Decorative paper print mark missing fault tolerance recognition system and method

Abstract

The present application relates to the technical field of printed image online detection, and discloses a system and method for fault-tolerant identification of missing printing marks on decorative paper. The method comprises: constructing a physical space-time coordinate system and obtaining an effective signal fingerprint model, using a pre-reading window and a centroid extraction calculation to obtain candidate trigger points carrying high-confidence labels; constructing a trigger deviation potential well to analyze intention energy, applying a damping regression operation to an intention state set, and performing inertia sliding compensation and weight smoothing switching to obtain a smooth trigger instruction; and based on full-image displacement vector analysis and phase reference reconstruction, a corrected trigger deviation potential well is obtained. The present application solves the problems of image random walk caused by physical mark edge jitter and system oscillation caused by missing marks, and realizes sub-microsecond trigger positioning stability and zero drift of the full-cycle monitoring picture.

Description

Technical Field

[0001] This invention relates to the field of online detection technology for printed images, and more specifically, to a system and method for identifying and correcting missing printing marks on decorative paper. Background Technology

[0002] With the increasing demands for production efficiency and finished product quality in the decorative paper printing industry, the synchronous stability of online monitoring of printed images has become a core concern. However, traditional monitoring systems, which rely on proximity switches to detect trigger signals generated by metal markers, generally face positioning accuracy bottlenecks caused by physical edge jitter. Even when the physical markers are present, the captured images still exhibit random horizontal wandering and breathing effects. Existing technologies mostly rely on a single threshold judgment of signal levels or simply switch to encoder counting mode when a marker is missing. They ignore the probability distribution distortion caused by electromagnetic interference and mechanical vibration at the signal edges under high-speed sampling, meaning that the trigger moment essentially follows a random distribution rather than a deterministic point. Furthermore, when the physical marker quality is critical or temporarily fails, there is a lack of proactive prediction and smooth transition mechanisms for phase drift trends, leading to severe logical oscillations and image displacement during fault-tolerant switching. In addition, simple feedforward control cannot detect the cumulative phase deviation caused by paper stretching or transmission slippage during long-term operation, causing the monitored image to slowly slip out of the field of view over time. Therefore, how to shift from static threshold triggering to dynamic intention energy analysis, transform discrete signal presence / absence judgment into continuous phase potential energy game, and establish a dual closed-loop correction of vision and mechanics to achieve sub-microsecond trigger positioning and zero drift monitoring under complex working conditions is a technical problem that urgently needs to be solved in this field.

[0003] In the prior art, Chinese patent application CN118438810A discloses an online printing quality inspection system and method for printing presses. This system includes a paper conveying device, an image acquisition device, an acquisition movement device, and an asynchronous linkage device. The linkage action is triggered by a print sensing device, achieving coordination between the image acquisition device and the paper conveying, enabling online inspection of print quality in high-speed printing scenarios, improving inspection efficiency and synchronous response speed. Chinese patent CN119840304A discloses an online inspection system and method based on inkjet printing, which includes a running status data acquisition unit, a deep learning prediction module, and a convolutional neural network recognition module. It predicts ink droplet states using fluid dynamics principles and generates correction parameters based on multispectral data, providing high-precision positioning for printing defect detection and achieving precise control of inkjet quality.

[0004] However, while the two patents mentioned above have some value in terms of printing inspection synergy and defect identification accuracy, they fail to address the core pain points in decorative paper printing monitoring, such as trigger jitter suppression, marker fault tolerance transition, and long-term zero drift. Specifically, the Chinese patent with publication number CN118438810A focuses on achieving synchronous acquisition through mechanical linkage, but lacks a compensation mechanism for trigger signal distortion caused by electromagnetic interference and mechanical vibration, thus failing to solve the image random walk problem. The Chinese patent with publication number CN119840304A focuses on ink droplet state prediction and defect correction, but lacks phase drift prediction and smooth switching logic for physical markers at critical or failed states, easily leading to system logic oscillations. Neither patent establishes a dual closed-loop correction architecture of vision and mechanics, making it difficult to perceive cumulative phase deviations caused by paper stretching or transmission slippage. This prevents sub-microsecond-level trigger positioning and long-term stable monitoring under complex operating conditions, failing to meet the high requirements of the decorative paper printing industry for monitoring synchronization stability. Summary of the Invention

[0005] This invention is applicable to high-speed decorative paper printing equipment such as gravure printing machines and flexographic printing machines, and can meet the high-precision monitoring requirements under strong interference environments. By constructing a physical spatiotemporal coordinate system, defining an effective signal fingerprint model, and applying a centroid extraction algorithm, the trigger positioning is improved from instantaneous edge jumps to full waveform mass center locking, eliminating image random walks caused by signal edge probability distribution distortion. By constructing a trigger deviation potential energy well to analyze the intent energy, applying damped regression to the intent state set to obtain the weighted trigger position, and performing inertial sliding compensation and weight smooth switching to generate a smooth trigger command, the system oscillation caused by abrupt logic switching is solved, ensuring the kinematic continuity of the shutter command generation logic. By obtaining the full-frame image analytical displacement vector and performing phase reference reconstruction, the corrected trigger deviation potential energy well and visual-mechanical dual closed-loop correction state are obtained, solving the long-period cumulative image shift caused by physical deformation, and ensuring that the monitored image remains absolutely stable with zero drift throughout the entire cycle.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Methods for identifying missing markings in decorative paper printing include:

[0008] The raw sensor data of the sensor in the high-speed printing scenario is acquired to construct a physical spatiotemporal coordinate system. The physical spatiotemporal coordinate system is quantized and defined to obtain an effective signal fingerprint model. Potential test signal waveform segments in the raw sensor data are screened. The test signal waveform segments are compared in all dimensions through the effective signal fingerprint model to obtain candidate trigger points. The candidate trigger points are serialized and arranged to obtain a preprocessed data stream.

[0009] A trigger deviation potential energy well is constructed based on the physical spacetime coordinate system. The preprocessed data stream is parsed to obtain the intention state set. Damped regression operation is applied to the intention state set to obtain the weighted trigger position. Inertial glide compensation and weight smooth switching are performed on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0010] The system acquires a full-frame image based on a smooth trigger command and performs absolute phase calibration and geometric deviation analysis to obtain a displacement vector. It then performs phase reference reconstruction based on the displacement vector to obtain a correction trigger deviation potential energy well that adaptively compensates for long-period cumulative phase deviation.

[0011] Furthermore, the effective signal fingerprint model includes:

[0012] The physical spacetime coordinate system is a two-dimensional reference system with absolute time as the horizontal axis (time axis) and the cumulative rotation angle of the printing roller (cumulative pulse count) as the vertical axis. The printing roller is a high-speed rotating cylinder that carries decorative textures and synchronization information in the printing production process of decorative paper. A physical mark is etched or attached to its surface, and a sensor is equipped facing the surface of the printing roller. An encoder is also equipped on the coaxial side of the printing roller.

[0013] The theoretical pulse width is determined based on the physical width of the physical marker and the resolution of the encoder;

[0014] Set an allowable deviation, and define the spatial span feature and the spatial span feature interval based on the allowable deviation and the theoretical pulse width. The spatial span feature interval includes an upper limit and a lower limit.

[0015] By setting a small observation window and a jitter threshold, a time-domain edge steepness feature is established to characterize the stability of level transitions;

[0016] The spatial span feature interval and the temporal edge steepness feature together constitute an effective signal fingerprint model.

[0017] Furthermore, the method for obtaining the candidate trigger point includes:

[0018] The raw sensing data includes the sensor's level state sequence and the encoder's pulse count value;

[0019] The level state sequence is detected, and when a rising edge transition occurs from low level to high level, a read-ahead window based on the physical space dimension is opened;

[0020] Obtain the pulse frequency output by the encoder at the current moment, and use the physical time length determined by the upper limit of the interval and the pulse frequency as the duration of the pre-read window;

[0021] During the duration of the pre-read window, the level state sequence and the encoder pulse count value are acquired to form a waveform segment of the signal under test;

[0022] The valid signal fingerprint model is used to verify the waveform segment of the signal under test to obtain the valid signal. The valid signal is then weighted and averaged to obtain the center of mass on the time axis as the centroid moment, and the centroid moment is marked as a candidate trigger point.

[0023] Furthermore, the method for acquiring the preprocessed data stream includes:

[0024] Calculate the absolute value of the spatial deviation between the encoder difference corresponding to the waveform segment of the signal under test and the theoretical pulse width, and determine the confidence level value based on the ratio of the absolute value of the spatial deviation to the allowable deviation. Here, the encoder difference refers to the difference between the encoder values ​​corresponding to the rising and falling edges of the high-level pulse.

[0025] The confidence scores are associated with and encapsulated with candidate trigger points in the form of confidence labels to obtain a trigger feature vector carrying digital quality proof.

[0026] The waveform segments of the test signal that do not meet the valid signal fingerprint model are identified as noise and removed. The candidate trigger points are then serialized and arranged according to their generation order in the physical spatiotemporal coordinate system to obtain the preprocessed data stream.

[0027] Furthermore, the intention energy includes:

[0028] The pulse count values ​​corresponding to the candidate trigger points recorded by the encoder in N consecutive historical printing cycles are retrieved. The arithmetic mean of the difference between the pulse count values ​​between two adjacent historical printing cycles is calculated as the pulse interval mean. The pulse interval mean is added to the pulse count value of the previous historical printing cycle to obtain the expected coordinate point as the theoretical predicted position. The printing cycle is the complete physical process of the printing roller completing a 360-degree mechanical rotation.

[0029] A trigger deviation axis is established with the mapping point of the theoretically predicted position on the time axis as the origin;

[0030] Define the potential energy function of the parabolic geometry on the trigger deviation axis;

[0031] The logical evaluation region is formed by the quadratic projection envelope of the potential energy function on the trigger deviation axis, and the trigger deviation potential energy well is obtained. The deviation is substituted into the potential energy function for calculation to obtain the intention energy characterizing the phase distortion weight of the physical marker. The deviation is defined as the algebraic difference between the centroid time corresponding to the physical marker and the theoretically predicted position on the time axis of the physical spatiotemporal coordinate system.

[0032] Furthermore, the method for obtaining the weighted trigger position includes:

[0033] A steady-state threshold and an escape threshold are preset. The intention energy is numerically compared with the steady-state threshold and the escape threshold to obtain the intention state set, which includes strong intention state, weak intention state and discrete state.

[0034] For weak intent states, the physical weight value and theoretical weight value are determined using intent energy, and a weighted summation operation is performed on the centroid moment of the current printing cycle and the theoretically predicted position to obtain the weighted trigger position.

[0035] Furthermore, the method for obtaining the smooth triggering instruction includes:

[0036] When the intent state set is discrete, the weighted trigger position determined by the last printing cycle in a strong intent state or a weak intent state is taken as the starting phase anchor point, and phase accumulation recursion is performed to obtain the virtual inertial trigger coordinates.

[0037] Monitor the time series variation of intention energy values ​​over M consecutive printing cycles, extract the numerical difference of intention energy between two adjacent printing cycles and divide it by the time interval between the two corresponding printing cycles to obtain the regression rate of intention energy; and dynamically generate a weighted evolution curve based on the regression rate.

[0038] Based on the weight evolution curve, the physical weight values ​​are faded in from zero to non-zero values ​​over M consecutive printing cycles. The inertial trigger coordinates and the re-extracted centroid time are used to perform weighted synthesis to obtain a smooth trigger command.

[0039] Furthermore, the method for obtaining the displacement vector includes:

[0040] A full-frame image is pre-captured using an industrial camera as a reference image, and a preset template containing geometric contours is extracted from the decorative texture of the reference image. The full-frame image is a pixel matrix composed of digital pixels.

[0041] The time axis scale and cumulative pulse count scale in the physical spatiotemporal coordinate system are mapped to the pixel matrix to establish the image coordinate space;

[0042] Calculate the absolute value of the brightness difference between each pixel and its neighboring pixels in the pixel matrix, and define the absolute value as the gray-level gradient change feature.

[0043] Using a preset template, a correlation search is performed within the full-frame image of the current frame, and the pixel region with the highest matching degree with the gray-level gradient change feature of the preset template is locked as a visual anchor point.

[0044] Obtain the geometric center pixel coordinates of the visual anchor point as the observation target.

[0045] The displacement vector is obtained by calculating the two-dimensional spatial vector difference between the observed target center and the preset ideal target center.

[0046] Furthermore, the method for obtaining the correction trigger deviation potential energy well includes:

[0047] The total length of horizontal pixels corresponding to a complete printing cycle in the image coordinate space of the pixel matrix of the full-frame image is identified, and the ratio of the total number of pulses generated by one rotation of the printing roller to the total length of the horizontal pixels is calculated to obtain the pulse equivalent.

[0048] The displacement vector is decomposed into a horizontal offset component along the horizontal axis of the image coordinate space and a vertical offset component along the vertical axis. The horizontal offset component is then multiplied by the pulse equivalent and a preset forgetting factor to obtain the incremental correction amount. The horizontal axis of the image coordinate space corresponds to the time axis of the physical spatiotemporal coordinate system, and the vertical axis of the image coordinate space corresponds to the vertical axis of the physical spatiotemporal coordinate system.

[0049] Subtract the incremental correction from the theoretical prediction position to obtain the corrected theoretical prediction position;

[0050] The geometric center of the trigger deviation potential energy well is re-anchored to the corrected theoretical prediction position to obtain the corrected trigger deviation potential energy well.

[0051] A system for identifying and correcting missing decorative paper printing marks, used to implement the aforementioned method for identifying and correcting missing decorative paper printing marks, the system comprising:

[0052] Signal purification and screening module: used to acquire raw sensor data in high-speed printing scenarios to construct a physical spatiotemporal coordinate system, perform quantization and definition on the physical spatiotemporal coordinate system to obtain an effective signal fingerprint model, screen potential test signal waveform segments in the raw sensor data, perform full-dimensional comparison of the test signal waveform segments through the effective signal fingerprint model to obtain candidate trigger points, and perform serialization and arrangement on the candidate trigger points to obtain a preprocessed data stream;

[0053] State parsing module: used to construct a trigger deviation potential energy well based on the physical spatiotemporal coordinate system, parse the preprocessed data stream to obtain the intention state set, apply damped regression operation to the intention state set to obtain the weighted trigger position, perform inertial glide compensation and weight smooth switching on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0054] Visual feedback reconstruction module: used to acquire the full-frame image based on the smooth trigger command and perform absolute phase calibration and geometric deviation analysis to obtain the displacement vector. Then, it performs phase reference reconstruction based on the displacement vector to obtain the correction trigger deviation potential energy well for adaptive compensation of long-period cumulative phase deviation.

[0055] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0056] This invention elevates trigger positioning from susceptible to interference-prone instantaneous edge jumps to full waveform mass center locking by constructing a physical spatiotemporal coordinate system and combining it with a centroid extraction algorithm. This statistically eliminates image random walk and breathing effects caused by signal edge jitter. The trigger deviation potential energy well and damped regression calculation transform traditional binary logic judgment into continuous state evolution based on intent energy. Through phase smoothing in weak intent states and weight smoothing switching in discrete states, the kinematic continuity of shutter command generation logic is effectively guaranteed, solving the problem of phase step and image displacement jump caused by mode switching. Based on the extraction of displacement vectors from the full-frame image, phase reference reconstruction is performed, realizing dynamic feedback of visual perception to theoretically predicted positions. By generating a corrected trigger deviation potential energy well, a visual-mechanical dual closed-loop correction state is constructed, solving the problem of accumulated phase deviation caused by physical deformation during long-cycle operation and ensuring that the online monitoring image remains absolutely stable with zero drift throughout the entire production cycle. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of the method for identifying and correcting missing decorative paper printing marks according to an embodiment of the present invention.

[0059] Figure 2 This is a side view structural diagram illustrating the positional relationship between a printing roller and a sensor, provided in an embodiment of the present invention.

[0060] Figure 3 A schematic diagram of the geometric topology and energy mapping of a trigger deviation potential energy well provided in an embodiment of the present invention;

[0061] Figure 4 A schematic diagram of an inertial gliding and weighted smoothing switching trajectory in a discrete state is provided for an embodiment of the present invention;

[0062] Figure 5 This is a schematic diagram of a displacement vector extraction based on image feedback provided in an embodiment of the present invention;

[0063] Figure 6 This is a functional module diagram of the decorative paper printing mark missing error recognition system provided in an embodiment of the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Example 1

[0066] Please see Figure 1 As shown, this embodiment provides a method for error-tolerant identification of missing markings in decorative paper printing, including:

[0067] Step S10: Obtain the original sensor data of the sensor in the high-speed printing scenario to construct a physical spatiotemporal coordinate system, perform quantization and definition on the physical spatiotemporal coordinate system to obtain an effective signal fingerprint model, screen potential test signal waveform segments in the original sensor data, perform full-dimensional comparison of the test signal waveform segments through the effective signal fingerprint model to obtain candidate trigger points, perform serialization and arrangement on the candidate trigger points to obtain a preprocessed data stream.

[0068] Further, step S10 includes:

[0069] Step S11: Obtain the original sensing data of the sensor in the high-speed printing scenario to construct a physical spatiotemporal coordinate system, perform quantization and definition on the physical spatiotemporal coordinate system, and obtain an effective signal fingerprint model.

[0070] In the printing process of decorative paper, the printing roller is a high-speed rotating cylinder carrying decorative textures and synchronization information. A physical mark is etched or attached to its surface; this mark is typically a geometrically regular color block or metal sheet with high reflectivity or magnetism. A sensor is fixedly mounted on the printing press frame, its sensing head facing the roller surface. When the physical mark passes through the sensor's sensing area as the roller rotates, the sensor outputs a pulse signal whose voltage level reverses. However, the printing environment is filled with strong electromagnetic interference and mechanical vibrations, causing the sensor's output voltage level to be a complex waveform mixed with numerous spikes and distortions, rather than an ideal square wave. To extract the true synchronization information from this contaminated signal, an observation mechanism capable of examining the signal simultaneously from both temporal and spatial dimensions is needed. The aim is to transform the judgment of a single moment in the signal into a matching of the spatiotemporal characteristics of the signal throughout its entire lifecycle, utilizing the inherent properties of the physical mark in the spatiotemporal domain as the basis for authenticity identification, providing a clean data source for subsequent precise triggering. See also Figure 2This is a side view schematic diagram of the positional relationship between a printing roller and a sensor according to an embodiment of the present invention. As shown in the figure, the printing roller rotates at high speed counterclockwise, and a physical marker, such as a highly reflective metal sheet, is embedded in a fixed position on its side. The sensor is fixed directly above the printing roller, and its detection beam is incident perpendicularly on the surface of the printing roller. When the rotation of the printing roller causes the physical marker to enter the coverage area of ​​the sensor's detection beam, the sensor's output circuit closes, generating a high-level signal; when the physical marker rotates out of this range, the signal returns to a low level. Simultaneously, an encoder coaxially mounted on one side of the printing roller outputs a pulse sequence corresponding to the rotation angle in real time. This mechanical structure determines a definite geometric mapping relationship between the physical width of the sensor output signal, the arc length of the physical marker, and the rotational speed of the printing roller.

[0071] Specifically, to achieve holographic capture of the sensor output signal, a pulse-time domain dual-track synchronous sampling stream is constructed. This pulse-time domain dual-track synchronous sampling stream consists of two parallel and clock-synchronized data acquisition channels. The first data acquisition channel is a time-domain channel, which uses a high-frequency clock within a field-programmable gate array (FPGA) to record the level state sequence of the sensor output port in real time. This level state sequence includes the absolute moment of each level transition and the duration of that level state, thus recording minute fluctuations in the signal along the time axis. The second channel is a spatial-domain channel, which latches the encoder's pulse count value completely synchronously with the time-domain channel. The encoder is coaxially coupled to the rotating shaft of the printing roller, converting the roller's mechanical rotational motion into digital electrical pulses. The encoder generates a fixed number of total pulses per revolution, meaning the pulse count value has a linear correspondence with the angular position of the printing roller's rotation. This ensures that each discrete pulse count value uniquely corresponds to a specific physical point on the roller's circumference, and the correspondence between the pulse count value and the physical position on the roller surface remains constant regardless of changes in the roller's rotational speed. The level state sequence is combined with the pulse count value to obtain the raw sensing data. Using a pulse time-domain dual-track synchronous sampling stream acquisition method, each sampling event is simultaneously anchored on both the time axis and spatial position coordinates, forming a physical spatiotemporal coordinate system. This physical spatiotemporal coordinate system is a two-dimensional reference system with absolute time as the horizontal axis (also defined as the time axis) and the cumulative rotation angle of the printing roller (i.e., the cumulative number of pulses) as the vertical axis. The physical spatiotemporal coordinate system can analyze not only the time dimension of the signal (e.g., how long it has been), but also the spatial dimension (e.g., on which arc length of the printing roller it occurred), thus completely eliminating the shortcomings of single-time dimension analysis affected by fluctuations in printing roller speed.

[0072] Based on the physical spatiotemporal coordinate system, an effective signal fingerprint model is defined. The effective signal fingerprint model is a multi-dimensional verification standard that maps the static geometric and optical boundary properties of a physical marker to dynamic digital features in a pulse-time domain dual-track synchronous sampling stream. The effective signal fingerprint model is used as a digital screen to distinguish between real physical markers and environmental noise signals. Specifically, the effective signal fingerprint model includes spatial span characteristics and temporal edge steepness characteristics. The spatial span characteristic is the increment of the pulse count value covered by the high-level signal output by the sensor on the vertical axis of the physical spatiotemporal coordinate system. This pulse count increment characterizes the actual geometric arc length of the physical source generating the signal in the circumferential direction of the printing roller. Since the physical marker has a definite physical width A on the printing roller surface, A is spatial entity evidence of the physical marker's existence. Based on the encoder's resolution B, the theoretical pulse width C is calculated, i.e. ,in, This indicates the diameter of the printing roller, which can be found in the parameter table provided with the equipment. The theoretical pulse width represents the encoder count that should be generated during the physical mark scanning process under ideal error-free installation and spot divergence conditions, serving as an ideal mapping reference from physical entity to digital signal. Considering the cumulative measurement errors introduced by physical factors such as the penumbra at the edge of the sensor detection spot causing a slight widening of the effective trigger width, minute angular deviations during mechanical installation leading to changes in the actual scanning chord length, and micron-level radial runout during high-speed rotation of the printing roller, an allowable deviation is set. The allowable deviation is an engineering redundancy boundary set based on the statistical distribution characteristics of physical error sources such as industrial site mechanical installation tolerances, spot edge effects, and radial runout. It aims to balance the system's adaptability to normal physical fluctuations with its ability to eliminate large-amplitude abnormal signals; for example, it is set to 10% of the theoretical pulse width. The purpose is to balance adaptability to normal physical fluctuations with its ability to eliminate large-amplitude abnormal signals. Based on the theoretical pulse width and the allowable deviation, the spatial span characteristic range is determined. ,in, The lower bound of the interval represents the minimum spatial scale threshold required to be recognized as a real physical marker. ; The upper bound of the interval represents the maximum spatial scale threshold allowed for a marker to be recognized as a real physical landmark. Simultaneously, the encoder values ​​corresponding to the rising and falling edges of the high-level pulse are acquired in real time, and the difference between the encoder values ​​corresponding to the rising and falling edges is calculated and defined as the encoder difference. The encoder difference is compared with the spatial span characteristic interval. If the encoder difference is within the closed interval defined by the spatial span characteristic interval, that is, the encoder difference is greater than or equal to the lower limit of the interval and less than or equal to the upper limit of the interval, it is determined that the spatial span characteristic is satisfied. If the encoder difference is less than the lower limit of the interval or greater than the upper limit of the interval, it is determined that the spatial span characteristic is not satisfied. The spatial span characteristic interval is based on a bandpass filter constructed according to the physical spatial geometric scale, which can quickly and definitively truncate all pseudo signals that do not conform to the physical marking geometric size characteristics based solely on the spatial length of the signal without the need for complex waveform analysis.

[0073] The temporal edge steepness characteristic is an index of the stability of the sensor output signal's level state within a tiny time window at the instant of level reversal. It characterizes the optical sharpness of the edge contour of the physical source generating the signal and the electromagnetic purity of the signal generation process. The physical edge of the physical marker is precisely machined, exhibiting sharp geometric abrupt changes. When the edge of the physical marker rapidly sweeps across the sensor's detection beam, the photoelectric conversion circuit inside the sensor should produce a monotonous and rapid level reversal. Conversely, electromagnetic interference signals originate from the oscillation of induced electromotive force within the circuit, manifesting as high-frequency, repetitive level jumps. To quantitatively capture this essential difference at the microscopic time scale and identify the true physical edges from complex signal waveforms, a time limit is defined to focus on analyzing the dynamic behavior of the levels. Therefore, a small observation window is set, which is based on covering the typical electromagnetic interference pulse width without including too much irrelevant data. For example, it is set to 5 microseconds. To quantitatively distinguish between the physical characteristics of monotonic flips and the noise characteristics of reciprocating oscillations, a jitter threshold is set, for example, 3 times. The time-domain edge steepness characteristic requires that within the small observation window triggered by the rising or falling edge of the signal, if the number of level state flips is less than or equal to the jitter threshold, then the time-domain edge steepness characteristic is satisfied, indicating that the edge is a monotonic physical edge; otherwise, the time-domain edge steepness characteristic is not satisfied, indicating that the edge is oscillating noise. The logic for determining the temporal edge steepness feature lies in the essential difference between the irreversible monotonicity of the physical event of a physical marker entering or leaving the sensor detection beam on the time scale and the random oscillation of electromagnetic noise signals on the same time scale. This eliminates complex noise that disguises itself as a physical marker in terms of spatial span but whose edge characteristics do not conform to physical laws. Through the joint constraint of spatial span features and temporal edge steepness features, the effective signal fingerprint model achieves the reverse reconstruction of the physical essence of the signal source, ensuring the purity and reliability of subsequent input data.

[0074] Step S12: Dynamically filter out potential signal waveform segments in the original sensor data using the pre-read window, compare the signal waveform segments with the effective signal fingerprint model in all dimensions, and obtain candidate trigger points carrying confidence labels.

[0075] The system continuously monitors raw sensor data. When the sensor's level changes from low to high (rising edge), a pre-read window is triggered. This pre-read window is a dynamic capture mechanism based on physical space. It obtains the current printing roller linear speed by reading the pulse frequency output by the encoder in real time. The upper limit of the spatial span characteristic interval is divided by the current printing roller linear speed to obtain the physical time length corresponding to the pre-read window. This physical time length is used as the duration of the pre-read window. During the duration of the pre-read window, all level state sequences and corresponding encoder pulse count values ​​are continuously acquired to form a complete waveform segment of the signal under test. The waveform segment of the signal under test is a time slice set containing dual data streams, including the level state sequence output by the sensor during the duration of the pre-read window and the encoder pulse count value corresponding to each level state change moment in the level state sequence. The read-ahead window allows the signal to fully exhibit its steady-state characteristics in the time and spatial domains within the waveform segment of the signal under test, avoiding single-point false triggering caused by transient jitter at the leading edge of the signal or high-frequency glitches. This ensures that the signal discrimination logic is based on the full-cycle waveform characteristics, rather than relying on local instantaneous level jumps.

[0076] The waveform segment of the signal under test is compared with the fingerprint model of the effective signal in all dimensions. Specifically, the spatial span feature is verified by extracting the encoder difference corresponding to the high-level state maintenance period in the waveform segment of the signal under test. It is determined whether the encoder difference falls within the spatial span feature range. If it does, it indicates that the high-level pulse in the waveform segment of the signal under test highly matches the actual physical mark in terms of spatial size, and has the geometric basis to become an effective signal. The spatial span feature verification is thus satisfied. After the spatial span feature verification is satisfied, the confidence value is quantized. Specifically, the absolute difference between the encoder difference corresponding to the waveform segment of the signal under test and the theoretical pulse width is calculated to obtain the absolute value of the spatial deviation. Dividing by the allowable deviation yields the deviation ratio. Subtracting the deviation ratio from 1 gives the confidence score. The confidence score is calculated using a linear decay model describing the physical signal fidelity: when the encoder difference perfectly matches the theoretical pulse width, the deviation ratio is zero, and the confidence score is the maximum value of 1, indicating perfect physical reliability. As the absolute value of the spatial deviation approaches the allowable deviation boundary, the confidence score decreases linearly, indicating increased quality degradation due to mechanical vibration or optical scattering. When the absolute value of the spatial deviation reaches the allowable deviation, the confidence score is 0. If the encoder difference does not fall within the spatial span characteristic range, the waveform segment of the test signal is determined to be noise and does not meet the spatial span characteristic verification. For waveform segments of the test signal that have passed the spatial span characteristic verification, the temporal edge steepness characteristic is verified by backtracking analysis of the number of level state transitions within the small observation window near the rising and falling edges of the waveform segment. If the number of occurrences is less than or equal to the jitter threshold, it indicates that the edge of the high-level pulse is clear and crisp, conforming to the optical characteristics of physical markings, and is therefore deemed to satisfy the time-domain edge steepness characteristic verification. Conversely, if the number of occurrences is greater than the jitter threshold, it indicates that the edge of the high-level pulse is accompanied by high-frequency oscillations, which may be a pulse disguised as electromagnetic interference, and is therefore deemed not to satisfy the time-domain edge steepness characteristic verification. A waveform segment of the signal under test that simultaneously satisfies both the time-domain edge steepness characteristic verification and the spatial span characteristic verification is defined as a valid signal.

[0077] For valid signals, the rising edge time is no longer simply taken as the trigger point. This is because, under high-speed printing, even for real physical markers, the rising edge time is subject to random drift due to spot edge effects, sampling quantization errors, or minor mechanical jitter. To eliminate jitter, a centroid extraction algorithm is introduced. Specifically, a time-domain weighted average is calculated for the sequence of level states contained in the valid signal. By accumulating the product of each discrete sampling time and its corresponding level state within the pre-read window, and dividing the resulting sum by the sum of all level states, the centroid time is obtained as the center of mass of the valid signal on the time axis of the physical spatiotemporal coordinate system. The centroid time, in its physical essence, is the time dimension coordinate of the valid signal in the physical spatiotemporal coordinate system, representing the geometric energy center when the physical marker sweeps across the sensor detection area. Compared to the rising edge time, which is susceptible to fluctuations due to single-point noise interference, the centroid time, by integrating the full waveform information of the level state sequence contained in the valid signal, eliminates the signal edge probability distribution distortion caused by electromagnetic jitter. The centroid moment extracted in each printing cycle is marked as a candidate trigger point, and a confidence value is bound to the candidate trigger point. The candidate trigger point is presented in the form of a confidence label. That is, the candidate trigger point is a binary data structure containing the centroid moment and the confidence value. The printing cycle is the complete physical process of the printing roller completing a 360-degree mechanical rotation, which corresponds to the count range of the total number of pulses generated by the printing roller rotating one revolution as described in step S11, from zero to zero.

[0078] Step S13: Perform serialization and arrangement on the candidate trigger points to obtain a preprocessed data stream that eliminates physical background noise interference.

[0079] Obtain candidate trigger points for each printing cycle and retrieve the confidence labels bound to the candidate trigger points. Extract the confidence values ​​contained in the confidence labels, and construct a trigger feature vector carrying digital quality proof by logically binding the candidate trigger points and confidence values ​​in terms of time and space. Retrieve the verification results of the full-dimensional comparison of the waveform segments of the signal under test in step S12. For waveform segments of the signal under test that are judged as noise, do not meet the spatial span feature verification, or do not meet the temporal edge steepness feature verification, they are defined as unstructured noise. Perform logical masking and data discarding on unstructured noise. Specifically, for waveform segments of the signal under test that are not defined as valid signals, refuse to merge their contained level state sequences and encoder pulse count values ​​into the preprocessed data stream to be generated, and trigger a memory reset operation to release the storage space previously occupied by the waveform segment of the signal under test in the preread window, thereby realizing the logical blocking and physical elimination of unstructured noise at the data link layer. The preprocessed data stream is obtained by sequentially arranging the candidate trigger points, each carrying a confidence score, according to their generation order on the time axis of the physical spatiotemporal coordinate system for each printing cycle. The preprocessed data stream is a structured data communication protocol packet containing the centroid time, confidence score, and corresponding encoder pulse count for each candidate trigger point in each printing cycle. This preprocessed data stream transforms the raw, discrete sensor data stream with physical background noise into a standardized feature dataset with deterministic physical meaning, thereby providing a unified logical interface for subsequent logic at the data level.

[0080] Step S10 addresses the technical challenges of signal authenticity in high-speed decorative paper printing environments, including the difficulty in distinguishing genuine signals from fakes due to strong electromagnetic interference and high-frequency mechanical vibrations, random phase jitter at trigger times, and high false trigger rates in traditional single-threshold detection under non-ideal waveforms. This is achieved through constructing a physical spatiotemporal coordinate system, defining an effective signal fingerprint model, full-dimensional comparison, and preprocessing the data stream. It realizes accurate restoration of the physical essence of the physical markers, stable trigger positioning, and standardized refinement of features at the data link level. Specifically, the physical spatiotemporal coordinate system provides a spatial and temporal scale mapping with deterministic physical meaning for signal determination, solving the problem of inconsistent judgment benchmarks caused by fluctuations in the printing roller speed in a single time dimension. The effective signal fingerprint model elevates the trigger position from easily disturbed instantaneous edge jumps to full-waveform quality center locking, resulting in a significant technical effect of eliminating random horizontal walks in the image caused by signal edge probability distribution distortion. The preprocessing data stream achieves a dimensional transformation from discrete, chaotic physical sampling to high-purity feature vectors with prior quality information, greatly reducing the real-time computational load and decision lag risk during subsequent logic execution.

[0081] Step S20: Construct a trigger deviation potential energy well based on the physical spatiotemporal coordinate system, parse the preprocessed data stream to obtain the intention state set, apply damped regression operation to the intention state set to obtain the weighted trigger position, perform inertial glide compensation and weight smoothing switching on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0082] Further, step S20 includes:

[0083] Step S21: Construct a trigger bias potential energy well based on the physical spatiotemporal coordinate system, parse the preprocessed data stream, and obtain the intention energy characterizing the phase distortion weight of the physical marker.

[0084] In the dynamic monitoring scenario of high-speed operation of decorative paper printing equipment, the preprocessed data stream provides highly pure physical feature inputs. To achieve synchronization stability, the phase deviation weights of candidate trigger points relative to the steady-state rotation of the printing roller are further explored. Since the theoretically predicted position determined by the pulse count value output by the encoder logically represents the ideal mechanical synchronization benchmark, while the candidate trigger points in the preprocessed data stream represent the physically measured positions affected by microscopic disturbances in the industrial environment, the real-time deviation between the two exhibits a statistically significant random distribution. Therefore, a geometric evaluation framework for quantifying the intensity of synchronization intent is constructed. By establishing a conversion mechanism that maps the displacement deviation of candidate trigger points to physical field energy states, continuous real-time grading of the strength of synchronization intent is achieved, providing core physical evaluation parameters for subsequent processing.

[0085] Specifically, to accurately quantify the offset of candidate trigger points relative to theoretically predicted positions, a trigger deviation potential energy well is constructed based on the physical spatiotemporal coordinate system. The theoretically predicted position is determined by retrieving the encoder pulse count values ​​corresponding to candidate trigger points recorded by the encoder in N consecutive historical printing cycles, calculating the average pulse interval between candidate trigger points in adjacent printing cycles, and adding this average pulse interval to the encoder pulse count value corresponding to the candidate trigger point in the previous printing cycle. This yields the expected coordinate point where the physical mark should appear in the current printing cycle, and this expected coordinate point is defined as the theoretically predicted position. The number of historical printing cycles, N, is set based on the statistical stability characteristics of the encoder output pulse frequency and the physical inertia of the printing roller rotation. The aim is to establish a statistical balance between smoothing instantaneous sampling noise caused by electromagnetic interference and maintaining real-time tracking capability of dynamic changes in the printing roller rotation speed; for example, it is set to 20. Using the mapping point of the theoretically predicted position on the time axis of the physical spatiotemporal coordinate system as the origin, a one-dimensional linear coordinate axis, i.e., the trigger deviation axis, is established, extending synchronously along the positive and negative directions of the horizontal axis of the physical spatiotemporal coordinate system, i.e., the time axis. The setting of the trigger deviation axis is based on the construction of a local deviation observation area with the theoretical prediction position as the core, so that any candidate trigger point falling into the physical spatiotemporal coordinate system can directly characterize the physical magnitude of the synchronous phase distortion through the displacement vector relative to the theoretical prediction position, thereby simplifying the absolute coordinate retrieval in the global range to the relative phase offset mapping for the ideal equilibrium point.

[0086] A potential energy function is defined on the trigger deviation axis to describe the distribution law of synchronization intention. The synchronization intention is the degree of phase fit between the centroid moment corresponding to the physical marker and the mechanical rotation sequence of the printing roller, characterizing the probabilistic tendency to maintain the stability of the trigger moment under the current printing conditions. The potential energy function adopts a parabolic geometry configuration. Its construction is based on introducing an elastic constraint feedback mechanism from physics, using a quadratic function to nonlinearly amplify the deviation, thereby simulating the physical characteristic that the restoring force of a mechanical system increases dramatically with distance when deviating from equilibrium. The formula for constructing the potential energy function is as follows: Where E represents the intention energy, used to quantify the degree of distortion of the current physical marker relative to the synchronous ideal state from the perspective of energy state; the synchronous ideal state is the physical state in which the centroid moment corresponding to the physical marker and the theoretically predicted position completely coincide on the time axis of the physical spacetime coordinate system; The deviation is represented by the algebraic difference between the centroid time of the candidate trigger point in the current printing cycle and the time coordinate of the theoretically predicted position in the preprocessed data stream. k represents the stiffness coefficient, which is dynamically adjusted in a positive correlation with the roller linear speed and confidence level. The purpose is to adaptively configure the penalty for synchronization phase deviation by changing the geometric steepness of the trigger deviation potential energy well; for example, it is set to 500. The trigger deviation potential energy well is a logical evaluation region within the physical spatiotemporal coordinate system, with the trigger deviation axis as the horizontal base and the intended energy as the vertical index. The geometric range of the trigger deviation potential energy well is enveloped by the quadratic projection of the potential energy function onto the trigger deviation axis. In the spatial framework of the physical spatiotemporal coordinate system, the theoretically predicted position corresponds to the geometric center and the lowest energy point of the trigger deviation potential energy well along the time axis; the trigger deviation axis defines the phase deviation span for which the candidate trigger point is included in the synchronous evaluation. The boundary of the trigger deviation potential energy well is formed by a parabolic wall determined by a potential energy function. This parabolic wall is tangent to the trigger deviation axis at the theoretically predicted location and extends towards higher energy states as the absolute value of the deviation increases. By constructing this trigger deviation potential energy well, the phase deviation of the physical marker relative to the ideal synchronization state is mapped to a potential energy height relative to the equilibrium position at the bottom of the well within the evaluation region, thereby achieving a geometrical representation of the synchronization intention in a two-dimensional spatiotemporal space. See also... Figure 3 This is a schematic diagram of the geometric topology and energy mapping of a triggered deviation potential energy well provided in an embodiment of the present invention. Figure 3As shown, this is established within the framework of a physical spacetime coordinate system. The bottom horizontal ray in the figure represents the horizontal axis of the physical spacetime coordinate system, and the left vertical ray represents the vertical axis. The thickened colored solid lines in the figure, such as the orange solid line, are defined as the trigger deviation axis, located above and parallel to the time axis, used to limit the activity space of candidate trigger points within the local phase. The green solid dots in the figure represent the theoretical prediction positions, which serve as the logical zeros of the trigger deviation axis, and the geometric center of the trigger deviation axis is anchored at the theoretical prediction positions. The parabolic geometry depicted by the blue solid line in the figure is the geometric envelope of the potential energy function. The closed semi-open region enclosed by the parabolic geometry and the trigger deviation axis below it is the logical evaluation region, which is physically equivalent to the trigger deviation potential energy well. In the diagram, the red solid dots on the parabolic geometry represent candidate trigger points for the current printing cycle. The distance between the vertical projection of the red solid dots onto the trigger deviation axis and the theoretically predicted position is labeled as the deviation amount, and the corresponding longitudinal energy height is labeled as the intended energy. Furthermore, the diagram uses two sets of parabolic dashed lines to illustrate the dynamic evolution of the stiffness coefficient k: steeper dashed lines with smaller radii of curvature indicate an increase in k, representing an enhancement of the synchronization constraint; while gentler dashed lines with larger radii of curvature indicate a decrease in k, representing an expansion of the fault tolerance buffer space. Through the trigger deviation potential energy well, the phase deviation of the physical marker relative to the ideal synchronization state is mapped to the energy state height within the logic space.

[0087] Step S22: Analyze the intent energy to obtain the intent state set, and apply damped regression operation to the intent state set to obtain the weighted trigger position of the image shutter opening command.

[0088] The resolved intent energy is used to perform a graded assessment of the reliability of physical markers within the current printing cycle. In the high-speed printing process of decorative paper, due to the dynamic changes in the physical characteristics of the physical markers and the detection environment, a decision-making mechanism capable of automatically adjusting trust weights based on real-time dynamic feedback is required. By establishing a non-binary control state domain, a refined characterization of the synchronization signal quality is achieved.

[0089] Specifically, two energy thresholds are preset: a steady-state threshold and an escape threshold. The steady-state threshold is set based on the minimum phase deviation tolerance required for image stabilization, representing the energy boundary where the physical marker and the theoretically predicted position highly coincide. The steady-state threshold is determined by performing statistical variance analysis on the deviation recorded by the encoder in historical printing cycles, using the fluctuation index of the deviation and superimposing a preset confidence gain. The deviation is defined as the algebraic difference between the centroid time corresponding to the physical marker and the theoretically predicted position on the time axis of the physical spatiotemporal coordinate system, reflecting the instantaneous phase drift of the physical signal relative to the mechanical logic reference. The preset confidence gain is based on the probability distribution envelope of the deviation in historical statistics, introducing a standard normal distribution confidence interval logic to set the value of the confidence gain. For example, the confidence gain is set to 3 to quantify the background noise reference of the mechanical transmission system, thereby defining the upper limit of the energy of the physical marker in an absolutely synchronous steady state. The escape threshold is determined based on the spatial overlap limit between the physical aperture of the sensor detection beam and the physical mark in the circumferential direction of the printing roller. This is achieved by converting the spatial overlap limit into the maximum allowable phase deviation time in the physical spatiotemporal coordinate system and substituting it into the potential energy function. This aims to define the physical boundary line where the physical mark signal transitions from valid to invalid. The preset energy critical threshold is based on constructing a three-stage control evaluation model with transition buffer characteristics. Setting the steady-state threshold ensures that the centroid time corresponding to the candidate trigger point can be locked when the phase deviation is extremely small, avoiding unnecessary computational intervention that could lead to phase response delay. Setting the escape threshold prevents false phase information from the physical mark due to severe contamination or misalignment. The intended energy of the current printing cycle is compared numerically with the steady-state threshold and the escape threshold, classifying it into strong intention state, weak intention state, and discrete state. When the intended energy is less than the steady-state threshold, the current physical mark is determined to be in a strong intention state. In a strong intent state, the region in the trigger deviation potential energy well where the intent energy value is less than the steady-state threshold is defined as the bottom steady-state region. The centroid time of the candidate trigger point falls into the bottom steady-state region, indicating that the position of the physical marker in the physical spatiotemporal coordinate system is highly consistent with the theoretically predicted position. A strong intent control strategy is implemented: the centroid time corresponding to the candidate trigger point in the current printing cycle in the preprocessed data stream is directly locked as the final weighted trigger position, and this centroid time is used to perform zero-deviation correction on the theoretically predicted position for the next printing cycle. Specifically, the centroid time of the current printing cycle is used as the absolute phase reference. By accumulating the total number of pulses generated by one rotation of the printing roller, the theoretically predicted position for the next printing cycle is recalculated and overwritten, thereby achieving real-time phase locking between the control reference and the high-quality physical marker. When the intent energy is greater than or equal to the steady-state threshold and less than the escape threshold, the current physical marker is determined to be in a weak intent state.In the weak intent state, the region in the trigger deviation potential energy well where the intent energy value is greater than or equal to the steady-state threshold but less than the escape threshold is defined as the wellbore slope region. The centroid of the candidate trigger point always falls into the wellbore slope region, indicating that the physical marker has phase jitter. A weak intent control strategy is executed, i.e., the weighted trigger position is calculated using the damped regression method. When the intent energy is greater than or equal to the escape threshold, or when the pre-read window does not collect a signal segment that satisfies the valid signal fingerprint model within the current printing cycle, the current state is determined to be a discrete state. In the discrete state, the physical marker is determined to be invalid within the logical evaluation region. A discrete state marking operation is performed, and the damped regression method is prohibited from intervening in the current cycle. The strong intent state, weak intent state, and discrete state are combined into an intent state set.

[0090] The weak intent control strategy employs a damped regression method for phase smoothing, constructing a weighted allocation model based on intent energy. Physical and theoretical weight values ​​are obtained through numerical calculations. Specifically, the quotient of the intent energy in the current printing cycle and the escape threshold is calculated to obtain an energy proportion coefficient. This energy proportion coefficient is used as the theoretical weight value, and the physical weight value is obtained by subtracting the energy proportion coefficient from the value of 1. The construction logic of the damped regression method is as follows: using intent energy as a weight adjustment parameter characterizing damping properties, the weighted trigger position is obtained by finding a dynamic equilibrium point between the measured phase point based on the centroid time and the inertial reference point based on the theoretically predicted position within the physical spatiotemporal coordinate system. The specific calculation method is as follows: multiplying the centroid time of the current printing cycle by the physical weight value yields a physical weighted term; multiplying the theoretically predicted position by the theoretical weight value yields a theoretical weighted term; and summing the physical and theoretical weighted terms yields the final weighted trigger position. The weighted trigger position is defined as the coordinates of the execution command sent to the industrial camera to open the image shutter. The industrial camera is a terminal aperture sensing component that maps decorative textures rotating with the printing roller within the physical spatiotemporal coordinate system to physical marks using a digital pixel matrix. By executing an exposure action based on a weighted trigger position, it freezes the physical marks and surrounding printed textures in a high-speed rotating state into a static image with spatiotemporal anchoring attributes, thereby providing raw visual data to capture the microscopic phase drift of the mechanical transmission system from a visual perspective. The role of the weighted trigger position is to use mathematical fusion to retain the real-time phase feedback characteristics provided by the physical marks while utilizing the kinematic stability of the theoretically predicted position to offset the random phase fluctuations remaining during the extraction of the physical marks at the centroid moment. The purpose of obtaining the weighted trigger position is to force the trigger reference of the industrial camera to approach the ideal theoretical synchronization trajectory when the physical marks are in a weak intention state, thereby ensuring that the exposure moment of the industrial camera maintains the consistency of its spatial position within the physical spatiotemporal coordinate system and effectively suppresses the random walk phenomenon of the image in the horizontal direction. Through the output of the weighted trigger position, depth filtering of high-frequency micro-jitter is achieved without losing physical mark calibration information, providing precise control data for obtaining sub-pixel level still images.

[0091] Step S23: Based on the weighted trigger position, perform inertial glide compensation and weight smooth switching on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0092] When the intent state set is discrete, inertial gliding compensation based on trend prediction is initiated. The aim is to maintain the determinism of the image acquisition phase using the kinematic stability of the mechanical system during the vacuum period when physical markers are missing. In the discrete state, the centroid time is no longer retrieved; instead, phase accumulation recursion is performed via the encoder. Specifically, the weighted trigger position is determined as the starting phase anchor point from the last printing cycle before the current printing cycle that was in a strong or weak intent state. The pulse frequency output by the encoder at the current moment is obtained, and a fixed number of total pulses are generated by one rotation of the printing roller. A numerical accumulation operation is performed based on the starting phase anchor point to generate a virtual coordinate point in the physical spatiotemporal coordinate system that simulates the trajectory of the physical marker, defined as the inertial trigger coordinate. The inertial trigger coordinate addresses the technical challenge of the system losing its spatial alignment reference due to the temporary disappearance of physical markers. While performing inertial gliding compensation, raw sensor data is continuously retrieved through a pre-read window. When step S12 extracts a valid signal that satisfies the valid signal fingerprint model again, and the intention energy of the current printing cycle calculated in step S21 shows a decreasing trend towards the escape threshold, a weight smoothing switch from the discrete state to the physical feedback state is performed. By monitoring the time-series change pattern of the intention energy value within M consecutive printing cycles, the regression rate of the intention energy is calculated. Specifically, the difference between the intention energy of the next cycle and the previous cycle is extracted during two adjacent printing cycles, and this difference is divided by the time interval between the two adjacent printing cycles to obtain the regression rate reflecting the rate of decrease of the intention energy. The number of printing cycles M is set according to the encoder resolution to ensure that the regression process covers the frequency cycle of the micro-vibrations generated by the mechanical system; for example, it is set to 5. The regression rate is used to dynamically generate a weight evolution curve, which is a nonlinear function that maps the regression rate of the intention energy to the weight distribution ratio. Based on the weight evolution curve, the physical weight value and the theoretical weight value are determined, causing the physical weight value to fade in from zero to a non-zero value within the M consecutive printing cycles.

[0093] To inject stability constraints based on mechanical inertia into the control commands, the inertial trigger coordinates of the current printing cycle are multiplied by the theoretical weight value to obtain an inertial compensation component. This inertial compensation component refers to the predicted phase component driven by encoder pulse logic within the physical spacetime coordinate system, aiming to provide a timing reference for the trigger logic that is independent of environmental noise fluctuations. Simultaneously, to introduce the absolute phase calibration capability provided by physical markers in real time, the newly extracted centroid time of the current printing cycle is multiplied by the physical weight value to obtain a measured feedback component. This measured feedback component refers to the position component transformed from the physical marker's projection into the real space of the current printing cycle, aiming to enable the trigger commands to dynamically track the phase changes of the actual printed pattern. The inertial compensation component and the measured feedback component are summed to obtain the final smooth trigger command. This smooth trigger command is defined as the final position command sent to the industrial camera for image capture. The aim is to mathematically constrain the slope of the trigger point's movement within the physical spacetime coordinate system at the instant of physical marker recovery. This avoids the phase step between the minute accumulated deviation caused by inertial recursion and the physically measured phase, ensuring the continuity of the trajectory during exposure of the industrial camera across different control modes and eliminating sudden image displacement jumps caused by hard mode switching. See also Figure 4 This is a schematic diagram of an inertial gliding and weighted smoothing switching trajectory in a discrete state, provided by an embodiment of the present invention. Figure 4 As shown, the entire system is built within a physical spacetime coordinate system. The trajectory line depicted by the solid green line is defined as the measured phase trajectory, representing the actual motion position of the physical marker before the discrete state occurs. The trajectory line depicted by the dashed blue line is defined as the inertial glide trajectory, which is formed by connecting the inertial trigger coordinates of each printing cycle, and the starting point of the inertial glide trajectory is anchored at the initial phase anchor point. The area enclosed by the gray dashed circle in the figure is marked as the physical marker missing area, used to show the break in the measured data caused by the failure of the physical marker. The curved trajectory depicted by the solid red line is defined as the shutter execution trajectory, which presents a smooth switching process constrained by the weighted evolution curve. The solid red dots distributed on the shutter execution trajectory in the figure represent the finally generated smooth trigger command. The solid green dot at the end of the shutter execution trajectory in the figure marks the centroid moment of the re-extracted physical marker. The area filled with light red shading in the figure is defined as the weighted smoothing switching area, which is used to visually demonstrate the process by which the system weights and sums the inertial compensation component and the measured feedback component, so that the shutter execution trajectory smoothly approaches the recaptured physical marker from the inertial sliding trajectory.

[0094] Step S20, through preprocessing the data stream, intent state set, inertial glide compensation, weighted smooth switching, and smooth triggering instructions, solves the technical challenges of image random walk caused by physical marker edge jitter in high-speed printing of decorative paper, and the discontinuous switching of triggering logic at critical quality moments or when physical markers fail, leading to system oscillations. It achieves a dimensional transformation of triggering decisions from binary logic judgment to continuous energy state evolution, ensuring the kinematic continuity and high phase consistency of shutter instruction generation logic. Specifically, parsing the preprocessed data stream enables geometric quantization of the strength of the synchronization intent; constructing the intent state set and performing hierarchical arbitration, through quantized limits of steady-state and escape thresholds, solves the interference of high-frequency random phase fluctuations on image positioning determinism; inertial glide compensation and weighted smooth switching achieve a smooth transition from discrete states to physical feedback states, eliminating phase jumps and image displacement abrupt changes during fault-tolerant mode switching.

[0095] Step S30: Acquire the full-frame image acquired based on the smooth trigger command and perform absolute phase calibration and geometric deviation analysis to obtain the displacement vector. Perform phase reference reconstruction based on the displacement vector to obtain the correction trigger deviation potential energy well for adaptive compensation of long-period cumulative phase deviation.

[0096] Further, step S30 includes:

[0097] Step S31: Acquire the full-frame image acquired based on the smooth trigger command, perform absolute phase calibration and geometric deviation analysis, and obtain the displacement vector representing the cumulative phase drift.

[0098] An industrial camera receives a smooth trigger command and opens the image shutter at the trigger time determined by the smooth trigger command on the time axis of the physical spatiotemporal coordinate system to perform an exposure operation, acquiring a full-frame image. This full-frame image is a digital pixel matrix captured by the industrial camera in a single exposure cycle, encompassing all decorative textures and physical markings carried by the printing roller throughout the printing cycle. Under the long-cycle operation conditions of decorative paper printing, due to the slight tensile elastic deformation caused by the physical tension of the decorative paper during high-speed transport in the printing equipment, slight slippage of the transmission mechanism, and thermal expansion and contraction of the mechanical transmission chain caused by changes in ambient temperature, the theoretically predicted position based on the smooth trigger command and the actual spatial phase of the decorative texture undergo a non-linear, slow decoupling. Visually, this manifests as follows: although the smooth trigger command eliminates high-frequency jitter through weighted trigger position, the observed full-frame image will slowly shift in one direction over time within the display field of view due to the cumulative drift of the phase reference at the mechanical level. The aim is to capture and quantify this low-frequency, cumulative mechanical phase error by introducing absolute position information feedback at the image level, providing a precise basis for subsequent visual compensation.

[0099] Specifically, the time axis scale and cumulative pulse count scale in the physical spatiotemporal coordinate system are mapped to the pixel matrix of the full-frame image to establish an image coordinate space. This image coordinate space is a two-dimensional discrete reference system in pixels, where the horizontal axis corresponds to the time axis of the physical spatiotemporal coordinate system, and the vertical axis corresponds to the vertical axis. A coordinate reference point representing the ideal synchronization state is preset within the image coordinate space, defined as the ideal bullseye. This ideal bullseye is represented by a set of fixed coordinate values, defined as static pixel coordinate points. Specifically, during the initialization phase of the printing operation, a reference full-frame image is pre-acquired using an industrial camera. From the decorative texture contained in the reference full-frame image, a slice containing a significant geometric contour is manually or automatically extracted and stored as a preset template. During the printing operation, the absolute value of the brightness difference between each pixel in the pixel matrix of the current full-frame image and its adjacent pixels is calculated. This absolute value of the brightness difference is defined as a grayscale gradient change feature, used to highlight the edge structure in the decorative texture. Using the preset template, a correlation search is performed within the full-frame image of the current frame. The pixel region with the highest matching degree of grayscale gradient change features with the preset template is locked as the visual anchor point. The geometric center pixel coordinates of the visual anchor point are obtained by calculating the arithmetic mean of the pixel region covered by the visual anchor point on the horizontal and vertical axes, and these coordinates are defined as the observation bullseye. The displacement vector is obtained by calculating the two-dimensional spatial vector difference between the observation bullseye and the ideal bullseye in the image coordinate space. To quantify the specific impact of phase shift in different dimensions of the physical spatiotemporal coordinate system, the displacement vector is decomposed into a horizontal offset component along the horizontal axis and a vertical offset component along the vertical axis. The displacement vector is a vector index characterizing the degree of deviation of the entire image from the ideal bullseye. The magnitude and polarity of the displacement vector precisely characterize the cumulative offset between the smooth trigger command of the current cycle and the true phase of the physical marker. Through the extraction of the displacement vector, the pattern shift features at the visual level are transformed into an error signal that can be used to correct the mechanical level synchronization reference, establishing an error feedback path from image detection results back to front-end control parameters. See also Figure 5 This is a schematic diagram of a displacement vector extraction based on image feedback provided in an embodiment of the present invention. Figure 5The diagram shows a full-frame image captured by the industrial camera. The thick solid rectangle in the diagram represents the pixel matrix boundary of the current full-frame image. The image coordinate space is represented by two rays intersecting perpendicularly at the lower left corner of the pixel matrix. The intersection of the central dashed crosshairs represents the preset ideal bullseye, whose position in the image coordinate space is the static pixel coordinate point. The closed area filled with blue shading represents the visual anchor point locked by the system through grayscale gradient changes. The red solid dot represents the geometric center pixel coordinates of the visual anchor point, i.e., the observation bullseye. The thick solid red line segment with arrows pointing from the ideal bullseye to the observation bullseye indicates the displacement vector. Vertical dashed lines extending from the observation bullseye to the horizontal and vertical axes respectively indicate the horizontal offset component of the displacement vector along the horizontal axis and the vertical offset component along the vertical axis. Through the shown geometric mapping relationship, the visual deviation after the smooth trigger command is perceived in real time, providing error data for closed-loop control to subsequently correct the theoretically predicted position.

[0100] Step S32: Perform phase reference reconstruction based on displacement vector to obtain a corrected trigger deviation potential energy well that adaptively compensates for long-period cumulative phase deviation.

[0101] In the continuous conveying scenario of high-speed printing of decorative paper, the horizontal displacement of the printed pattern in the full-frame image is directly mapped to the phase deviation along the time axis in the physical spatiotemporal coordinate system. Since the mechanical deviation at the physical level exhibits a low-frequency cumulative characteristic that evolves slowly over time, it is difficult to extract this long-period, minute phase slip from high-frequency noise simply by capturing the instantaneous signal of the physical marker. Therefore, the horizontal offset component distributed along the horizontal axis of the image coordinate space is extracted from the displacement vector. Phase reference reconstruction is performed using this horizontal offset component to initiate a reverse correction process from visual absolute feedback to mechanical execution reference.

[0102] Specifically, the spatial mapping ratio between the image dimension and the pulse dimension is determined. By retrieving the total pixel scale corresponding to the horizontal axis in the image coordinate space, the pixel span corresponding to the total number of pulses generated by one rotation of the printing roller is identified, and this pixel span is defined as the total horizontal pixel length. The total horizontal pixel length represents the pixel sampling depth of the full-frame image covering a complete printing cycle in the time dimension. To unify the geometric deviation dimension at the visual level to the pulse dimension in the physical spatiotemporal coordinate system, the total number of pulses generated by one rotation of the printing roller is divided by the total horizontal pixel length to obtain the pulse equivalent reflecting the number of pulses corresponding to a unit pixel. The pulse equivalent represents the mechanical rotation step size represented by each pixel at the current image resolution; its function is to serve as a linear conversion coefficient between the visual displacement and the mechanical phase correction. The horizontal offset component in the displacement vector is multiplied by the pulse equivalent to convert the pixel deviation in the image domain into a pulse deviation value in the mechanical domain. The pulse deviation value is a digital compensation variable used to characterize the degree of deviation of the full-frame image from the theoretical synchronization phase in spatial scale, under the vertical axis dimension of the physical spatiotemporal coordinate system. Its function is to eliminate the unit barrier between the image pixel domain and the mechanical pulse domain, and to transform the abstract pixel displacement into phase adjustment data that can be directly used to offset physical mechanical deviations.

[0103] To ensure the kinematic smoothness of the correction process and avoid drastic changes in control commands caused by instantaneous calculation fluctuations in single-frame full-frame image recognition, a forgetting factor is introduced to perform iterative correction. The forgetting factor is defined as a proportional coefficient used to adjust the intervention weight of the pulse deviation value on the theoretical prediction position. Its value is set based on the encoder's sampling frequency and the evolution period of the physical deformation of the decorative paper, aiming to balance the real-time response speed of the system correction and the global stability of the control loop; for example, it is set to 0.05. The pulse deviation value and the forgetting factor are multiplied to obtain the incremental correction amount. The theoretical prediction position is then translated using the incremental correction amount: the theoretical prediction position of the current printing cycle is subtracted from the incremental correction amount to obtain the corrected theoretical prediction position. Based on the corrected theoretical prediction position, spatial topology reconstruction is performed on the trigger deviation potential energy well. Specifically, the geometric center of the trigger deviation potential energy well in the physical spatiotemporal coordinate system is re-anchored to the corrected theoretical prediction position to obtain the corrected trigger deviation potential energy well. Through the above reverse correction, a visual-mechanical dual-closed-loop correction state is obtained. The described visual-mechanical dual-loop correction state is a balanced steady state constrained by real-time phase calibration provided by physical markers and absolute position feedback provided by the full-frame image. The aim is to ensure that the opening moment of the industrial camera's image shutter not only microscopically offsets the random jitter of the physical markers through centroid extraction and damped regression, but also macroscopically follows the long-period physical deformation of the decorative paper to perform reference migration, thereby ensuring that the full-frame image remains absolutely stable with zero drift within the display field of view. By reconstructing the trigger deviation potential energy well, a corrected trigger deviation potential energy well is obtained, achieving dynamic drift compensation for the reliability assessment benchmark of the physical markers during subsequent printing cycles. This ensures automatic calibration even during inertial gliding when physical markers are missing, following the deformation of the physical entity.

[0104] Step S30 solves the problem of slow image shifting and slipping out of the observation field caused by physical tension deformation, mechanical temperature drift, and slight slippage of the transmission mechanism in the synchronous monitoring system of high-speed decorative paper printing scenarios through full-frame image features, displacement vectors, phase reference reconstruction, and correction trigger deviation potential energy well. It realizes real-time feedback correction of the mechanical level synchronization reference at the visual level, achieving the technical effect of zero drift of the monitoring image throughout the entire cycle and visual-mechanical dual closed-loop correction. Among them, full-frame image features establish an absolute spatial reference in the complex decorative texture, providing a stable visual base for error extraction; displacement vectors realize accurate measurement of the cumulative phase deviation generated by the feedforward control logic; phase reference reconstruction uses pulse equivalent and forgetting factor to smoothly map the pixel deviation at the visual level to the pulse deviation value at the mechanical level and predict the position based on the translation theory, so that the correction trigger deviation potential energy well can perform dynamic spatial topology reconstruction with the deformation of physical entities, ensuring the long-term phase consistency of online monitoring of high-speed printing images.

[0105] Example 2

[0106] This embodiment, based on Embodiment 1, provides a fault-tolerant identification system for missing markings in decorative paper printing, such as... Figure 6 As shown, it includes:

[0107] Signal purification and screening module: used to acquire raw sensor data in high-speed printing scenarios to construct a physical spatiotemporal coordinate system, perform quantization and definition on the physical spatiotemporal coordinate system to obtain an effective signal fingerprint model, screen potential test signal waveform segments in the raw sensor data, perform full-dimensional comparison of the test signal waveform segments through the effective signal fingerprint model to obtain candidate trigger points, and perform serialization and arrangement on the candidate trigger points to obtain a preprocessed data stream;

[0108] State parsing module: used to construct a trigger deviation potential energy well based on the physical spatiotemporal coordinate system, parse the preprocessed data stream to obtain the intention state set, apply damped regression operation to the intention state set to obtain the weighted trigger position, perform inertial glide compensation and weight smooth switching on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0109] Visual feedback reconstruction module: used to acquire the full-frame image based on the smooth trigger command and perform absolute phase calibration and geometric deviation analysis to obtain the displacement vector. Then, it performs phase reference reconstruction based on the displacement vector to obtain the correction trigger deviation potential energy well for adaptive compensation of long-period cumulative phase deviation.

[0110] In the signal purification and screening module, the original sensor data obtained in the high-speed printing scenario is used to construct a physical spatiotemporal coordinate system. This system is then quantized and defined to obtain an effective signal fingerprint model. Potential waveform segments of the signal to be tested are screened from the original sensor data. The effective signal fingerprint model is used to perform a full-dimensional comparison of these waveform segments to obtain candidate trigger points. These candidate trigger points are then serialized and arranged to obtain a preprocessed data stream, including:

[0111] Step S11: Obtain the raw sensing data of the sensor in the high-speed printing scenario to construct a physical spatiotemporal coordinate system, perform quantization and definition on the physical spatiotemporal coordinate system, and obtain an effective signal fingerprint model.

[0112] Step S12: Dynamically filter out potential signal waveform segments in the original sensor data using the pre-read window, compare the signal waveform segments with the effective signal fingerprint model in all dimensions, and obtain candidate trigger points carrying confidence labels.

[0113] Step S13: Perform serialization and arrangement on the candidate trigger points to obtain a preprocessed data stream that eliminates physical background noise interference.

[0114] In the state parsing module, the step of constructing a trigger deviation potential energy well based on the physical spatiotemporal coordinate system, parsing the preprocessed data stream to obtain the intention state set, applying damped regression calculation to the intention state set to obtain the weighted trigger position, and performing inertial glide compensation and weight smoothing switching on the intention state set to obtain a smooth trigger command that eliminates phase step, including:

[0115] Step S21: Construct a trigger bias potential energy well based on the physical spatiotemporal coordinate system, analyze the preprocessed data stream, and obtain the intention energy characterizing the phase distortion weight of the physical marker.

[0116] Step S22: Analyze the intent energy to obtain the intent state set, and apply damped regression operation to the intent state set to obtain the weighted trigger position of the image shutter opening command.

[0117] Step S23: Based on the weighted trigger position, perform inertial glide compensation and weight smooth switching on the intention state set to obtain a smooth trigger command that eliminates phase step.

[0118] In the visual feedback reconstruction module, the process of acquiring a full-frame image based on a smooth trigger command, performing absolute phase calibration and geometric deviation analysis to obtain a displacement vector, and performing phase reference reconstruction based on the displacement vector to obtain a correction trigger deviation potential energy well for adaptive compensation of long-period cumulative phase deviation includes:

[0119] Step S31: Acquire the full-frame image acquired based on the smooth trigger command and perform absolute phase calibration and geometric deviation analysis to obtain the displacement vector representing the cumulative phase drift.

[0120] Step S32: Perform phase reference reconstruction based on displacement vector to obtain a corrected trigger deviation potential energy well that adaptively compensates for long-period cumulative phase deviation.

[0121] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0122] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for error-tolerant identification of missing printed markings on decorative paper, characterized in that, The method includes: The raw sensor data of the sensor in the high-speed printing scenario is acquired to construct a physical spatiotemporal coordinate system. The physical spatiotemporal coordinate system is quantized and defined to obtain an effective signal fingerprint model. Potential test signal waveform segments in the raw sensor data are screened. The test signal waveform segments are compared in all dimensions through the effective signal fingerprint model to obtain candidate trigger points. The candidate trigger points are serialized and arranged to obtain a preprocessed data stream. A trigger deviation potential energy well is constructed based on the physical spacetime coordinate system. The preprocessed data stream is parsed to obtain the intention state set. Damped regression operation is applied to the intention state set to obtain the weighted trigger position. Inertial glide compensation and weight smooth switching are performed on the intention state set to obtain a smooth trigger command that eliminates phase step. The full-frame image acquired based on the smooth trigger command is obtained and absolute phase calibration and geometric deviation analysis are performed to obtain the displacement vector. Phase reference reconstruction based on the displacement vector is performed to obtain the correction trigger deviation potential energy well for adaptive compensation of long-period cumulative phase deviation. The method for obtaining the effective signal fingerprint model includes: the physical spatiotemporal coordinate system is a two-dimensional reference system with absolute time as the horizontal axis (time axis) and the cumulative rotation angle of the printing roller (cumulative pulse count) as the vertical axis. The printing roller is a high-speed rotating cylinder that carries decorative texture and synchronization information in the printing production process of decorative paper. A physical mark is etched or attached to its surface, and a sensor is provided facing the surface of the printing roller. At the same time, an encoder is provided on the coaxial side of the printing roller. The theoretical pulse width is determined based on the physical width of the physical marker and the resolution of the encoder; Set an allowable deviation, and define the spatial span feature and the spatial span feature interval based on the allowable deviation and the theoretical pulse width. The spatial span feature interval includes an upper limit and a lower limit. By setting a small observation window and a jitter threshold, a time-domain edge steepness feature is established to characterize the stability of level transitions; The spatial span feature interval and the temporal edge steepness feature together constitute an effective signal fingerprint model; The method for obtaining the intent state set is as follows: a steady-state threshold and an escape threshold are preset, and the intent energy is numerically compared with the steady-state threshold and the escape threshold to obtain the intent state set, which includes strong intent state, weak intent state and discrete state. The damped regression operation refers to determining the physical weight value and theoretical weight value using the intention energy for a weak intention state, and performing a weighted summation operation on the centroid moment of the current printing cycle and the theoretically predicted position to obtain the weighted trigger position. The absolute phase calibration and geometric deviation analysis process is as follows: a full-frame image is pre-acquired using an industrial camera as a reference image, and a preset template containing geometric contours is extracted from the decorative texture of the reference image. The full-frame image is a pixel matrix composed of digital pixels. The time axis scale and cumulative pulse count scale in the physical spatiotemporal coordinate system are mapped to the pixel matrix to establish the image coordinate space; Calculate the absolute value of the brightness difference between each pixel and its neighboring pixels in the pixel matrix, and define the absolute value as the gray-level gradient change feature. Using a preset template, a correlation search is performed within the full-frame image of the current frame, and the pixel region with the highest matching degree with the gray-level gradient change feature of the preset template is locked as a visual anchor point. Obtain the geometric center pixel coordinates of the visual anchor point as the observation target. The displacement vector is obtained by calculating the two-dimensional spatial vector difference between the observed target center and the preset ideal target center.

2. The method for identifying missing decorative paper printing marks as described in claim 1, characterized in that, The method for obtaining the candidate trigger point includes: The raw sensing data includes the sensor's level state sequence and the encoder's pulse count value; The level state sequence is detected, and when a rising edge transition occurs from low level to high level, a read-ahead window based on the physical space dimension is opened; Obtain the pulse frequency output by the encoder at the current moment, and use the physical time length determined by the upper limit of the interval and the pulse frequency as the duration of the pre-read window; During the duration of the pre-read window, the level state sequence and the encoder pulse count value are acquired to form a waveform segment of the signal under test; The valid signal fingerprint model is used to verify the waveform segment of the signal under test to obtain the valid signal. The valid signal is then weighted and averaged to obtain the center of mass on the time axis as the centroid moment, and the centroid moment is marked as a candidate trigger point.

3. The method for identifying missing decorative paper printing marks as described in claim 2, characterized in that, The method for acquiring the preprocessed data stream includes: Calculate the absolute value of the spatial deviation between the encoder difference corresponding to the waveform segment of the signal under test and the theoretical pulse width, and determine the confidence level value based on the ratio of the absolute value of the spatial deviation to the allowable deviation. Here, the encoder difference refers to the difference between the encoder values ​​corresponding to the rising and falling edges of the high-level pulse. The confidence scores are associated with and encapsulated with candidate trigger points in the form of confidence labels to obtain a trigger feature vector carrying digital quality proof. The waveform segments of the test signal that do not meet the valid signal fingerprint model are identified as noise and removed. The candidate trigger points are then serialized and arranged according to their generation order in the physical spatiotemporal coordinate system to obtain the preprocessed data stream.

4. The method for identifying missing decorative paper printing marks as described in claim 3, characterized in that, The specific process of parsing the preprocessed data stream includes: The pulse count values ​​corresponding to the candidate trigger points recorded by the encoder in N consecutive historical printing cycles are retrieved. The arithmetic mean of the difference between the pulse count values ​​between two adjacent historical printing cycles is calculated as the pulse interval mean. The pulse interval mean is added to the pulse count value of the previous historical printing cycle to obtain the expected coordinate point as the theoretical predicted position. The printing cycle is the complete physical process of the printing roller completing a 360-degree mechanical rotation. A trigger deviation axis is established with the mapping point of the theoretically predicted position on the time axis as the origin; Define the potential energy function of the parabolic geometry on the trigger deviation axis; The logical evaluation region is formed by the quadratic projection envelope of the potential energy function on the trigger deviation axis, and the trigger deviation potential energy well is obtained. The deviation is substituted into the potential energy function for calculation to obtain the intention energy characterizing the phase distortion weight of the physical marker. The deviation is defined as the algebraic difference between the centroid time corresponding to the physical marker and the theoretically predicted position on the time axis of the physical spatiotemporal coordinate system.

5. The method for identifying missing decorative paper printing marks as described in claim 4, characterized in that, The method for obtaining the smooth triggering command includes: When the intent state set is discrete, the weighted trigger position determined by the last printing cycle in a strong intent state or a weak intent state is taken as the starting phase anchor point, and phase accumulation recursion is performed to obtain the virtual inertial trigger coordinates. Monitor the time series variation of intention energy values ​​over M consecutive printing cycles, extract the numerical difference of intention energy between two adjacent printing cycles and divide it by the time interval between the two corresponding printing cycles to obtain the regression rate of intention energy; and dynamically generate a weighted evolution curve based on the regression rate. Based on the weight evolution curve, the physical weight values ​​are faded in from zero to non-zero values ​​over M consecutive printing cycles. The inertial trigger coordinates and the re-extracted centroid time are used to perform weighted synthesis to obtain a smooth trigger command.

6. The method for identifying missing decorative paper printing marks as described in claim 5, characterized in that, The method for obtaining the corrected trigger deviation potential energy well includes: The total length of horizontal pixels corresponding to a complete printing cycle in the image coordinate space of the pixel matrix of the full-frame image is identified, and the ratio of the total number of pulses generated by one rotation of the printing roller to the total length of the horizontal pixels is calculated to obtain the pulse equivalent. The displacement vector is decomposed into a horizontal offset component along the horizontal axis of the image coordinate space and a vertical offset component along the vertical axis. The horizontal offset component is then multiplied by the pulse equivalent and a preset forgetting factor to obtain the incremental correction amount. The horizontal axis of the image coordinate space corresponds to the time axis of the physical spatiotemporal coordinate system, and the vertical axis of the image coordinate space corresponds to the vertical axis of the physical spatiotemporal coordinate system. Subtract the incremental correction from the theoretical prediction position to obtain the corrected theoretical prediction position; The geometric center of the trigger deviation potential energy well is re-anchored to the corrected theoretical prediction position to obtain the corrected trigger deviation potential energy well.

7. A system for identifying and correcting missing decorative paper printing marks, used to implement the method for identifying and correcting missing decorative paper printing marks as described in any one of claims 1-6, characterized in that, The system includes: Signal purification and screening module: This module acquires raw sensor data from sensors in high-speed printing scenarios to construct a physical spatiotemporal coordinate system. It then performs quantization and definition on this system to obtain an effective signal fingerprint model. Potential test signal waveform segments are screened from the raw sensor data. The effective signal fingerprint model is used to perform a full-dimensional comparison of these segments to obtain candidate trigger points. These candidate trigger points are then serialized and arranged to obtain a preprocessed data stream. The method for acquiring the effective signal fingerprint model includes: the physical spatiotemporal coordinate system is a two-dimensional reference system with absolute time as the horizontal axis (time axis) and the cumulative rotation angle of the printing roller (cumulative pulse count) as the vertical axis. The printing roller is a component used in the printing process of decorative paper. In the production process, a high-speed rotating cylinder carrying decorative textures and synchronization information has a physical mark etched or attached to its surface. A sensor is positioned directly opposite the printing roller surface, and an encoder is located on the coaxial side of the printing roller. Based on the physical width of the physical mark and the encoder's resolution, the theoretical pulse width is determined. An allowable deviation is set, and based on this deviation and the theoretical pulse width, a spatial span feature and a spatial span feature interval are defined. The spatial span feature interval includes an upper and lower limit. A small observation window and a jitter threshold are set to establish a temporal edge steepness feature to characterize the stability of level transitions. The spatial span feature interval and the temporal edge steepness feature together constitute an effective signal fingerprint model. The state analysis module is used to construct a trigger deviation potential energy well based on the physical spatiotemporal coordinate system, analyze the preprocessed data stream to obtain the intention state set, apply damped regression operation to the intention state set to obtain the weighted trigger position, and perform inertial glide compensation and weight smooth switching on the intention state set to obtain a smooth trigger command that eliminates phase step. The method for obtaining the intention state set is as follows: preset steady-state threshold and escape threshold, compare the intention energy with the steady-state threshold and escape threshold to obtain the intention state set, which includes strong intention state, weak intention state and discrete state; the damped regression operation refers to, for the weak intention state, using the intention energy to determine the physical weight value and theoretical weight value, and performing a weighted summation operation on the centroid time of the current printing cycle and the theoretical predicted position to obtain the weighted trigger position; The visual feedback reconstruction module is used to acquire a full-frame image based on a smooth trigger command and perform absolute phase calibration and geometric deviation analysis to obtain a displacement vector. It then performs phase reference reconstruction based on the displacement vector to obtain a correction trigger deviation potential energy well that adaptively compensates for long-period cumulative phase deviation. The absolute phase calibration and geometric deviation analysis process involves: pre-acquiring a full-frame image using an industrial camera as a reference image; extracting a preset template containing geometric contours from the decorative texture of the reference image; and using a pixel matrix composed of digital pixels. The time axis scale in the physical spatiotemporal coordinate system is then... The cumulative pulse count scale is mapped to the pixel matrix to establish an image coordinate space; the absolute value of the brightness difference between each pixel and its adjacent pixels is calculated, and the absolute value is defined as the gray-level gradient change feature; a correlation search is performed on the full-frame image of the current frame using a preset template, and the pixel region with the highest matching degree with the gray-level gradient change feature of the preset template is locked as the visual anchor point; the geometric center pixel coordinates of the visual anchor point are obtained as the observation target center; the two-dimensional spatial vector difference between the observation target center and the preset ideal target center is calculated to obtain the displacement vector.

Images