formula:

[0141] Δx vibration =Asin(2πft+θ),

[0142] Where: A is the vibration amplitude; f is the vibration frequency; θ is the initial phase of vibration.

[0143] In a possible implementation, the image processing module also uses a Kalman filter algorithm to perform real-time smoothing on the sensor data. The update formula of the Kalman filter is:

[0144] x k|k =x k|k-1 +K k (z k -Hx k|k-1 ),

[0145] K k =P k|k-1 HT(HP k|k-1 H T +R) -1 ,

[0146] Where: x k|k is the state after filtering; z k is the measured value at the current moment; K k is the Kalman gain.

[0147] It can be understood that the data processed by the Kalman filter algorithm is smoother and provides high-quality input data for the multi-physics field model.

[0148] It should be noted that the image processing module in this embodiment not only supports real-time detection, but also can store and analyze historical data. In some embodiments, the image processing module is combined with the storage module to train and optimize model parameters using stored image and sensor data to adapt to different production conditions. For example, in high-speed operation mode, the sampling frequency and parameters of the optical distortion correction algorithm are adjusted; in low-speed operation mode, the resolution is increased to detect more subtle size changes.

[0149] In another possible implementation, the image processing module can also be extended to support multi-camera image fusion. Specifically, after the image data from multiple cameras are aligned, a three-dimensional reconstruction model of the carton can be generated, thereby improving the comprehensiveness and accuracy of size detection.

[0150] In this embodiment, the image processing module achieves high-precision detection of carton size and ensures stable operation of the system in a complex dynamic environment by collaborating with the industrial camera module, dynamic speed sensor and mechanical vibration sensor, combined with dynamic modeling and optimization algorithms. It should be emphasized that the implementation of this module can be adjusted according to actual production needs to meet the accuracy and real-time requirements in different scenarios. The above design and algorithm provide core technical support for the present invention.

[0151] A control module, used for real-time feedback and control of the test results;

[0152] The control module is an important component of the intelligent detection system of the carton production line of the present invention. Its main function is to dynamically adjust the system operating parameters according to the detection data provided by the image processing module and the sensor, so as to realize the real-time optimization and compensation of the carton size detection process. Through the feedback control mechanism, the control module can effectively respond to changes in the dynamic environment, including conveyor belt speed fluctuations, vibration interference and changes in lighting conditions, thereby ensuring the high-precision operation of the system in a complex production environment. It should be noted that the control module works closely with the industrial camera module, dynamic speed sensor, mechanical vibration sensor and image processing module, and is the core control unit for realizing intelligent detection.

[0153] In this embodiment, the control module uses a real-time feedback control algorithm combined with reinforcement learning technology to dynamically adjust the key parameters of the system. The input of the control module includes the data collected by the sensor in real time and the error information generated by the image processing module, and the output includes the optimized values ​​of the camera acquisition timing, light source intensity and conveyor belt operation parameters. Through feedback control and optimization, the system can achieve a balance between size detection accuracy and efficiency under dynamic conditions.

[0154] Specifically, the control module uses the error feedback control model to optimize the detection parameters, which is mathematically expressed as:

[0155] u(t)=-K·e(t),

[0156] Where: u(t) is the control input (such as camera frame rate adjustment, light source intensity adjustment); e(t) is the current error, that is, the deviation between the system detection value and the ideal value; K is the feedback gain matrix.

[0157] As an option, the control module in this embodiment dynamically adjusts the camera trigger timing to achieve accurate synchronization between camera acquisition and the running state of the conveyor belt. Specifically, when the dynamic speed sensor detects that the conveyor belt speed fluctuates, the control module corrects the camera acquisition time in real time according to the fluctuation data to ensure that the acquisition position of each frame image is consistent. It should be noted that this synchronization mechanism significantly improves the stability of size detection.

[0158] In one possible implementation, the control module also combines a reinforcement learning algorithm to globally optimize system parameters. The reinforcement learning algorithm learns the optimal control strategy to reduce errors and energy consumption by training the policy function. Its core mathematical model is based on the Bellman equation, which is described as:

[0159] Q(s,a)=r+γmax a′ Q(s′,a′),

[0160] in:

[0161] Q(s,a) is the reward function of state s and action a; r is the immediate reward, which indicates the amount by which the current error is reduced; γ is the discount factor, which indicates the influence weight of future rewards.

[0162] Exemplarily, in some embodiments, the control module uses a deep Q network (DQN) to implement a reinforcement learning strategy. Specifically, DQN approximates Q(s, a) through a neural network and dynamically updates the policy function based on real-time error data. This method can quickly converge to the optimal control strategy in a complex dynamic environment.

[0163] It should be noted that the control module also considers multi-objective constraints during the optimization process, such as the trade-off between detection accuracy, system real-time performance, and resource consumption. The control module constructs a multi-objective optimization function and solves it using the Lagrangian relaxation method. Its mathematical model is:

[0164]

[0165] Where: J(x,u) is the optimization objective function, including detection error, detection time and resource consumption; g i (x,u) is the system constraint; λ i is the Lagrange multiplier.

[0166] In another possible implementation, the control module also integrates dynamic light source adjustment function to adapt to different lighting conditions in the production environment. For example, when the sensor detects insufficient light, the control module automatically increases the light intensity; when the conveyor speed increases, the camera exposure time is shortened to avoid image blur. Through this dynamic adjustment mechanism, the system can maintain high-quality image acquisition when the lighting conditions change.

[0167] As an extended application, the control module can also be linked with other equipment on the production line. For example, when the abnormal increase in conveyor belt vibration is detected, the control module can trigger the conveyor belt tension regulator to automatically adjust the tension to reduce the vibration amplitude. In addition, the control module can also collaborate with the production line management system to upload the detection data to the cloud for real-time monitoring and statistical analysis.

[0168] It is understandable that the control module in this embodiment is not limited to feedback control and reinforcement learning optimization, and its functions can be expanded according to the specific needs of the production line. For example:

[0169] Under the requirement of high-precision detection, the camera resolution and sampling frequency can be optimized and controlled;

[0170] In low-cost production lines, choose a more lightweight control strategy to reduce hardware requirements.

[0171] In this embodiment, the control module works in coordination with other modules through feedback control, reinforcement learning optimization and dynamic adjustment functions, significantly improving the robustness and adaptability of the intelligent detection system of the carton production line. The design of the control module fully considers the complexity in the dynamic environment, provides an efficient real-time optimization solution, and provides key support for the technical implementation of the present invention.

[0172] Storage module, used for data storage and model training.

[0173] The storage module in the carton production line intelligent detection system of the present invention is mainly used to store the data generated during the real-time detection process, including image data, sensor acquisition data, and detection result data. By managing and storing these data, the storage module supports the subsequent analysis, model training, and parameter optimization of the system. It should be noted that the storage module works closely with the image processing module and the control module, and is an important component for realizing a closed data loop. At the same time, this module provides a basis for the backtracking of historical data and the continuous optimization of the system.

[0174] In this embodiment, the storage module adopts a high-speed storage device and a distributed database to manage massive test data. The high-speed storage device is mainly used to store temporary data in the production process in real time, while the distributed database is used to store and retrieve historical data for a long time.

[0175] Specifically, the storage module classifies and manages the data collected in real time according to the data type. For example, the data is divided into the following three categories:

[0176] Image data: original image from industrial camera module and processed image after optical distortion correction.

[0177] Sensor data: includes real-time operating data collected by dynamic speed sensors and mechanical vibration sensors.

[0178] Detection result data: carton size detection results and error compensation data output by the image processing module.

[0179] As an option, the storage module adopts a hierarchical storage structure, storing real-time data in a cache and historical data in a distributed database. It should be noted that this architecture design can improve the access efficiency of historical data while ensuring real-time requirements.

[0180] In one possible implementation, the storage module is combined with an edge computing device to support local preprocessing of the collected data. For example, before the image data is stored, the storage module can compress and encode the image to reduce the storage space occupied. In addition, for the sensor data, the storage module can use a sliding window method to extract key feature values, such as the maximum deviation of speed fluctuations or the changing trend of vibration frequency.

[0181] Exemplarily, the storage module also supports batch analysis of historical data and model training. Specifically, the storage module transmits the error data recorded during the detection process to the training module for optimizing the parameters of the dynamic multi-physics model. The optimization objective function is:

[0182]

[0183] Where: x i Input data, such as speed fluctuations and vibration characteristics collected by sensors; i is the actual measurement result; f(x i ,θ) is the prediction result of the multi-physics field model, and θ is the model parameter.

[0184] It should be noted that through the data managed by the storage module, the model parameters can be dynamically adjusted to adapt to different production environments.

[0185] In another possible implementation, the storage module also supports abnormality detection and alarm functions. For example, when the system detects that speed fluctuations or vibration characteristics exceed a preset range, the storage module can mark the relevant data as abnormal data and trigger an alarm signal through the control module. This function can significantly improve the safety and reliability of production line operation.

[0186] As an extended application, the data of the storage module can also be used for statistical analysis of the operation status of the production line. For example, by analyzing historical data, key indicators such as the average speed of the conveyor belt, vibration intensity, and the qualified rate of carton size detection can be calculated. These statistical information can not only provide a decision-making basis for production line optimization, but also serve as an important data support for quality management.

[0187] It is understandable that the functions of the storage module are not limited to data storage. Its management and analysis capabilities provide key support for the efficient operation of the entire system. For example:

[0188] When the production line switches to cartons of different specifications, the storage module can quickly retrieve and load the corresponding model parameters, thereby reducing adjustment time.

[0189] In long-running scenarios, the storage module can periodically clean up historical data to free up storage space while retaining key data to meet long-term analysis needs.

[0190] It should be emphasized that the storage module also has high scalability. Its hardware configuration and storage strategy can be flexibly adjusted according to the scale of the production line and the complexity of the inspection task. For example, under the demand for high-precision inspection, higher capacity and higher performance storage devices can be selected; and in cost-constrained scenarios, resource consumption can be reduced by optimizing storage strategies.

[0191] In this embodiment, the storage module provides reliable data management and analysis capabilities for the intelligent detection system of the carton production line through data classification storage, hierarchical management and edge computing support. The storage module works in coordination with other modules to form a closed loop of data flow and analysis, which is an important basis for achieving high-precision and high-efficiency operation of the system. The above design ensures the adaptability and practicality of the storage module in different application scenarios.

[0192] Embodiment 2:

[0193] Please see attached Figure 2 The embodiment of the present invention provides an intelligent detection method for a carton production line, comprising the following steps:

[0194] S1. Collect carton images through industrial cameras to obtain initial data for size detection;

[0195] In this embodiment, an industrial camera is used to collect the appearance image of the carton to provide initial data for size detection. The industrial camera uses a high-definition camera with a resolution of 1920×1080 pixels and a collection rate of 120 frames per second. In order to ensure image quality, the camera is combined with an LED ring light source to provide uniform and stable lighting conditions.

[0196] Specifically, industrial cameras work based on the pinhole camera model, and the mapping relationship between their three-dimensional space points and image plane points is:

[0197]

[0198] Where: (x, y, z) is the physical coordinates of the space point; (u, v) is the image plane coordinates; f x , f y is the focal length, c x , c y is the image center offset.

[0199] In a possible implementation, the camera data is transmitted to the image processing module via a Gigabit Ethernet interface to ensure real-time performance and data integrity.

[0200] S2, monitor the speed fluctuation of the conveyor belt through a dynamic speed sensor and record the fluctuation data;

[0201] The speed fluctuation of the conveyor belt is monitored in real time through a dynamic speed sensor (such as a laser speed meter), and the speed data is recorded to provide a basis for subsequent dynamic compensation. The sensor samples the speed at intervals of 1ms, and the speed fluctuation monitoring model is:

[0202] v(t)=v0+∈sin(ωt+φ),

[0203] Where: v0 is the conveyor belt reference speed; ∈sin is the fluctuation amplitude; ω is the fluctuation frequency; φ is the phase offset.

[0204] Specifically, the dynamic speed sensor sends the collected data to the image processing module through a high-speed data bus (such as a CAN bus) to calculate the impact of speed fluctuations on dimensional measurement.

[0205] S3, detecting the vibration disturbance of the conveyor belt through a mechanical vibration sensor;

[0206] The vibration characteristics of the conveyor belt during operation are monitored by mechanical vibration sensors (such as accelerometers), mainly recording the vibration amplitude, frequency and phase. The sensor measures the vibration displacement based on a single degree of freedom dynamic model, which is described as:

[0207]

[0208] Where: m is the mass of the conveyor belt; c is the damping coefficient; k is the stiffness; F(t) is the external interference force.

[0209] The sensor data is smoothed by the Kalman filter algorithm, and the filtered vibration data is used as the input parameter for error compensation.

[0210] S4, establishing a dynamic multi-physics coupling model based on the image data, velocity fluctuation data and vibration data;

[0211] The image data, velocity fluctuation data and vibration data are integrated to establish a dynamic multi-physics coupling model to describe the impact of different interference factors on size detection. The total error description of the model is:

[0212] Δx=Δx optical +Δx speed +Δx vibration ,

[0213] in:

[0214] Δx optical The error is caused by optical distortion;

[0215] Δx speed is the error caused by speed fluctuation, calculated as:

[0216]

[0217] Where Δv(t) is the velocity deviation;

[0218] Δx vibration is the error caused by vibration interference, calculated as:

[0219] Δx vibration =Asin(2πft+θ),

[0220] Where A is the amplitude, f is the frequency, and θ is the initial phase.

[0221] S5. Construct an optimization objective function, decompose it into multiple sub-problems through Lagrangian relaxation method and solve them in parallel;

[0222] Based on the dynamic multi-physics coupling model, an optimization objective function is constructed to minimize detection error and resource consumption while meeting real-time constraints. The objective function is described as:

[0223] J(x,u)=α1||Δx|| 2 +α2C time (x,u)+α3C resource (u),

[0224] Where: C time (x,u) is the detection time; C resource (u) is resource consumption.

[0225] Through the Lagrangian relaxation method, the objective function is decomposed into multiple sub-problems, including:

[0226] Optical distortion correction sub-problem;

[0227] Speed ​​fluctuation compensation sub-problem;

[0228] Vibration disturbance compensation subproblem.

[0229] The dual function of the optimization problem is:

[0230]

[0231] where g i (x,u) represents the constraint condition, λ i is the Lagrange multiplier.

[0232] Exemplarily, the sub-problems are solved in parallel using a distributed algorithm, which significantly improves computational efficiency.

[0233] S6. Perform real-time feedback control on the solution results and dynamically adjust the detection parameters to optimize the detection accuracy and efficiency.

[0234] The optimization results are passed to the control module to adjust the detection parameters in real time, including camera frame rate, light intensity and conveyor speed. The control algorithm is based on the error feedback control principle, which is expressed as:

[0235] u(t)=-K·e(t),

[0236] Where: u(t) is the control input; e(t) is the error feedback value; K is the feedback gain matrix.

[0237] In one possible implementation, the control module combines the reinforcement learning algorithm to achieve global optimization by training the optimal policy function. The core model is:

[0238] Q(s,a)=r+γmax a′ Q(s,a′),

[0239] Where r is the reward function and γ is the discount factor.

[0240] This embodiment realizes intelligent detection of carton production lines through a comprehensive method of image acquisition, speed and vibration data monitoring, dynamic modeling, optimization solution and real-time control. All technical details are designed based on the actual operating environment to ensure that the method has good reproducibility and industrial application value. The collaborative work of the above steps builds a complete intelligent detection process, providing an efficient and reliable solution for carton production lines.

[0241] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent detection system for carton production line, characterized by: include: Industrial camera module, used to capture images of carton appearance; Dynamic speed sensor to monitor conveyor belt speed; Mechanical vibration sensors to detect conveyor belt vibration disturbances; An image processing module, used for correcting and analyzing the collected images; A control module, used for real-time feedback and control of the test results; Storage module, used for data storage and model training.

2. The intelligent detection system for carton production line according to claim 1 is characterized in that: The industrial camera module adopts a pinhole camera model, the parameters of the model are obtained through calibration, and image distortion is corrected.

3. The intelligent detection system for carton production line according to claim 1 is characterized in that: The dynamic speed sensor is based on a speed fluctuation model, which includes a reference value, fluctuation amplitude, frequency and phase offset of the conveyor belt speed.

4. The intelligent detection system for carton production line according to claim 1 is characterized in that: The mechanical vibration sensor describes the vibration characteristics of the conveyor belt based on a single degree of freedom dynamic equation, wherein the dynamic equation includes the mass of the conveyor belt, the damping coefficient, the stiffness and the external interference force.

5. An intelligent detection method for a carton production line, applied to a system as claimed in any one of claims 1 to 4, characterized in that: The following steps are involved: Capture carton images through industrial cameras to obtain initial data for size detection; Monitor the speed fluctuation of the conveyor belt through a dynamic speed sensor and record the fluctuation data; Detect vibration disturbance of the conveyor belt through mechanical vibration sensors; Establishing a dynamic multi-physics coupling model based on the image data, velocity fluctuation data and vibration data; Construct an optimization objective function and decompose it into multiple sub-problems through Lagrangian relaxation method for parallel solution; Conduct real-time feedback control on the solution results and dynamically adjust the detection parameters to optimize the detection accuracy and efficiency.

6. The intelligent detection method for a carton production line according to claim 5 is characterized in that: The dynamic multi-physics coupling model includes an optical distortion compensation model, a velocity fluctuation compensation model and a vibration interference compensation model.

7. The intelligent detection method for a carton production line according to claim 6 is characterized in that: The optical distortion compensation model is established according to the internal parameter matrix and distortion parameters of the pinhole camera, and the distortion parameters are solved by nonlinear optimization.

8. The intelligent detection method for a carton production line according to claim 6 is characterized in that: The speed fluctuation compensation model calculates the error correction value of the speed fluctuation on the detection size through integration according to the influence of the conveyor belt speed fluctuation on the detection result.

9. The intelligent detection method for a carton production line according to claim 6 is characterized in that: The vibration interference compensation model estimates the influence of conveyor belt vibration on the detection result in real time through Kalman filtering, and dynamically adjusts the detection parameters.

10. The intelligent detection method for a carton production line according to claim 5, characterized in that: The optimization objective function is decomposed into an optical distortion correction subproblem, a speed fluctuation compensation subproblem and a vibration interference correction subproblem by using the Lagrangian relaxation method, and global optimization is achieved by updating the dual variables.

Images