Intelligent anti-fake color box printing positioning control method

Abstract

The invention discloses an intelligent anti-counterfeiting color box printing positioning control method, and relates to the technical field of printing positioning control, and the method comprises the steps: dividing a color box layout into a plurality of detection regions, setting a cross-shaped and circular combined positioning mark in each region, and collecting positioning information in real time through a high-resolution image sensor to establish a feature database; extracting regional deformation and position features by using a multilayer convolutional neural network, calculating actual deviation by combining a residual network, constructing a mapping model of process parameters and position deviation, and establishing a differential compensation model by using a deep reinforcement learning algorithm; and the compensation parameters are distributed to a servo control unit, a motion control instruction is generated, a compensation effect evaluation index is adopted to carry out real-time evaluation on position deviation, a decision model is established to realize compensation parameter adaptive optimization, and regional compensation adaptive adjustment is realized. According to the method, intelligent monitoring and optimization of the anti-counterfeiting printing process are achieved, and the printing quality stability and the production efficiency of the anti-counterfeiting color box in batch production are greatly improved.

Description

Technical Field

[0001] The invention relates to the technical field of printing positioning control, in particular to an intelligent anti-counterfeiting color box printing positioning control method. Background Art

[0002] With the rapid development of anti-counterfeiting printing technology, the quality of printed products and the requirements for anti-counterfeiting effects are constantly increasing, and precise positioning control has become a key factor in ensuring printing quality. Existing printing positioning control technology mainly relies on traditional photoelectric detection and mechanical compensation methods, using fixed detection positioning marks and a single compensation parameter for adjustment. This method has obvious limitations when dealing with complex deformation characteristics. Especially in the multi-color overprinting process, due to the coupling of multiple factors such as paper tension fluctuations, temperature and humidity changes, and ink viscosity, the printed products show nonlinear dynamic deformation characteristics in different areas. Traditional positioning control methods are difficult to accurately capture and compensate for these deformations.

[0003] At the same time, the existing technology lacks systematic modeling and analysis of various process parameters in the printing process, and cannot establish a mapping relationship between parameter changes and position deviations, making the compensation strategy lack of pertinence and adaptability. In addition, in actual production, printed products from different batches and different regions often show differentiated deformation characteristics, and the traditional unified compensation method is difficult to meet the needs of differentiated adjustment, and lacks effective quality assessment and parameter optimization mechanisms, resulting in difficulty in fundamentally improving printing positioning accuracy and quality stability. Therefore, it is urgent to develop a new printing control method that can intelligently perceive deformation characteristics, adaptively adjust compensation parameters, and achieve precise positioning control. Summary of the invention

[0004] In view of the problems existing in the existing intelligent anti-counterfeiting color box printing positioning control method, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that it is difficult to meet the differentiated requirements of deformation characteristics in different regions and the lack of an intelligent compensation mechanism makes it difficult to ensure the overprinting accuracy and quality stability.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides an intelligent anti-counterfeiting color box printing positioning control method, which includes: dividing the color box layout into multiple detection areas, each area is provided with a combined positioning system including a cross mark and a circular mark, and positioning information is collected in real time by an image sensor to establish a detection area feature database; based on the feature database, a multi-layer convolutional neural network is used to extract the local deformation and global position features of each detection area, and the actual deviation value is calculated in combination with the residual network, and a mapping model between process parameters and position deviations is established. Through deep reinforcement learning, a differentiated compensation model is constructed and the compensation parameters are optimized to form a compensation database; based on the differentiated compensation database, the compensation model parameters are allocated to the corresponding servo motor control unit, the compensation adjustment amount is calculated and a control instruction is generated, and the position deviation is monitored in real time through evaluation indicators to achieve adaptive optimization of compensation parameters and regional compensation adjustment.

[0008] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, wherein: the combined positioning system includes a cross mark for positioning in the X and Y directions and a circular mark for rotation angle positioning; based on the divided detection area, an image sensor array is installed, and the field of view angle of the image sensor is adjusted to ensure that the field of view range of the image sensor completely covers the detection area, wherein the field of view angle is calculated by the ratio of the diagonal length value of the detection area to the focal length value of the sensor; the sampling frequency is set, and the image resolution is calculated by dividing the product of the number of sensor pixels and the actual physical size of the sampling area by the sensor imaging field size.

[0009] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, wherein: based on the image sensor array, the detection mark group image data is collected and preprocessed, and the image data includes grayscale images of cross marks and circular marks in the detection area; the preprocessing includes image enhancement and binarization processing; based on the binarized image data, the centroid coordinates of the detection mark group are obtained by calculating the weighted average of the coordinate values ​​of each pixel point in the mark area and its binarization value.

[0010] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method of the present invention, wherein: a tension sensor is synchronously deployed to collect paper tension parameters;

[0011] A detection area feature database is established based on the centroid coordinate calculation result, and the feature database includes a detection area identification code, centroid coordinates, tension parameters at the acquisition time, and image enhancement parameters k1 and k2.

[0012] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method of the present invention, the multi-layer convolutional neural network includes three convolutional layers and two fully connected layers, and the local deformation features, texture features and global position features of the detection area are extracted based on the three convolutional layers, and the dimensionality reduction processing is performed to obtain the feature vector of the detection area; the deviation value between the actual centroid position and the ideal centroid position of each detection area is calculated based on the feature vector, and a deviation calculation model is established using a residual network structure, wherein the X-direction deviation is obtained by subtracting the ideal standard position X-coordinate from the actual centroid X-coordinate, the Y-direction deviation is obtained by subtracting the ideal standard position Y-coordinate from the actual centroid Y-coordinate, and the angle deviation is calculated by dividing the X-direction deviation by the inverse tangent of the Y-direction deviation, and the output feature of the residual module is obtained by superimposing the residual mapping function and the input feature.

[0013] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, wherein: based on the feature database, tension parameters are extracted, printing speed and ambient temperature are collected in real time, and the parameters are normalized to form a process parameter set; a parameter mapping model of process parameters and position deviations is established based on a double-layer nonlinear mapping network, the process parameter set is input into the first layer mapping network and the first layer bias vector is superimposed, and after being processed by the activation function, a deviation value vector including X-direction deviation, Y-direction deviation and angle deviation is obtained through the second layer mapping, and finally the second layer bias vector is superimposed to obtain the prediction result.

[0014] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, wherein: based on the optimized parameter mapping model, a mapping relationship between process parameters and position deviations is established, and combined with the characteristic differences in different detection areas, an independent differentiated compensation model is constructed for each area, and the update of the compensation model parameters is achieved by superimposing the product of the learning rate of the parameter value at the current moment and the difference between the deviation values ​​at two adjacent moments; a differentiated compensation database is constructed, and the optimized compensation model parameters are stored in the differentiated compensation database.

[0015] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, wherein: the process parameter set is input into the compensation model parameters to obtain the compensation adjustment amount; the motion control instruction of the servo motor is generated based on the compensation adjustment amount; the motion control instruction includes the adjustment direction, the number of step pulses and the acceleration and deceleration timing, and the position sensor signal is collected during the compensation process to obtain the real-time position deviation value; based on the real-time position deviation value, a compensation effect evaluation index is established, and the evaluation index is obtained by multiplying the absolute value of the position deviation value, the absolute value of the deviation change at adjacent moments and the standard deviation of the deviation sequence by the corresponding evaluation weight coefficient and then summing them up.

[0016] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, the accuracy, responsiveness and stability of the compensation effect are comprehensively evaluated through weighted combination based on the effect evaluation index. If the effect evaluation index exceeds the preset threshold range, the compensation parameter optimization is triggered, and the optimized compensation model parameters are stored in the differentiated compensation database.

[0017] As a preferred solution of the intelligent anti-counterfeiting color box printing positioning control method described in the present invention, based on the data in the differential compensation database, the printing process is continuously monitored by a high-precision photoelectric encoder, and a real-time monitoring mechanism is established in combination with the calculation results of the compensation effect evaluation index.

[0018] The beneficial effect of the present invention is that the present invention adopts a mathematical model based on printing positioning deviation, combined with a differentiated compensation database and compensation effect evaluation indicators, to achieve intelligent control of the overprint accuracy of the anti-counterfeiting pattern. The multi-dimensional evaluation system based on the compensation effect evaluation index can accurately identify and timely compensate for various deviations in the printing process, effectively solving the problems of low overprint overlap of the anti-counterfeiting pattern, accumulation of overprint deviations between multiple printing units, and abnormal quality of anti-counterfeiting details. Intelligent monitoring and optimization of the anti-counterfeiting printing process is achieved, which greatly improves the printing quality stability and production efficiency of anti-counterfeiting color boxes in mass production, and provides reliable technical support for the large-scale application of intelligent anti-counterfeiting color boxes. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0020] Figure 1 The present invention is a flow chart of the intelligent anti-counterfeiting color box printing positioning control method. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1

[0025] Reference Figure 1 , which is the first embodiment of the present invention, and provides an intelligent anti-counterfeiting color box printing positioning control method, comprising:

[0026] S1: Perform regional recognition on the color box layout and divide it into multiple detection areas according to preset rules. A combined positioning system including a cross positioning mark and a circular positioning mark is set up in each area. The positioning mark information of each area is collected in real time through a high-resolution image sensor to establish a detection area feature database.

[0027] The color box layout is divided into N detection areas according to the preset size of 100-200mm×100-200mm. The detection area is determined by the regional boundary coordinates, and the boundary range of each detection area is accurately defined to provide a spatial reference for the detection mark group; the regional identification code R is marked at the boundary of the detection area i , used to distinguish different detection areas.

[0028] A detection mark group is set in the detection area, and the spatial arrangement coordinates of the detection mark group are obtained through the distribution function, which is expressed as:

[0029] P mn =(M c +mΔ x ,N c +nΔ y )

[0030] Among them, P mn Indicates the layout coordinates of the detection mark group in the detection area, M c and N c Respectively represent the center coordinate values ​​of the detection area on the X-axis and Y-axis, Δ x and Δ y They represent the spacing values ​​of adjacent detection mark groups in the X-axis and Y-axis directions respectively, and m and n represent the row number and column number of the detection mark group respectively.

[0031] The detection mark group includes a cross mark for positioning in the X and Y directions and a circular mark for positioning in the rotation angle, forming an array distribution to ensure uniform distribution of the detection mark group, facilitating image acquisition and feature extraction.

[0032] According to the divided detection area, a CCD image sensor array with a resolution of 2400dpi is installed on the printing equipment. To ensure that the field of view of the CCD image sensor completely covers the detection area, the field of view angle of the CCD image sensor is further calculated and expressed as:

[0033]

[0034] Where, θ represents the field of view of the CCD image sensor, L d Indicates the diagonal length of the detection area, f l Indicates the focal length value of the sensor.

[0035] Each sensor of the CCD image sensor array is matched with a detection area one by one to achieve multi-area synchronous acquisition.

[0036] The image data of the detection marker group is collected by the CCD image sensor array, the acquisition frequency is set, and the image resolution is calculated, which is expressed as:

[0037]

[0038] Among them, R s Indicates the actual image resolution, Q p Indicates the number of pixels of the sensor, S a represents the actual physical size of the sampling area, F v Indicates the size of the sensor's imaging field of view.

[0039] Synchronously collect paper tension parameters T based on tension sensor p , expressed as:

[0040]

[0041] Among them, T p Indicates the paper tension value per unit width, F s Indicates the tension value measured by the tension sensor, W p Indicates the paper width.

[0042] The tension value is collected by a strain type tension sensor installed at the paper feed roller. The sampling frequency is synchronized with the image acquisition and is used to record the stress state of the paper during the printing process. The sensitivity coefficient K of the strain type tension sensor is s satisfy:

[0043]

[0044] Among them, K s represents the sensor sensitivity coefficient, ΔV represents the change in sensor output voltage, V0 represents the sensor excitation voltage, and ε represents the strain value.

[0045] The detection mark group image data collected by the CCD image sensor array is preprocessed, and the image data includes grayscale images of the cross mark and the circular mark in the detection area.

[0046] Preprocessing includes image enhancement and binarization. Image enhancement is used to compensate for the lack of image contrast caused by uneven lighting in the printing environment. The formula is expressed as:

[0047] E(x,y)=k1I(x,y)+k2

[0048] Among them, E(x,y) represents the gray value after image enhancement, and I(x,y) represents the gray value after image enhancement with a resolution of R. s The original grayscale value collected by the CCD image sensor, k1 represents the contrast adjustment coefficient, and k2 represents the brightness adjustment coefficient.

[0049] The features of the detection marker group are highlighted by image enhancement, and the enhanced image is binarized, which is expressed as:

[0050]

[0051] Among them, B(x,y) represents the binary image, T h Represents the binarization threshold.

[0052] Based on the binary image data, the centroid coordinates of the detection marker group are calculated and expressed as:

[0053]

[0054] Among them, X g and Y g Respectively represent the X and Y coordinates of the centroid of the detection marker group, x i and i represents the coordinate value of the i-th pixel in the marked area, w i Indicates the binarized value of the corresponding pixel.

[0055] Based on the calculation results of the centroid coordinates, a detection area feature database is established. The feature database includes the detection area identification code R i , center of mass coordinates (X g ,Y g ), the tension parameter T at the acquisition time t p, image enhancement parameters k1 and k2 and other information; it can provide the center of mass coordinate value under standard working conditions as an ideal position reference, and provide basic data support for real-time anti-counterfeiting printing positioning control.

[0056] S2: Based on the feature database, a multi-layer convolutional neural network is used to extract the local deformation features and global position features of each inspection area to build a feature extraction model. The actual deviation value of each inspection area is calculated in combination with the residual network structure, and a multi-parameter coupling mapping model of process parameters and position deviation is established. A deep reinforcement learning algorithm is used to build a differentiated compensation model and dynamically optimize the compensation parameters to build a differentiated compensation database.

[0057] Perform deep feature extraction on the data in the feature database based on the centroid coordinates (X g ,Y g ), tension parameter T p , image enhancement parameters k1 and k2 construct the input matrix, and the local deformation features and global position features of each detection area are extracted through a multi-layer convolutional neural network to construct a feature extraction model.

[0058] The multi-layer convolutional neural network consists of three convolutional layers and two fully connected layers. The forward propagation process of the three convolutional layers is expressed as:

[0059] H n =σ(W n *H n-1 +B n )

[0060] F i =φ(H3)

[0061] Among them, H n represents the output feature matrix of the nth layer, W n Represents the convolution kernel weight matrix of the nth layer, B n represents the bias matrix of the nth layer, σ represents the ReLU activation function; φ represents the feature mapping function.

[0062] The first convolution layer of the three-layer convolution layer uses 32 3×3 convolution kernels to extract low-level features, the second layer uses 64 3×3 convolution kernels to extract middle-level features, and the third layer uses 128 3×3 convolution kernels to extract high-level features. Through the multi-layer structure, the local deformation features, texture features, and global position features of the detection area are extracted.

[0063] The local deformation features, texture features and global position features of the detection area are reduced in dimension through two fully connected layers. The first fully connected layer reduces the feature dimension to 512 dimensions and passes through the ReLU activation function. The second fully connected layer further reduces the feature dimension to 128 dimensions to obtain the detection area R i The feature vector of .

[0064] Calculate each detection area R based on the feature vector after dimension reduction i The deviation value between the actual centroid position and the ideal centroid position is calculated by using the residual network structure to establish a deviation calculation model, which is expressed as:

[0065] ΔX i =X g -X r

[0066] ΔY i =Y g -Y r

[0067]

[0068] Z out =G(Z in )+Z in

[0069] Among them, X g and Y g represents the actual centroid coordinates, X r and Y r represents the ideal standard position coordinate, ΔX i , ΔY i and Δθ i Respectively represent the deviation values ​​in the X direction, Y direction and angle, Z out represents the output feature of the residual module, G(Z in ) represents the residual mapping function, Z in Represents the input features.

[0070] Construct a multi-parameter coupling mapping model to establish the relationship between process parameters and position deviation, including:

[0071] Extract the tension parameter T from the detection area feature database p , real-time collection of printing speed V p and ambient temperature T e , these parameters are normalized to form the process parameter set P t ;

[0072] Due to the complex nonlinear relationship between process parameters and position deviation, a parameter mapping model is established through a double-layer nonlinear mapping network, which is expressed as:

[0073] M i =[ΔX i ,ΔY i ,Δθ i ]=U2·σ(U1P t +D1)+D2

[0074] Among them, Mi It represents the deviation value vector predicted by the model, U1 represents the first-layer mapping weight matrix with a dimension of [64×3], which is used to extract the primary features of the process parameters, U2 represents the second-layer mapping weight matrix with a dimension of [3×64], which is used to generate the final deviation prediction value, D1 and D2 represent the bias vectors of the two layers respectively, which are used to adjust the sensitivity of the mapping, and σ represents the activation function, which introduces nonlinear transformation capability.

[0075] The parameter mapping model is optimized by minimizing the loss function, which is expressed as:

[0076]

[0077] in, represents the actual observed deviation value vector, λ represents the regularization coefficient, and n represents the total number of training samples.

[0078] Continuously iterate and optimize the loss function to make the deviation value M predicted by the model i As close as possible to the actual observed value And keep the weight matrices U1 and U2 of moderate size.

[0079] Based on the optimized mapping model, the mapping relationship between process parameters and position deviation is established, and the established mapping relationship is combined with different detection areas R i Based on the differences in characteristics, an independent differentiated compensation model C is constructed for each region. i , the update equation of the compensation model parameters is expressed as:

[0080] C i (t+1)=C i (t)+β[M i (t)-M i (t-1)]

[0081] Among them, β represents the learning rate.

[0082] The decision evaluation equation is expressed as:

[0083]

[0084] Among them, C i (t) and C i (t+1) represents the compensation model parameters at time t and time t+1 respectively, β represents the learning rate parameter, which controls the step size of parameter update; M i (t) and M i (t-1) represents the deviation mapping value at time t and time t-1 respectively; Q(S t ,A t ) indicates that in the current state S t Take compensatory action A tThe total value, R t represents the immediate reward, γ is the discount factor, Indicates the next state S t+11 The maximum value among all possible actions.

[0085] By observing the current state S t , including the current deviation M i (t), current compensation parameter C i (t) and other process parameters, use the Q equation to evaluate possible adjustment actions, calculate the Q values ​​of different β values ​​and adjustment directions, and select the action with the largest Q value as the optimal decision; substitute the optimal decision into the compensation parameter update equation, use the selected β value, and calculate the new compensation parameter C i (t+1), apply the new compensation parameter, obtain the new deviation value, and calculate the reward R t , updated in a loop.

[0086] The deep reinforcement learning algorithm is used to dynamically optimize the compensation model parameters. The deep reinforcement learning algorithm includes three parts: state space, action space and reward mechanism.

[0087] The position deviation value of the current detection area, the process parameters collected in real time, and the historical compensation effect are taken as components of the state space;

[0088] The action space defines the set of compensation operations that can be taken. For each compensation parameter, three basic adjustment directions are defined: increase, decrease, and maintain. The adjustment step size is dynamically determined according to the size of the current deviation.

[0089] The reward function is constructed by comprehensively considering the reduction degree of position deviation, the adjustment range of compensation parameters and the stability of the system, which is expressed as:

[0090] R t =-ω1||M i (t)||2-ω2||C i (t)-C i (t-1)||2

[0091] Among them, ω1 and ω2 are weight coefficients, C i (t-1) represents the compensation model parameters at time t-1.

[0092] If the compensation effect is good and the parameter adjustment is stable, positive rewards will be given;

[0093] If overcompensation or parameter oscillation occurs, negative rewards will be given.

[0094] Building a differentiated compensation database D c , store the optimized compensation model parameters in the differential compensation database D c , Dc Contains the detection area identification code R i , compensation model parameters C i , deviation mapping relationship M i , process parameter set P t and timestamp T s The five core data fields record compensation control information of different dimensions.

[0095] Analyze historical data, understand the changing patterns of compensation parameters, evaluate the effectiveness of compensation strategies, and provide data support for process optimization.

[0096] S3: Based on the differentiated compensation database, the compensation model parameters are assigned to the corresponding servo motor control unit according to the detection area identification code, the compensation adjustment amount is calculated to generate the motion control instruction, the compensation effect evaluation index is used to evaluate the position deviation in real time, and a decision model is established to realize adaptive optimization of the compensation parameters and adaptive adjustment of regional compensation.

[0097] The differential compensation database D c The compensation model parameter C in i According to the detection area identification code R i Assigned to corresponding servo motor control units, wherein each servo motor control unit is configured with an independent parameter receiving module.

[0098] The parameter receiving module includes a parameter parsing unit and a data buffer area, which is used to parse the compensation model parameters C i Adjustment direction and step size information in;

[0099] The servo motor control unit adopts a distributed architecture to ensure that each detection area R i The compensation adjustments are independent of each other and do not interfere with each other;

[0100] Compensation model parameter C i Contains position compensation gain coefficient, speed response coefficient and acceleration limit parameters to accurately control the motion characteristics of the servo motor.

[0101] The process parameter set P t Input compensation model parameters C i , based on the compensation parameter update equation, the compensation adjustment amount is obtained, which is expressed as:

[0102] ΔC i =η·C i +τ·(M i -M0)

[0103] Where, ΔC i represents the compensation adjustment amount of the i-th detection area, η represents the compensation parameter adjustment coefficient, C irepresents the current compensation model parameters, τ represents the deviation correction coefficient, M i It indicates the position deviation value of the current detection area, and M0 indicates the target deviation threshold.

[0104] Based on the compensation adjustment ΔC i Generate motion control instructions for the servo motor; the motion control instructions include adjustment direction, number of step pulses and acceleration and deceleration timing, which are transmitted to the servo motor control unit through differential encoding, and the adjustment amount ΔC i After limiting processing, the compensation action is ensured to be smooth and controllable.

[0105] According to the motion control instructions, each servo motor is driven to perform position compensation. During the compensation process, the position sensor signal is collected to obtain the real-time position deviation value M. i , and the position deviation value M i Recorded in the differential compensation database D c In the example, the position deviation value M i Used to evaluate compensation effects and optimize compensation strategies.

[0106] Based on the differentiated compensation database D c The position deviation value M in i Establish compensation effect evaluation index E i , expressed as:

[0107] E i =λ1·|M i |+λ2·|ΔM i |+λ3·δ i

[0108] Among them, E i represents the compensation effect evaluation index of the i-th detection area, λ1, λ2, λ3 represent the evaluation weight coefficients, |ΔM i | represents the deviation change between adjacent moments, δ i Represents the standard deviation of the deviation series.

[0109] Effect evaluation index E i The accuracy, responsiveness and stability of the compensation effect are comprehensively evaluated by weighted combination. If the effect evaluation index E i Exceeding the preset threshold range [E imin ,E imax ] triggers compensation parameter optimization.

[0110] The compensation parameters are optimized by using the differential compensation database D c Extract the most recent N groups of historical data as the basis for optimization, the data includes each timestamp T s The corresponding process parameter set P t and position deviation value M i .

[0111] Based on the acquired data, a decision-making model is established using a value evaluation equation based on Q-learning. The evaluation equation guides the selection of the optimal compensation strategy by combining immediate rewards and future benefits, thereby achieving adaptive optimization of compensation parameters.

[0112] Among them, the state-action value function of the decision model reflects the expected long-term benefits of taking compensatory actions under a specific system state. The system state is represented by the position deviation M i and process parameters P t Together, they represent the operating state of the system; the compensation action corresponds to the specific compensation parameter adjustment ΔC i , directly affects the compensation effect; the immediate reward value is based on the compensation effect evaluation index E i Calculated and used to evaluate the effect of a single compensation action.

[0113] The optimized compensation model parameters C i Stored in the differentiated compensation database D c At the same time, the compensation parameters in each servo motor control unit are updated in a progressive manner to establish an adaptive optimization mechanism for compensation control.

[0114] Based on the differentiated compensation database D c The data in the printing process is continuously monitored by a high-precision photoelectric encoder, combined with the compensation effect evaluation index E i Based on the calculation results, a real-time monitoring mechanism is established.

[0115] The control instructions are updated in a progressive manner, and abnormal situations detected are handled intelligently, including:

[0116] If the detected anti-counterfeiting pattern overprint overlap is lower than the standard threshold T1, the current compensation effect evaluation index E is calculated. i , from the differential compensation database D c Extract optimized compensation parameters, and execute compensation parameter update to improve overprinting accuracy;

[0117] If the monitoring shows that the overprint deviation between multiple printing units is increasing cumulatively, the current compensation effect index E is evaluated. i The changing trend of c Select the compensation parameters with higher stability and perform smooth adjustment of the compensation parameters;

[0118] If the quality of anti-counterfeiting details is abnormal at the specified printing speed, calculate the compensation effect index E under the current working conditions i , establish the correlation model between printing speed and compensation parameters, and update the differential compensation database D c ;

[0119] If the anti-counterfeiting spot color overprinting has periodic deviation, the compensation effect evaluation index E is analyzed. i The periodic characteristics of c Extract compensation experience of similar working conditions and implement targeted compensation strategies;

[0120] If there are regional differences in the overprinting effect of the anti-counterfeiting pattern in batch production, the compensation effect evaluation index E of each region is calculated. i , establish a regional compensation parameter database and implement regional compensation adjustment.

[0121] In summary, the present invention adopts a mathematical model based on printing positioning deviation, combined with a differentiated compensation database and compensation effect evaluation index, to achieve intelligent control of the overprint accuracy of the anti-counterfeiting pattern. The multi-dimensional evaluation system based on the compensation effect evaluation index can accurately identify and timely compensate for various deviations in the printing process, effectively solving the problems of low overprint overlap of anti-counterfeiting patterns, accumulation of overprint deviations between multiple printing units, and abnormal quality of anti-counterfeiting details. Intelligent monitoring and optimization of the anti-counterfeiting printing process is achieved, which greatly improves the printing quality stability and production efficiency of anti-counterfeiting color boxes in mass production, and provides reliable technical support for the large-scale application of intelligent anti-counterfeiting color boxes.

[0122] This embodiment further provides an intelligent anti-counterfeiting color box printing positioning control system, including:

[0123] The acquisition module is used to perform regional recognition on the color box layout, divide it into multiple detection areas according to preset rules, set a combined positioning system including a cross positioning mark and a circular positioning mark in each area, collect the positioning mark information of each area in real time through a high-resolution image sensor, and establish a detection area feature database;

[0124] The compensation module is used to extract the local deformation features and global position features of each detection area based on the feature database using a multi-layer convolutional neural network to build a feature extraction model, calculate the actual deviation value of each detection area in combination with the residual network structure, establish a multi-parameter coupling mapping model of process parameters and position deviations, use a deep reinforcement learning algorithm to build a differentiated compensation model and dynamically optimize the compensation parameters to build a differentiated compensation database;

[0125] The control module is used to allocate the compensation model parameters to the corresponding servo motor control unit according to the detection area identification code based on the differentiated compensation database, calculate the compensation adjustment amount to generate motion control instructions, use the compensation effect evaluation index to perform real-time evaluation of the position deviation, establish a decision model to achieve adaptive optimization of the compensation parameters, and realize adaptive adjustment of regional compensation.

[0126] This embodiment also provides a computer device, which is suitable for the case of an intelligent anti-counterfeiting color box printing positioning control method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the intelligent anti-counterfeiting color box printing positioning control method proposed in the above embodiment.

[0127] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through Wi-Fi, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0128] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing intelligent anti-counterfeiting color box printing positioning control as proposed in the above embodiment is implemented.

[0129] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent anti-counterfeiting color box printing positioning control method, characterized in that: include: The color box layout is divided into multiple detection areas, and a combined positioning system including a cross mark and a circular mark is set up in each area. The positioning information is collected in real time through the image sensor to establish a detection area feature database; Based on the feature database, a multi-layer convolutional neural network is used to extract the local deformation and global position features of each detection area, and the actual deviation value is calculated in combination with the residual network, a mapping model between process parameters and position deviations is established, and a differentiated compensation model is constructed through deep reinforcement learning and the compensation parameters are optimized to form a compensation database; Based on the differentiated compensation database, the compensation model parameters are allocated to the corresponding servo motor control unit, the compensation adjustment amount is calculated and the control instruction is generated. The position deviation is monitored in real time through evaluation indicators to achieve adaptive optimization of compensation parameters and regional compensation adjustment.

2. The intelligent anti-counterfeiting color box printing positioning control method according to claim 1, characterized in that: The combined positioning system includes a cross mark for positioning in the X and Y directions and a circular mark for positioning in the rotation angle; Based on the divided detection area, an image sensor array is installed, and the field of view angle of the image sensor is adjusted to ensure that the field of view range of the image sensor completely covers the detection area, wherein the field of view angle is calculated by the ratio of the diagonal length value of the detection area to the focal length value of the sensor; Set the sampling frequency and calculate the image resolution by multiplying the number of sensor pixels by the actual physical size of the sampling area divided by the sensor imaging field size.

3. The intelligent anti-counterfeiting color box printing positioning control method according to claim 2, characterized in that: Based on the image sensor array, the image data of the detection mark group is collected and preprocessed, wherein the image data includes grayscale images of the cross mark and the circular mark in the detection area; The preprocessing includes image enhancement and binarization processing; Based on the binarized image data, the centroid coordinates of the detection mark group are obtained by calculating the weighted average of the coordinate value of each pixel point in the mark area and its binarized value.

4. The intelligent anti-counterfeiting color box printing positioning control method according to claim 3, characterized in that: Synchronously deploy tension sensors to collect paper tension parameters; A detection area feature database is established based on the centroid coordinate calculation result, and the feature database includes a detection area identification code, centroid coordinates, tension parameters at the acquisition time, and image enhancement parameters k1 and k2.

5. The intelligent anti-counterfeiting color box printing positioning control method according to claim 4, characterized in that: The multi-layer convolutional neural network includes three convolutional layers and two fully connected layers; Based on the three convolutional layers, local deformation features, texture features and global position features of the detection area are extracted, and dimension reduction processing is performed to obtain a feature vector of the detection area; The deviation value between the actual centroid position and the ideal centroid position of each detection area is calculated based on the feature vector, and a deviation calculation model is established using a residual network structure, wherein the X-direction deviation is obtained by subtracting the X-coordinate of the ideal standard position from the actual centroid X-coordinate, the Y-direction deviation is obtained by subtracting the Y-coordinate of the ideal standard position from the actual centroid Y-coordinate, and the angular deviation is calculated by dividing the X-direction deviation by the inverse tangent of the Y-direction deviation. The output features of the residual module are obtained by superimposing the residual mapping function and the input features.

6. The intelligent anti-counterfeiting color box printing positioning control method according to claim 5, characterized in that: Based on the feature database, tension parameters are extracted, printing speed and ambient temperature are collected in real time, and the parameters are normalized to form a process parameter set; A parameter mapping model of process parameters and position deviations is established based on a double-layer nonlinear mapping network. The process parameter set is input into the first-layer mapping network and the first-layer bias vector is superimposed. After being processed by the activation function, a deviation value vector including X-direction deviation, Y-direction deviation and angle deviation is obtained through the second-layer mapping. Finally, the second-layer bias vector is superimposed to obtain the prediction result.

7. The intelligent anti-counterfeiting color box printing positioning control method according to claim 6, characterized in that: Based on the optimized parameter mapping model, the mapping relationship between process parameters and position deviation is established. Combined with the characteristic differences in different detection areas, an independent differentiated compensation model is constructed for each area. The update of the compensation model parameters is achieved by superimposing the product of the learning rate of the current parameter value and the difference between the deviation values ​​at two adjacent moments. A differentiated compensation database is constructed, and the optimized compensation model parameters are stored in the differentiated compensation database.

8. The intelligent anti-counterfeiting color box printing positioning control method according to claim 7, characterized in that: Inputting the process parameter set into compensation model parameters to obtain compensation adjustment amount; Generate a motion control instruction for the servo motor based on the compensation adjustment amount; The motion control instructions include adjusting direction, step pulse number and acceleration / deceleration timing, and collecting position sensor signals to obtain real-time position deviation values ​​during compensation; A compensation effect evaluation index is established based on the real-time position deviation value, and the evaluation index is obtained by summing up the absolute value of the position deviation value, the absolute value of the deviation change at adjacent moments, and the standard deviation of the deviation sequence multiplied by corresponding evaluation weight coefficients.

9. The intelligent anti-counterfeiting color box printing positioning control method according to claim 8, characterized in that: Based on the effect evaluation index, the accuracy, responsiveness and stability of the compensation effect are comprehensively evaluated by weighted combination. If the effect evaluation index exceeds the preset threshold range, the compensation parameter optimization is triggered, and the optimized compensation model parameters are stored in the differentiated compensation database.

10. The intelligent anti-counterfeiting color box printing positioning control method according to claim 8, characterized in that: Based on the data in the differential compensation database, the printing process is continuously monitored by a high-precision photoelectric encoder, and a real-time monitoring mechanism is established in combination with the calculation results of the compensation effect evaluation index.

Images