A printing roller positioning information acquisition and deviation compensation method

Abstract

The application provides a printing roller positioning information acquisition and deviation compensation method, which comprises the following steps: obtaining the diameter, surface roughness and rotating direction of a printing roller, and establishing a circumferential positioning origin of the roller surface by using an encoder zero position pulse; when the elongation length of a light spot exceeds a preset length threshold, the photoelectric signal of the corresponding circumferential arc segment is intercepted, the adjacent sampling values are differentiated, and the extreme value position of the change rate is extracted to generate a second photoelectric signal with an edge position index; the edge blur width is adaptively clustered and divided in intervals according to the curvature of the corresponding arc segment of the second photoelectric signal, and a matching dynamic evaluation threshold is set for each width change interval in combination with the surface roughness of the printing roller; the dynamic evaluation threshold of the corresponding interval is selected by using the edge position index in the second photoelectric signal, and the second photoelectric signal is subjected to outlier rejection and sliding window smoothing processing to obtain a third photoelectric signal that retains the marked edge jump.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for collecting and compensating for the positioning information of printing rollers. Background Technology

[0002] In modern high-speed printing production, multi-layer overprinting requires that the same pattern on all printing rollers precisely overlap at the same position on the substrate. To achieve precise overprinting, the equipment must first know the positioning mark of each printing roller during its rotation. This timing is obtained by a photoelectric sensor installed near the printing roller, which emits a beam of light onto the roller surface and receives the light reflected back from the marked area. The edge of the mark is determined by the abrupt change in the intensity of the reflected light. Once the edge of the positioning mark passes the detection point, the reflected light signal will change from bright to dark or from dark to bright. Based on this change, the relative positional deviation between the printing rollers is calculated, and the roller speed or phase is adjusted in real time to ensure overprinting accuracy. As printing speeds increase, the rotational speed of the printing rollers becomes faster and faster, and the time window for the edge of the positioning mark to pass through the detection point is significantly shortened. The photoelectric sensor must accurately capture the change edge of the reflected light in a very short time; otherwise, the obtained positioning time will be delayed or blurred, leading to inaccurate overprinting compensation. However, in actual operation, the printing roller surface is covered with ink layers of varying thicknesses. These ink layers have strong specular reflective properties, concentrating and reflecting incident light along the reflection angle. The positioning markers themselves are typically made of matte or rough-surfaced materials, relying on diffuse reflection to generate edge transition signals. The photoelectric sensor needs to receive diffusely reflected light from the marker area to identify the brightness difference between the marker and the surrounding ink surface, thus generating the edge transition signal. Currently, most acquisition schemes use factory-set fixed optical axis angles and fixed edge detection thresholds. The optical axis angle refers to the angle between the centerline of the incident beam emitted by the sensor and the normal to the printing roller surface. This fails to consider variations in roller diameter, ink batches, and roller surface condition during operation, leading to significant fluctuations in signal quality after roller or material changes, edge detection time drift, and gradually amplifying registration errors. Intuitively, changing the sensor's optical axis from perpendicular to the roller surface to an angle deviating from the normal would allow the specular reflection of the ink to avoid the receiving window. This would deflect the strong specular reflection from the ink layer along the reflection angle, thus avoiding the receiving window. In this way, only diffuse reflection light from the marking area would enter the probe, reducing interference in the signal and making the edges appear clearer and sharper. However, the roller surface is curved. Once the light is tilted, the light spot projected onto the curved surface is elongated along the curvature. The originally clear boundary between light and dark becomes a slowly transitioning blurry band, blurring the marking edges and making it difficult to accurately lock onto transition moments. The same action of increasing the tilt angle benefits glare suppression but loses edge sharpness; both occur simultaneously in the same device, but in opposite directions.For example, when the production line switches from a small-diameter printing roller to a large-diameter printing roller, in order to suppress the more obvious mirror reflection on the new roller surface, the device often needs a larger tilt angle. The increased tilt angle leads to a more significant light spot elongation effect. The collected mark edge changes from a steep jump of tens of microseconds to a widened undulating slope. If the trigger circuit still uses the original fixed threshold to determine the edge arrival time, the position corresponding to the fixed threshold value will move left and right on the slope due to signal noise or amplitude fluctuations, causing the determination time to be earlier or later. The deviation compensation module receives a distorted positioning time and believes that the phase deviation between the rollers has deviated. Therefore, it outputs a reverse speed adjustment command. However, because this positioning time itself is inaccurate, the actual overprinting deviation after adjustment not only does not decrease, but will become larger and larger. Summary of the Invention

[0003] This invention provides a method for collecting and compensating for positioning information of printing rollers, mainly including: The printing roller diameter, surface roughness, and rotation direction are obtained, and the circumferential positioning origin of the roller surface is established using the encoder zero-position pulse; The first photoelectric signal is continuously acquired in the color mark area, and the rotation speed, time index and encoder angle index are recorded simultaneously. The signal is decomposed into specular reflection component and diffuse reflection component, mapped to the actual arc position and the length of the light spot is determined. When the length of the light spot exceeds the preset length threshold, the photoelectric signal of the corresponding arc segment is intercepted, the adjacent sampled values ​​are differentiated and the extreme position of the rate of change is extracted to generate a second photoelectric signal with an edge position index. The edge fuzzy width is adaptively clustered and divided according to the curvature of the arc segment corresponding to the second photoelectric signal, and a dynamic evaluation threshold is set for each width variation interval in combination with the surface roughness. By using the edge position index to select the dynamic evaluation threshold, outlier removal and sliding window smoothing are performed on the second photoelectric signal to obtain a third photoelectric signal that retains the edge transition. The third photoelectric signal is compared with the dynamic evaluation threshold to identify the first set of sampling points that continuously cross the threshold. The starting point time index is used as the target jump time. The target rotation angle is determined according to the encoder angle index. The time offset is obtained by subtracting the preset positioning time. Based on the time offset and the rotation direction, the circumferential compensation distance is determined, and a circumferential position deviation compensation instruction for the printing plate is generated.

[0004] Furthermore, the step of obtaining the printing roller diameter, surface roughness, and rotation direction, and establishing the circumferential positioning origin of the roller surface using the encoder zero-position pulse, includes: The roller diameter is obtained by a contact measuring device, the roller surface roughness parameters are collected by a surface profilometer, and the rotation direction is obtained by reading the feedback signal of the drive motor. The roller surface angle position corresponding to the rising edge of the zero pulse is used as the circumferential positioning origin, and the roller surface arc length is angularly indexed and encoded.

[0005] Furthermore, the step of continuously acquiring the first photoelectric signal in the color mark area, simultaneously recording the rotation speed, time index, and encoder angle index, decomposing the signal into specular reflection and diffuse reflection components, mapping it to the actual arc segment position, and determining the light spot elongation length includes: Photoelectric probes are deployed in the color mark area to collect the intensity of reflected light to form the first photoelectric signal, and the rotation speed, timestamp and angle index are read simultaneously; the specular reflection component and diffuse reflection component are separated based on the waveform difference between narrow peak high amplitude and wide peak low amplitude and mapped to the actual arc position; the first sampling point and the last sampling point of the switching of the dominant relationship between the two components of specular reflection component and diffuse reflection component are calibrated to determine the length of the light spot elongation.

[0006] Furthermore, the determination of the first and last sampling points for switching the dominance relationship between the two components, and the determination of the spot elongation length, includes: The dominance relationship between the specular reflection component and the diffuse reflection component is determined point by point along the rotation direction of the printing roller. Sampling points where the amplitude of the specular reflection component is greater than that of the diffuse reflection component are marked as specular dominance, and sampling points where the amplitude of the diffuse reflection component is greater than that of the specular reflection component are marked as diffuse reflection dominance. The sampling point that switches from specular dominance to diffuse reflection dominance is captured as the first sampling point, and the sampling point that switches back from diffuse reflection dominance to specular dominance is captured as the last sampling point. By tracing back the actual arc position of the first and last sampling points, the elongation length of the light spot is determined.

[0007] Furthermore, when the light spot elongates beyond a preset length threshold, the photoelectric signal of the corresponding arc segment is intercepted, the difference between adjacent sampled values ​​is calculated, and the extreme value position of the rate of change is extracted to generate a second photoelectric signal with an edge position index, including: The length of the light spot is compared with a preset length threshold. If it is greater than the threshold, the circumferential arc range is defined based on the arc positions of the first and last sampling points. The set of sampling points falling within the range is extracted from the first photoelectric signal. The first-order difference is performed on the extracted sampling points to obtain the rate of change sequence. The position of the sampling point with the largest absolute value of the rate of change is captured as the extreme position of the rate of change. The angle index of the extreme position of the rate of change and the actual arc position are recorded as the edge position index.

[0008] Furthermore, the step of adaptively clustering the edge fuzziness width interval according to the curvature of the arc segment corresponding to the second photoelectric signal, and setting a dynamic evaluation threshold for each width variation interval in combination with surface roughness, includes: The actual arc position and edge ambiguity width of each sampling point are extracted from the second photoelectric signal to form a two-dimensional feature vector. K-means clustering is used to divide the interval along the ambiguity width dimension and the mean value is taken as the cluster center width. The threshold bias is calculated according to the surface roughness and the bias is superimposed on the cluster center width to obtain the dynamic evaluation threshold.

[0009] Furthermore, the step of selecting a dynamic evaluation threshold using edge position indexing, performing outlier removal and sliding window smoothing on the second photoelectric signal to obtain a third photoelectric signal that retains edge transitions includes: Read the edge fuzzy width bound to the edge position index sampling point, classify it into the corresponding width change interval according to the upper and lower boundary classification, retrieve the dynamic evaluation threshold of the corresponding width change interval and convert it into the amplitude judgment threshold; traverse the second photoelectric signal along the rotation direction, when the absolute value of the difference between the amplitude of the sampling point and the median of the adjacent sampling points exceeds the threshold, it is judged as an outlier and replaced by the median; perform sliding window smoothing on the signal after removal, the window width is divided by the width of the cluster center of the interval and rounded down, and the original amplitude of the edge position index sampling point is not included in the mean replacement.

[0010] Furthermore, the step of comparing the third photoelectric signal with the dynamic evaluation threshold, identifying the first set of sampling points that consecutively cross the threshold, using the starting point time index as the target transition time, determining the target rotation angle based on the encoder angle index, and subtracting the time offset from the preset positioning time to obtain the time offset includes: The smoothed amplitude is compared with the amplitude judgment threshold of the interval and marked as having crossed the threshold. The sampling point sequence with the crossing mark is traversed, and the first set of three or more consecutive sampling points that have crossed the threshold is captured. The first sampling point is taken as the starting point, and the starting point time index and angle index are read to calculate the target rotation angle relative to the circumferential positioning origin.

[0011] Furthermore, the step of determining the circumferential compensation distance based on the time offset and the rotation direction, and generating the circumferential position deviation compensation command for the printing plate, includes: The current instantaneous rotation speed of the printing roller is read from the servo driver, the roller diameter is retrieved from the initial parameter set of the roller surface, and the circumferential linear velocity of the roller surface is calculated by multiplying the roller diameter by half the angular velocity corresponding to the instantaneous rotation speed. The time offset is multiplied by the circumferential linear velocity to obtain the circumferential overprinting position offset of the color mark edge from the preset positioning position.

[0012] Furthermore, after multiplying the time offset by the circumferential linear velocity to obtain the circumferential overprint position offset of the color mark edge from the preset positioning position, the process includes: The inverse of the circumferential overprint position offset is used as the current deviation input for the PID controller. The PID compensation output value is obtained by multiplying the current deviation, the cumulative deviation, and the change in deviation by the proportional, integral, and derivative coefficients, respectively. The circumferential compensation distance is determined according to the rotation direction. A compensation command is generated based on the circumferential compensation distance and sent to the servo control interface. The target rotation angle is superimposed with the angle increment, and the target jump moment is superimposed with the time increment, and then written to the positioning reference register of the next cycle.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for acquiring and compensating for positioning information of printing rollers. Addressing the core problems of inaccurate color mark edge positioning and difficulty in accurately correcting circumferential overprinting deviations caused by roller surface geometry and the elongation effect of the light spot during printing, this invention proposes a complete solution from signal acquisition to closed-loop deviation compensation. The invention establishes the circumferential positioning origin of the roller surface using the encoder zero-position pulse, separates the specular reflection and diffuse reflection components in the photoelectric signal and maps them to the actual arc segment position. The elongation length of the light spot is quantified by calibrating the first and last sampling points of the switching of the dominant relationship between the two components, thereby effectively eliminating the interference of light spot deformation on edge recognition. Based on this, the invention adaptively divides the edge ambiguity interval according to the arc segment curvature and roller surface roughness and dynamically sets the evaluation threshold. Combined with outlier removal and sliding window smoothing, it accurately locks the first target jump moment that continuously crosses the threshold. Then, it uses the encoder angle index to determine the target rotation angle, converts the time offset into a circumferential overprinting position offset by combining the current rotation speed and roller diameter, generates a deviation compensation command through PID compensation, and injects the result into the positioning reference for the next printing cycle, achieving high-precision adaptive closed-loop correction of circumferential overprinting deviations. Attached Figure Description

[0014] Figure 1 This is a flowchart of a printing roller positioning information acquisition and deviation compensation method according to the present invention.

[0015] Figure 2 This is a schematic diagram of a printing roller positioning information acquisition and deviation compensation method according to the present invention.

[0016] Figure 3 This is another schematic diagram of a printing roller positioning information acquisition and deviation compensation method according to the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0018] like Figures 1-3 The method for collecting and compensating for positioning information of printing rollers in this embodiment may specifically include: Step S101: Obtain the printing roller diameter, surface roughness, and rotation direction, and establish the circumferential positioning origin of the roller surface using the encoder zero-position pulse.

[0019] The printing roller diameter is obtained through a contact measuring device, and the surface roughness parameters of the printing roller surface are collected using a surface profilometer. The rotation direction of the printing roller is obtained by reading the feedback signal from the drive motor. The roller diameter, roughness, and rotation direction together constitute the initial parameter set of the roller surface. The zero-position pulse signal output by the encoder is collected, and the rising edge of the zero-position pulse is captured. The roller surface angle position corresponding to this moment is used as the circumferential positioning origin. According to the rotation direction in the initial parameter set of the roller surface, the arc length of the roller surface is angularly indexed and encoded along the rotation path of the printing roller to obtain the circumferential positioning origin of the roller surface.

[0020] In one embodiment, a caliper-type contact diameter measuring device is attached to the outer circumferential surface of the middle section of the printing roller body, and the roller diameter value is obtained by reading the span scale of the two measuring jaws of the caliper. The measuring point is selected at a position far away from the chamfer of the shaft end to avoid the processing transition zone. A stylus-type surface profilometer is used to slide along the generatrix of the roller surface at a uniform speed of 0.5 mm per second. The vertical displacement signal of the stylus is amplified by an amplifier by 100-500 times and then sampled discretely at intervals of 0.01 mm to output the roller surface roughness profile curve. The arithmetic mean deviation of the absolute value of the profile curve from the centerline within the evaluation length is taken as the roughness parameter.

[0021] For example, for gravure printing rollers, the roller diameter measurement point is usually sampled at the middle section of the roller body circumference to avoid measurement deviation caused by the chamfered area at the shaft end.

[0022] Specifically, the drive motor feedback signal comes from the encoder output interface of the servo driver. The rotation direction of the printing roller is determined by judging the phase lead relationship between the A-phase and B-phase pulses of the encoder. When the A-phase leads the B-phase, it is determined to be forward rotation, and vice versa, it is reverse rotation. The roller diameter, roughness parameters, and rotation direction together constitute the initial parameter set of the roller surface.

[0023] In one possible implementation, the encoder outputs a zero-position pulse signal per revolution. The sampling circuit captures the rising edge of the zero-position pulse level from low to high and marks the physical position on the roller surface at that moment, which is directly opposite the detection point of the photoelectric sensor, as the circumferential positioning origin.

[0024] It is understood that, based on the rotation direction in the initial parameter set of the roller surface, the circumferential angle index is encoded incrementally along the rotation path of the printing roller. Each incremental pulse output by the encoder corresponds to the roller surface rotating through an angle increment Δθ. The arc length ΔL of the roller surface corresponding to this angle increment satisfies ΔL=R*Δθ, where R is half of the diameter of the printing roller, and Δθ is in radians. Each sampling moment corresponds to a unique angle index and arc length position, thus obtaining the circumferential positioning origin of the roller surface.

[0025] Step S102: Continuously collect the first photoelectric signal in the color mark area, synchronously record the rotation speed, time index and encoder angle index, decompose the signal amplitude into specular reflection component and diffuse reflection component, map each sampling point to the actual arc position relative to the circumferential positioning origin of the roller surface, and determine the length of the light spot elongation.

[0026] A photoelectric probe is arranged in the color mark area on the printing roller surface. The intensity of reflected light is continuously collected at a fixed sampling frequency along the rotation direction of the printing roller to form a first photoelectric signal. Simultaneously, the instantaneous rotation speed of the printing roller is read from the servo driver to generate a rotation speed index, a timestamp is read from the sampling clock to generate a time index, and an angle index is read from the encoder incremental pulse accumulator. The rotation speed index, time index, and angle index are bound to each sampling point of the first photoelectric signal, resulting in a first photoelectric signal sampling sequence with triple indices. For the amplitude of each sampling point in the first photoelectric signal sampling sequence, based on the difference between narrow-peak high-amplitude waveforms and wide-peak low-amplitude waveforms in the spatial distribution characteristics of reflected light, the amplitude is separated into a specular reflection component corresponding to the concentrated reflection of the ink layer and a diffuse reflection component corresponding to the matte scattered light of the color mark. Based on the angle index and the circumferential positioning origin of the roller surface, each sampling point is mapped to the actual arc segment position relative to the circumferential positioning origin according to the product relationship between the roller diameter and the corresponding radian value of the angle index, resulting in an arc segment mapping sequence marked with specular and diffuse reflection components. In the arc segment mapping sequence, the magnitudes of the specular reflection component and the diffuse reflection component are compared point by point along the rotation direction of the printing roller. The first sampling point where the specular reflection component amplitude is higher than the diffuse reflection component amplitude is switched to the point where the diffuse reflection component amplitude is higher than the specular reflection component amplitude is calibrated, as well as the last sampling point where the switch is reversed. The actual arc segment positions corresponding to the first and last sampling points are traced back, and the circumferential arc segment span along the rotation path of the printing roller between the two positions is measured to obtain the light spot elongation length.

[0027] In one embodiment, the photoelectric probe continuously sweeps along the color mark area on the surface of the printing roller at a fixed sampling frequency. The sampling frequency is selected to match the instantaneous linear velocity of the printing roller. The higher the sampling frequency, the denser the sampling points obtained per unit arc segment. The intensity of the collected reflected light is converted by photoelectric conversion and pre-amplified to form a first photoelectric signal. The first photoelectric signal is sent into the sampling buffer in digital form to provide the original data basis for subsequent index binding.

[0028] Specifically, the rotational speed index is taken from the rotational speed feedback interface output by the servo driver. The feedback signal contains a digital value of the number of revolutions per minute of the printing roller. The sampling circuit extracts the instantaneous rotational speed value at each sampling point as the rotational speed index of that sampling point. The time index is taken from the millisecond-level timestamp of the sampling clock. The timestamp increases monotonically in the order of the sampling points. The angle index is taken from the encoder incremental pulse accumulator. The accumulator is cleared to zero at the circumferential positioning origin of the roller surface. The accumulator value is incremented by one for each received incremental pulse. The accumulated value corresponds to the angle that the roller surface has rotated relative to the circumferential positioning origin.

[0029] It should be noted that the rotation speed index, time index, and angle index are synchronously latched at each sampling moment. The three are bound one-to-one with the corresponding sampling point of the first photoelectric signal. The bound data set is arranged in the order of the time index, forming the first photoelectric signal sampling sequence with triple index.

[0030] In one possible implementation, the amplitude of each sampling point in the first photoelectric signal sampling sequence is separated into components based on the spatial distribution characteristics of the reflected light.

[0031] It is understandable that the surface of the ink layer is relatively smooth, and most of the energy of the incident light is concentrated and emitted along the reflection angle after encountering the ink layer. The part of the light entering the receiving window of the photodetector exhibits a narrow peak high amplitude waveform with extremely high instantaneous amplitude and extremely short duration. On the other hand, the surface of the color mark is a matte and rough material. After the incident light encounters the color mark surface, the energy is scattered in all directions. The part of the light entering the receiving window of the photodetector exhibits a wide peak low amplitude waveform with low instantaneous amplitude and long duration.

[0032] For example, let the amplitude of the first photoelectric signal at a certain sampling point be S. The component corresponding to the narrow-peak, high-amplitude waveform is denoted as Sm as the specular reflection component, and the component corresponding to the wide-peak, low-amplitude waveform is denoted as Sd as the diffuse reflection component. The two satisfy the amplitude superposition relationship S = Sm + Sd. The separation process is performed point-by-point using a sliding window with a width of 8 sampling points. First, the absolute value sequence of amplitude differences between adjacent sampling points within the window is calculated. If the absolute value of the amplitude difference at a sampling point k exceeds twice the average amplitude difference within the window, and the amplitudes at the two points before and after sampling point k all fall back to within 0.1*A0, then it is determined that a narrow-peak, high-amplitude waveform exists at sampling point k. At this time, for sampling point k itself, Sm = S - A0 and Sd = A0, meaning the bulge above the baseline is classified as the specular reflection component, and the remaining portion as the diffuse reflection component. For the remaining sampling points within the window, since no narrow-peak feature is detected, Sm = 0 and Sd = S, and all amplitudes are classified as diffuse reflection components. Thus, each sampling point yields a pair of Sm and Sd, and always satisfies the superposition relationship of S = Sm + Sd.

[0033] Preferably, the separated specular reflection component corresponds to the concentrated reflective component of the ink layer, and the diffuse reflection component corresponds to the scattering component of the color mark. Both components are jointly marked on the original sampling point, forming a set of sampling points with component markings. Further, based on the angle index and the circumferential positioning origin of the roller surface, each sampling point with component markings is mapped to the actual arc segment position relative to the circumferential positioning origin. The mapping relationship follows the conversion L=R*θ, where L is the actual arc segment position of the sampling point relative to the circumferential positioning origin, R is half of the printing roller diameter, and θ is the radian value corresponding to the angle index of the sampling point.

[0034] In one embodiment, after mapping, each sampling point has a unique actual arc segment position label. Several sampling points are arranged in ascending order of arc segment position to obtain an arc segment mapping sequence labeled with specular reflection component and diffuse reflection component. The arc segment mapping sequence simultaneously retains triple index and two component labeling information.

[0035] Specifically, the magnitudes of the specular reflection component Sm and the diffuse reflection component Sd are compared point by point along the rotation direction of the printing roller. If the previous position of a certain sampling point satisfies Sm>Sd, and the next 5 consecutive sampling points from that sampling point all satisfy Sd>Sm, then that sampling point is designated as the first sampling point; if the previous position of a certain sampling point satisfies Sd>Sm, and the next 5 consecutive sampling points from that sampling point all satisfy Sm>Sd, then that sampling point is designated as the last sampling point. The first sampling point corresponds to the position where the edge of the color mark just enters the coverage area of ​​the light spot, and the last sampling point corresponds to the position where the edge of the color mark just leaves the coverage area of ​​the light spot. In one embodiment, the actual arc segment positions bound to the first and last sampling points are traced back, and the circumferential arc segment span along the rotation path of the printing roller between the two positions is measured. The span value is the total length of the arc segment of the roller surface covered by the light spot after being stretched due to curvature when projected onto the curved roller.

[0036] It is understandable that the length of the light spot elongation reflects the actual widening scale of the boundary blurring band formed by the inclined optical axis on the arc roller. The larger the arc span, the wider the blurring band and the smoother the transition between light and dark at the edge of the mark.

[0037] One way to determine the elongation length of the light spot is to retrieve the specular reflection component and diffuse reflection component mapped to the actual arc segment position, calibrate the first and last sampling points where the dominance relationship of the two components switches along the rotation direction of the printing roller, trace back the first and last sampling points to their respective actual arc segment positions, measure the circumferential arc segment span between the actual arc segment position corresponding to the first sampling point and the actual arc segment position corresponding to the last sampling point, and output the elongation length of the light spot.

[0038] From the acquired arc segment mapping sequence, retrieve the specular reflection component and diffuse reflection component labeled at each sampling point. Determine the dominance relationship of the magnitudes of the specular reflection component and diffuse reflection component at each sampling point along the rotation direction of the printing roller. If the magnitude of the specular reflection component at a sampling point is greater than that of the diffuse reflection component, the sampling point is marked as specular dominance; if the magnitude of the diffuse reflection component is greater than that of the specular reflection component, the sampling point is marked as diffuse reflection dominance, resulting in a sampling point sequence labeled with dominance relationship. Traverse the sampling point sequence labeled with dominance relationship along the rotation direction of the printing roller, capturing the sampling point that switches from specular dominance to diffuse reflection dominance as the first sampling point, and capturing the sampling point that switches back from diffuse reflection dominance to specular dominance as the last sampling point. Backtrack the first and last sampling points to their respective actual arc segment positions bound in the arc segment mapping sequence to obtain the arc segment positions of the first and last sampling points. The length of the continuous arc segment between the first sampling point arc segment position and the last sampling point arc segment position is accumulated along the rotation path of the printing roller to obtain the circumferential arc segment span. The circumferential arc segment span is output as the light spot elongation length.

[0039] In one implementation, the specular reflection component and diffuse reflection component amplitudes marked on each sampling point are read point by point from the acquired arc segment mapping sequence. The two component amplitudes are stored in two fields of the sampling point data structure in the form of floating-point numbers. During reading, the sampling points are retrieved sequentially along the rotation direction of the printing roller according to their arrangement order in the arc segment mapping sequence, providing point-by-point input for determining the dominant relationship.

[0040] Specifically, the dominance relationship is determined by directly comparing the magnitudes of the two components. Let the amplitude of the specular reflection component at a certain sampling point be Sm, and the amplitude of the diffuse reflection component be Sd. If Sm is greater than Sd, the dominance relationship label for that sampling point is set to specular dominance; if Sd is greater than Sm, the dominance relationship label is set to diffuse reflection dominance.

[0041] It is understandable that within the ink layer region, Sm is at a high level and Sd is at a low level, and the difference between the two component amplitudes is typically greater than 30% of the full-scale amplitude of the photodetector output. The full-scale amplitude of the photodetector output refers to the digital amplitude corresponding to the maximum voltage allowed to be output by the photoelectric conversion circuit. Within the color mark region, Sd is at a high level and Sm is at a low level, and the difference between the two component amplitudes is also significant. Only in the transition region at the edge of the color mark are the amplitudes of Sm and Sd close. If the absolute value of the difference between the two component amplitudes is less than 5% of the full-scale amplitude of the photodetector output, then the dominant relationship label of that sampling point is set to a transition state and does not participate in subsequent switching determination. Furthermore, the dominant relationship label of each sampling point, along with its bound angle index, time index, and actual arc segment position, are saved together to obtain a sampling point sequence with dominant relationship labels. The sampling point sequence is arranged in ascending order of arc segment position.

[0042] In one possible implementation, the sampling point sequence with dominant relationship labels is advanced point by point from the beginning to the end along the rotation direction of the printing roller, and the switching of dominant relationship is identified by comparing adjacent pairs of points. Let the previous sampling point be the reference point and the next sampling point be the comparison point. The two points are directly adjacent in the sequence, and the advancement step size is 1 sampling point. If the dominant relationship label of the reference point is specular dominant and the dominant relationship label of the comparison point is diffuse dominant, then the comparison point is marked as the first sampling point; continuing to advance along the rotation direction, if the dominant relationship label of the reference point is diffuse dominant and the dominant relationship label of the comparison point is specular dominant, then the comparison point is marked as the last sampling point.

[0043] It should be noted that the switching determination process skips the transition state label sampling point. When a transition state label is encountered, the reference point remains unchanged, and the comparison point is postponed to the next non-transition state sampling point before comparison, so as to avoid the component amplitude near the edge being mistakenly judged as a switching point.

[0044] Specifically, a backtracking operation is performed on the first sampling point and the last sampling point respectively. The backtracking is based on the actual arc position field that is bound in the data structure of each sampling point, and the arc position Lf of the first sampling point and the arc position Lt of the last sampling point are obtained. Both positions are represented by the arc length value relative to the circumferential positioning origin of the roller surface.

[0045] Preferably, the position increment is accumulated segment by segment along the printing roller rotation path from Lf towards Lt. For normal segments, the arc length is calculated by multiplying the roller radius R by the radian value θ corresponding to the angle index difference between adjacent sampling points, and then accumulated. To identify sudden changes in the angle index caused by encoder counter overflow and return to zero or brief pulse loss, twice the number of pulses generated by the printing roller at its highest linear speed within a single sampling period is taken as the jump judgment threshold. When the absolute value of the index difference between adjacent sampling points exceeds this threshold, it is judged as a jump. At this time, the difference is first moduloed by the total number of pulses N per revolution and wound back to the normal range. Then, R is multiplied by the wound radian value to make up the actual arc length of the segment, avoiding accumulation deviation caused by counter reversal. The accumulated length of the continuous arc segment between the two positions is obtained as the circumferential arc segment span.

[0046] It is understandable that the circumferential arc span output is used as the light spot elongation length, which reflects the actual scale of the color mark marking area widening along the curvature direction after being covered by the light spot on the arc roller.

[0047] In step S103, when the length of the light spot exceeds the preset length threshold, the photoelectric signal of the corresponding circumferential arc segment is intercepted, the adjacent sampled values ​​are differentiated and the extreme position of the rate of change is extracted to generate a second photoelectric signal with an edge position index.

[0048] The length of the stretched light spot is compared with a preset length threshold. If the stretched length of the light spot is greater than the preset length threshold, then based on the circumferential arc range defined by the measured arc positions of the first and last sampling points, a set of sampling points falling within the circumferential arc range is extracted from the first photoelectric signal sampling sequence to obtain the extracted photoelectric signal corresponding to the arc. For the sampling points within the extracted photoelectric signal, a first-order difference operation is performed on two adjacent sampling values ​​along the rotation direction of the printing roller. The difference between the amplitude of the latter sampling value and the former sampling value is taken as the signal change rate at that sampling point position. The signal change rates at each position are arranged sequentially to obtain a change rate sequence. The change rate sequence is traversed, and the sampling point position where the absolute value of the change rate reaches the maximum value in the sequence is captured as the change rate extreme value position. The angle index corresponding to the change rate extreme value position and the actual arc position are jointly recorded as the edge position index. The edge position index is appended to the corresponding sampling point of the extracted photoelectric signal to obtain a second photoelectric signal with an edge position index.

[0049] In one implementation, the determined elongation length of the light spot is obtained as the input basis for threshold comparison, and the preset length threshold is stored in the threshold register in advance according to the matching relationship between the printing roller diameter and the ideal light spot diameter of the photoelectric probe.

[0050] For example, for printing rollers with a diameter between 0.1 meters and 0.3 meters, the preset length threshold is set to 1.5 times the ideal spot diameter of the photoelectric probe, and the length of the stretched spot is compared with the preset length threshold.

[0051] It should be noted that if the length of the light spot is not greater than the preset length threshold, it indicates that the current width of the light spot on the roller surface is still within the allowable range, and the transition between light and dark at the edge of the color mark is steep enough to not trigger the subsequent arc segment interception action.

[0052] Specifically, if the elongation of the light spot is greater than the preset length threshold, then the circumferential arc range defined by the measured arc position of the first sampling point and the arc position of the last sampling point is used as the boundary. All sampling points whose actual arc position field falls within the circumferential arc range are selected from the first photoelectric signal sampling sequence with triple index. The photoelectric signals are then formed in ascending order of arc position. Each sampling point in the captured photoelectric signal retains the original speed index, time index and angle index binding relationship.

[0053] Understandably, first-order difference operation directly characterizes the rate of amplitude change between two adjacent sampled values, reflecting the steepness of the photoelectric signal transition from dark to bright or from bright to dark at the edge of the color mark. First-order difference operation is performed point-by-point on the captured photoelectric signal along the rotation direction of the printing roller. Let the amplitude of the k-th sampled point in the captured photoelectric signal be Vk, and the amplitude of the (k+1)-th sampled point be Vk+1. Then, the signal change rate dk at the k-th sampled point satisfies dk = Vk+1 - Vk, where Vk and Vk+1 are both taken from the original amplitude within the captured photoelectric signal, and dk retains its positive or negative sign to distinguish the direction of the bright-dark transition. Further, the signal change rate corresponding to each sampled point is recorded sequentially according to the order of the sampled points in the captured photoelectric signal, forming a change rate sequence arranged along the rotation direction of the printing roller. Each element in the change rate sequence is aligned with the k-th sampled point in the captured photoelectric signal. The sampled point at the end of the sequence does not participate in the difference operation because it lacks a subsequent sampled point.

[0054] Preferably, the rate of change sequence is traversed, and the absolute value of the signal rate of change at each position is taken and compared. The rate of change sequence is then split into a positive rate of change subset and a negative rate of change subset according to the sign of the signal. The position with the largest amplitude in the positive rate of change subset is captured as the extreme position of the incoming edge rate of change, and the position with the largest absolute value in the negative rate of change subset is captured as the extreme position of the outgoing edge rate of change.

[0055] For example, when a color mark moves from a dark area to a bright area of ​​ink, its intensity increases, corresponding to the position of the positive rate of change extreme value; when a color mark moves from a bright area of ​​ink to a dark area of ​​color mark, its intensity decreases, corresponding to the position of the negative rate of change extreme value. The two rate of change extreme value positions together cover the entry edge and exit edge within a single acquisition.

[0056] Specifically, the sampling points corresponding to the extreme positions of the ingress and egress rates of change are traced back, and the angle index and actual arc position field bound to the sampling points are read. The two fields are combined into an edge position index, and the edge position index is written into the additional field of the sampling point corresponding to the truncated photoelectric signal to obtain a second photoelectric signal with an edge position index.

[0057] It is understandable that the second photoelectric signal, while retaining the original amplitude waveform, carries a clear edge position index mark, providing a signal carrier with a position label for the edge positioning of the color mark.

[0058] Step S104: Adaptive interval clustering is performed on the edge fuzziness width according to the curvature of the arc segment corresponding to the second photoelectric signal, and a matching dynamic evaluation threshold is set for each width variation interval in combination with the surface roughness of the printing roller.

[0059] The actual arc segment position corresponding to each sampling point is extracted from the second photoelectric signal. The edge fuzziness width is calculated based on the arc length span between two positions where the absolute value of the signal change rate before and after the sampling point at the edge position index drops to within 10% of the extreme value of the change rate. The actual arc segment position and the edge fuzziness width form a two-dimensional feature vector, resulting in a feature vector set covering the entire arc segment of the second photoelectric signal. For this feature vector set, a K-means clustering algorithm is used to adaptively cluster the feature vectors along the fuzziness width dimension, grouping feature vectors with similar fuzziness widths into the same width variation interval. The mean edge fuzziness width within each width variation interval is taken as the cluster center width of that interval, resulting in several width variation intervals with cluster center widths. The surface roughness value of the printing roller is retrieved. Based on the linear proportional relationship between the roughness value and the threshold bias, the roughness value is multiplied by a preset bias coefficient to obtain the threshold bias. The threshold bias is then superimposed on the cluster center width of each width variation interval to obtain a dynamic evaluation threshold matching that width variation interval.

[0060] In one embodiment, the actual arc segment position field bound to each sampling point is read sequentially from the second photoelectric signal with edge position index. The actual arc segment position field value is stored in the form of the arc length relative to the circumferential positioning origin of the roller surface. The field value increases monotonically as the sampling point advances along the rotation direction of the printing roller. The read actual arc segment positions constitute a position sequence arranged along the rotation direction, providing a position reference for measuring the edge fuzzy width.

[0061] Specifically, the edge blur width reflects the actual widening scale of the bright-dark transition zone under the coverage of the light spot on the edge of the color mark. Taking the sampling point marked by the edge position index in the second photoelectric signal as the center, the absolute value of the signal change rate at the adjacent positions of the sampling point is scanned point by point in both forward and backward directions. Let the absolute value of the signal change rate at the edge position index be dmax, that is, the extreme value of the change rate; scan forward until a sampling point is encountered where the absolute value of the signal change rate first drops to within the value obtained by multiplying dmax by 0.1, and record it as the front boundary point; scan backward until a sampling point is encountered where the absolute value of the signal change rate first drops to within the same threshold, and record it as the back boundary point; measure the absolute value of the difference between the actual arc segment positions bound to the front boundary point and the back boundary point, and use this difference as the edge blur width.

[0062] It should be noted that the above measurement operation is performed on each edge position index within the second photoelectric signal. The measured edge fuzziness width is then bound to the actual arc segment position of the sampling point where the edge is located, and the two form a two-dimensional feature vector. The two-dimensional feature vectors corresponding to all edge position indices are arranged in ascending order of arc segment position, forming a feature vector set covering the entire arc segment of the second photoelectric signal.

[0063] In one possible implementation, for the set of feature vectors, the K-means clustering algorithm is used to perform adaptive interval clustering partitioning along the fuzzy width dimension.

[0064] Understandably, the K-means clustering algorithm receives a set of data to be grouped and divides the data into K clusters with the goal of minimizing the sum of the squared distances from the data within a cluster to the cluster center.

[0065] For example, the edge blur width values ​​in the feature vector set are used as clustering inputs, the number of clusters K is preset to an integer between 3 and 5, and the clustering process is executed in the following order: randomly selecting the initial cluster center, assigning each feature vector to the corresponding cluster according to the nearest distance, recalculating the cluster center as the mean of all edge blur widths in the cluster, and iterating until the change amplitude of the cluster center is less than the preset convergence threshold of 0.001 mm.

[0066] Preferably, after clustering is completed, each cluster corresponds to a width variation interval. The average value of the fuzzy width of all edges within the cluster is used as the cluster center width of the width variation interval. The minimum and maximum values ​​of the fuzzy width of the edges within the cluster are used as the lower and upper boundaries of the width variation interval, resulting in several width variation intervals with cluster center widths.

[0067] Specifically, the surface roughness value of the printing roller is retrieved from the initial parameter set of the roller surface. The roughness value, expressed in micrometers, represents the arithmetic mean deviation of the roller surface profile. Furthermore, the roughness value and the threshold offset amount follow a linear proportional relationship. Let the roughness value be Ra, expressed in micrometers; let the preset offset coefficient be α, a conversion factor with millimeters per micrometer, where α takes a value between 0.05 and 0.15; let the threshold offset amount B, expressed in millimeters, have the same dimension as the arc length of the cluster center width, then B satisfies B = α * Ra.

[0068] For example, when the roughness value Ra is 0.8 micrometers and the offset coefficient α is 0.1 millimeters per micrometer, the threshold offset B is 0.08 millimeters, which is in the same arc length dimension as the cluster center width. The threshold offset is added to the cluster center width for each width variation interval to obtain a dynamic evaluation threshold that matches the width variation interval.

[0069] Understandably, the dynamic evaluation threshold is tiered based on the difference in the cluster center width within the width variation range, and is also biased according to the actual value of the surface roughness of the printing roller, forming a set of thresholds that adapt to the arc ambiguity state and the roller surface state.

[0070] Step S105: Select the dynamic evaluation threshold of the corresponding interval using the edge position index in the second photoelectric signal, perform outlier removal and sliding window smoothing on the second photoelectric signal, and obtain the third photoelectric signal with retained mark edge jumps.

[0071] The edge fuzzy width bound to the sampling point marked by the edge position index is read from the second photoelectric signal. The edge fuzzy width is compared with the lower and upper boundaries of each width change interval. Sampling points whose edge fuzzy width falls between the upper and lower boundaries of a certain width change interval are assigned to that width change interval. The dynamic evaluation threshold corresponding to the width change interval is retrieved, and the corresponding amplitude judgment threshold is calculated according to the ratio of the arc length to the sampling interval in the threshold value, which is used as the amplitude judgment threshold of the sampling point. The second photoelectric signal is traversed along the rotation direction of the printing roller. For each sampling point, the absolute value of the difference between the current amplitude of the sampling point and the median of the amplitudes of the five adjacent sampling points before and after it is taken. If the absolute value of the difference is greater than the retrieved amplitude judgment threshold, the sampling point is determined to be an outlier. The original amplitude of the sampling point is replaced by the median of the amplitudes of the five adjacent sampling points before and after it, and the photoelectric signal after outlier removal is obtained. After removing the outliers, the photoelectric signal is smoothed point by point along the rotation direction of the printing roller using a sliding window. The window width is calculated by dividing the width of the cluster center of the width variation interval to which the corresponding sampling point belongs by the sampling interval and rounding up to obtain the window width in units of the number of sampling points. The average amplitude within the window is used as the smoothed amplitude of the sampling point. At the same time, the original amplitude of the sampling point marked by the edge position index is retained and does not participate in the mean replacement, thus obtaining the third photoelectric signal with the marked edge jump.

[0072] In one implementation, each sampling point marked by an edge position index is located from the second photoelectric signal with edge position index, and the edge blur width value bound to the sampling point is read. The edge blur width, in millimeters, represents the actual arc length of the color mark edge after it has been widened on the arc surface.

[0073] Specifically, the obtained edge fuzzy widths are sequentially compared with the lower and upper boundaries of several width variation intervals obtained from clustering. If the fuzzy width of a certain edge is greater than or equal to the lower boundary of a certain width variation interval and less than or equal to the upper boundary of that width variation interval, then the sampling point marked by the edge position index is determined to belong to that width variation interval. After the assignment is completed, the dynamic evaluation threshold corresponding to the width variation interval is retrieved from a pre-established threshold mapping table. The threshold mapping table stores the one-to-one correspondence between the cluster center width and the dynamic evaluation threshold in the previous step.

[0074] Understandably, the dynamic evaluation threshold uses arc length as its dimension to characterize the allowable edge widening scale; while outlier determination needs to be performed in the amplitude dimension, so the dynamic evaluation threshold needs to be converted into an amplitude determination threshold.

[0075] For example, the conversion process is performed according to the slope relationship between the roller surface arc length and the signal amplitude. Let the dynamic evaluation threshold be G, in millimeters; the sampling interval be Δs, in millimeters, representing the distance between adjacent sampling points in the direction of the roller surface arc length; the extreme value of the rate of change is dmax, with the dimension of amplitude per sampling point, and its value is taken from the maximum absolute value of the difference sequence dk obtained from S103, that is, the maximum rate of change of signal amplitude at the edge position reflected by the unit sampling step size. β is a dimensionless conversion factor, with a value between 0.3 and 0.5, used to convert the sampling point span corresponding to the allowable edge widening scale into the upper limit of the allowable amplitude fluctuation within that span. Usually, it is taken as 0.3 to 0.35 when the edge is relatively steep and the noise is low, and 0.4 to 0.5 when the edge is gentle or there are slight burrs. The conversion process begins by using the ratio of G to Δs to obtain the allowable sampling point span N, i.e., N = G / Δs. Then, N is multiplied by dmax to obtain the theoretical upper limit of amplitude fluctuation within this span without reduction. Finally, a conversion factor β is added to provide a safety margin, resulting in the amplitude judgment threshold Ga, calculated as Ga = N * dmax * β, equivalent to Ga = G / Δs * dmax * β. The priority of operations other than division precedes multiplication is fixed as explained in the previous steps. Ga has the same dimensions as the output amplitude of the photodetector and can be directly compared with the amplitude fluctuation within the intercepted signal.

[0076] In one possible implementation, the second photoelectric signal is traversed along the rotation direction of the printing roller, and outlier determination is performed for each sampling point.

[0077] Specifically, an 11-point neighborhood window is formed by taking the five neighboring sampling points before and after the current sampling point. The median Vmed of the amplitudes of these 11 sampling points within the neighborhood window is calculated. The median is the value in the middle after the amplitudes within the window are sorted by size. Let the amplitude of the current sampling point be Vcur. The absolute value of the difference between Vcur and Vmed is taken as the deviation value D, where D = |Vcur - Vmed|. If D is greater than the amplitude judgment threshold Ga, the current sampling point is determined to be an outlier.

[0078] Preferably, for sampling points identified as outliers, their original amplitude Vcur is directly replaced by the median Vmed in their neighborhood window. After the replacement, the amplitudes of the remaining sampling points in the neighborhood window remain unchanged. After the traversal is completed, the photoelectric signal after the outliers are removed is obtained.

[0079] It should be noted that the sliding window smoothing process is a conventional smoothing method that replaces the amplitude of the sampled point with the average of its neighbors. The sliding window smoothing process is performed point-by-point along the rotation direction of the printing roller on the photoelectric signal after outlier removal. Furthermore, the width of the sliding window is adaptively determined according to the width of the cluster center of the width variation interval to which the current sampled point belongs. Let the cluster center width be Wc in millimeters; let the sampling interval be Δs in millimeters; then the window width Nw, in units of the number of sampled points, satisfies Nw = Wc / Δs rounded up.

[0080] For example, when Wc is 1.0 mm and Δs is 0.05 mm, the window width Nw is 20 sampling points. A sliding window is formed by Nw sampling points centered on the current sampling point, and the arithmetic mean of the amplitudes of all sampling points within the window is used as the smoothed amplitude of the current sampling point. When Nw is even, a "left-more, right-less" bias method is adopted, that is, taking the current sampling point as a reference, taking Nw / 2 sampling points forward and Nw / 2-1 sampling points backward, and including the current sampling point in the previous segment, for a total of Nw sampling points; for the case of Nw=20, the current sampling point and the 9 points in front and the 10 points behind together form the sliding window. When the window touches the starting point or ending point of the trajectory, a mirror extension strategy is adopted to fill in the missing sampling points, that is, with the boundary point as the axis of symmetry, the amplitudes of the existing sampling points in the window are copied in reverse along the trajectory, so that the window always contains Nw valid samples; if Nw is odd, (Nw-1) / 2 sampling points are taken before and after, and no bias is required. After all sampling points are processed, a third photoelectric signal is obtained that retains the edge transition of the mark. The third photoelectric signal retains the original amplitude transition characteristics at the edge position of the color mark while suppressing random fluctuations.

[0081] Step S106: Compare the third photoelectric signal point by point with the selected dynamic evaluation threshold, identify the first set of sampling points that continuously cross the dynamic evaluation threshold, take the starting point time index as the target jump time, and determine the target rotation angle of the target jump point relative to the circumferential positioning origin based on the encoder angle index. Subtract the target jump time from the preset positioning time to obtain the time offset.

[0082] Amplitude comparison is performed point-by-point on the third photoelectric signal along the rotation direction of the printing roller. For each sampling point, an amplitude judgment threshold is calculated based on its width variation range. The smoothed amplitude of the sampling point is compared with the amplitude judgment threshold, and each sampling point is marked as being in a state that has exceeded the amplitude judgment threshold, resulting in a sampling point sequence with the crossing mark. The sampling point sequence with the crossing mark is traversed along the rotation direction of the printing roller, and the first set of sampling points where three or more consecutive sampling points are in a state that has exceeded the amplitude judgment threshold is captured. The sampling point at the beginning of this set is taken as the starting point. The time index bound to the starting point is read as the target jump time. The encoder angle index bound to the starting point is read, and the angle index is converted into an angle value relative to the circumferential positioning origin of the roller surface according to the angle value corresponding to each incremental pulse, thus obtaining the target rotation angle of the target jump point relative to the circumferential positioning origin of the roller surface. Retrieve the preset positioning time from the positioning reference register, perform a time difference calculation between the target jump time and the preset positioning time, take the time difference value obtained by subtracting the two as the time offset, and bind the time offset to the target turning angle for output.

[0083] In one embodiment, the third photoelectric signal with edge abrupt changes along the rotation direction of the printing roller is progressively advanced from the starting sampling point to the ending sampling point, and an amplitude comparison operation is performed for each currently scanned sampling point. During the comparison, the amplitude judgment threshold corresponding to the sampling point marked by the nearest edge position index on the arc segment position of the current sampling point is used as the basis, and the amplitude judgment threshold of the current sampling point is directly adopted as the comparison threshold of the current sampling point. Optionally, a binding is performed before the comparison: each rising edge and falling edge identified on the third photoelectric signal is assigned an edge position index i, where i is from 1 to N, and N is the total number of edges in the arc segment; all sampling points between two adjacent edges constitute a width variation interval Wi, and the amplitude judgment threshold corresponding to this interval is recorded as Ti. Ti has been calibrated according to the mark width in S105, for example, 0.65 times the full width for the wide mark interval and 0.35 times the full width for the narrow mark interval.

[0084] Specifically, the smoothed amplitude value of the sampling point is compared with the amplitude judgment threshold value. If the smoothed amplitude is greater than the amplitude judgment threshold, a 1 is written to the crossover flag field of the sampling point data structure, indicating that the sampling point is in a crossover state; if the smoothed amplitude is not greater than the amplitude judgment threshold, a 0 is written to the crossover flag field, indicating that the sampling point is not crossover. After the point-by-point scanning is completed, each sampling point in the third photoelectric signal is bound to a 0 or 1 crossover flag, forming a sampling point sequence with crossover flags.

[0085] It should be noted that the sampling point sequence with the crossing mark is arranged in ascending order of arc segment position, and each sampling point in the sequence retains the original binding relationship of the three fields: time index, encoder angle index, and crossing mark.

[0086] In one possible implementation, the sequence of sampling points marked with crossings is traversed along the rotation direction of the printing roller, and the first set of sampling points that consecutively crosses the threshold is identified by a continuous counting method.

[0087] It is understandable that when the edge of the color mark actually reaches the detection point, the photoelectric signal exhibits the characteristic of multiple consecutive sampling points being in a state of high amplitude; while the overshoot caused by outliers or residual noise usually only covers a single sampling point.

[0088] For example, a continuous counter is set. During the traversal, if the current sampling point's crossing marker is 1, the continuous counter is incremented by 1, and the current sampling point is written to the temporary sampling point buffer; if it is 0, the continuous counter is cleared, and the temporary sampling point buffer is emptied. When the continuous counter value first reaches 3 or more, all the sampling points accumulated in the temporary sampling point buffer are taken as the first set of sampling points that continuously cross the amplitude determination threshold.

[0089] Preferably, the sampling point located at the foremost position along the rotation direction of the printing roller in the sampling point set is taken as the starting point, the time index field value bound in the starting point data structure is read, and the field value is used as the target jump time. The target jump time is represented in the form of a millisecond-level timestamp.

[0090] Specifically, the encoder angle index field value bound to the starting point data structure is read, and the angle index is stored in the form of incremental pulse accumulation value. Let the total number of pulses output per encoder revolution be N, and the angle corresponding to a single incremental pulse be 2π / N, in radians; let the angle index value bound to the starting point be n. Then, the target rotation angle θt of the target jump point relative to the circumferential positioning origin of the roller surface satisfies θt=n*2π / N, where θt, in radians, represents the angle rotated by the starting point relative to the circumferential positioning origin of the roller surface along the rotation direction of the printing roller.

[0091] It should be noted that the positioning reference register is a dedicated storage unit within the equipment controller. The preset positioning time is calculated in advance based on the theoretical time required for the printing roller to rotate from the circumferential positioning origin to the photoelectric probe detection point at the nominal rotation speed, and is written into the positioning reference register as a reference for comparing jump times. Further, the preset positioning time is retrieved from the positioning reference register. Let the preset positioning time be Tref and the target jump time be Tt. The difference between the two is the time offset Δt, where Δt = Tt - Tref. Δt is a signed value, and the sign represents the leading or lagging relationship between the target jump time and the preset positioning time. The time offset Δt is bound to the target rotation angle θt into a data pair and output to the downstream compensation calculation unit.

[0092] Step S107: Combine the current rotation speed and the printing roller diameter to convert the time offset into the circumferential overprint position offset. According to the opposite direction of the overprint position deviation and combined with the rotation direction of the printing roller, determine the circumferential compensation distance through PID compensation, generate the circumferential position deviation compensation command for the printing plate, and inject the compensated target rotation angle and target jump time into the positioning reference of the next printing cycle.

[0093] The current instantaneous rotational speed of the printing roller is read from the servo driver. The roller diameter is retrieved from the initial parameter set of the roller surface. The circumferential linear velocity of the roller surface is calculated by multiplying half of the roller diameter by the angular velocity corresponding to the instantaneous rotational speed. The time offset is multiplied by the circumferential linear velocity to obtain the circumferential overprinting position offset of the color mark edge from the preset positioning position in the circumferential direction of the roller surface. The opposite number of the circumferential overprinting position offset is taken to obtain the offset in the opposite direction. The offset in the opposite direction is used as the current deviation input for PID compensation calculation. The PID compensation output value is obtained by summing the current deviation, the cumulative deviation, and the change in deviation according to the preset proportional coefficient, the preset integral coefficient, and the preset differential coefficient. The sign of the PID compensation output value is determined according to the convention that the rotation direction of the printing roller is the positive direction, and the circumferential compensation distance is obtained. A circumferential position deviation compensation command for the printing plate is generated based on the circumferential compensation distance and sent to the printing plate roller servo control interface. At the same time, the angle increment is obtained by dividing the circumferential compensation distance by half of the roller diameter, and the time increment is obtained by dividing the circumferential compensation distance by the circumferential linear velocity of the roller surface. The target rotation angle of the current cycle is superimposed with the angle increment, and the target jump time of the current cycle is superimposed with the time increment, and then written together into the positioning reference register of the next printing cycle.

[0094] In one implementation, the current instantaneous rotational speed of the printing roller is read from the rotational speed feedback interface of the servo driver, and the obtained instantaneous rotational speed is given in the form of revolutions per minute; at the same time, the roller diameter value is retrieved from the initial parameter set of the roller surface, and the roller diameter value is in millimeters.

[0095] It is understandable that the instantaneous rotational speed reflects the actual motion state of the roller surface at the moment of data acquisition, and together with the roller diameter, determines the magnitude of the linear velocity of the roller surface in the circumferential direction.

[0096] Specifically, the conversion of the circumferential linear velocity of the roller surface is performed by multiplying half of the roller diameter by the angular velocity corresponding to the instantaneous rotational speed. Let the instantaneous rotational speed be n revolutions per minute (rpm), then the corresponding angular velocity ω satisfies ω = 2πn / 60, in radians per second; the roller diameter is D, in millimeters; then the circumferential linear velocity v of the roller surface satisfies v = ωD / 2, in millimeters per second. Multiplying the time offset Δt by the circumferential linear velocity v yields the arc length value of the edge of the color mark deviating from the preset positioning position along the circumferential direction of the roller surface. This arc length value is used as the circumferential overprinting position offset. The circumferential overprinting position offset is represented by a signed arc length value, the sign inheriting the positive and negative relationship of the time offset: a positive sign indicates that the printing roller is ahead of the preset positioning position along the rotational direction, and a negative sign indicates that it is behind the preset positioning position along the rotational direction.

[0097] It should be noted that the direction of the overprint compensation should be opposite to the direction of the overprint position deviation itself. Therefore, the opposite number operation is performed on the circumferential overprint position offset to obtain the offset in the opposite direction, and the offset in the opposite direction is used as the current deviation input for the PID compensation operation.

[0098] Understandably, PID compensation is a classic closed-loop feedback method in the field of printing equipment control, consisting of the sum of three parts: proportional, integral, and derivative. The proportional term directly reflects the magnitude of the current deviation; the integral term reflects the cumulative effect of historical deviations and can suppress long-term steady-state deviations; the derivative term reflects the rate of change of deviation and can suppress sudden fluctuations in deviation.

[0099] Specifically, let the current deviation of the current printing cycle be e(k), where k is the current printing cycle number; the deviation of the previous printing cycle be e(k-1); the preset proportional coefficient is Kp, the preset integral coefficient is Ki, and the preset differential coefficient is Kd. The proportional term is obtained in the manner of Kpe(k); the deviation accumulation value ΣE is obtained by accumulating the deviations e(j) of each printing cycle from the initial printing cycle j=0 to the current printing cycle j=k, that is, ΣE=e(0)+e(1)+…+e(k); the integral term is obtained in the manner of KiΣE; the differential term is obtained in the manner of Kd*(e(k)-e(k-1)), reflecting the difference between the deviation of the current cycle and the previous cycle. The PID compensation output value u(k) is obtained by summing the three terms of proportional, integral, and differential, that is, u(k)=Kpe(k)+KiΣE+Kd*(e(k)-e(k-1)).

[0100] For example, in a typical overprint compensation scenario, Kp takes values ​​between 0.5 and 1.2, Ki takes values ​​between 0.05 and 0.2, and Kd takes values ​​between 0.01 and 0.05. Further, following the convention that the printing roller rotation direction is the positive direction, the PID compensation output value u(k) is directly used as the circumferential compensation distance Lcom with a direction indicator. A positive Lcom indicates that the compensation action is performed along the printing roller rotation direction, and a negative Lcom indicates that the compensation action is performed in the opposite direction of the printing roller rotation.

[0101] In one possible implementation, a circumferential position deviation compensation command for the printing plate is generated based on the circumferential compensation distance Lcom. The compensation command is encapsulated in a message format that includes the compensation distance value, direction mark and current printing cycle number, and is sent to the printing plate roller servo control interface via fieldbus. After receiving the compensation command, the printing plate roller servo control unit adjusts the real-time phase of the printing plate roller according to the message content.

[0102] Preferably, to synchronously reflect the compensation effect in the positioning reference of the next printing cycle, the circumferential compensation distance Lcom is converted into two components: an angular increment and a time increment. Let half the roller diameter be R, then the angular increment Δθcom satisfies Δθcom = Lcom / R, in radians; let the circumferential linear velocity of the roller surface be v, then the time increment Δtcom satisfies Δtcom = Lcom / v, in seconds.

[0103] Specifically, the target rotation angle θt of the current printing cycle is added to the angle increment Δθcom to obtain the compensated target rotation angle θt′, and the target jump time Tt of the current printing cycle is added to the time increment Δtcom to obtain the compensated target jump time Tt′. θt′ and Tt′ are written together into the positioning reference register of the next printing cycle as an update reference for the jump time comparison of the next printing cycle.

[0104] It is understandable that the compensated target rotation angle θt′ serves as the positioning reference angle at the start of the next printing cycle, and the compensated target jump time Tt′ replaces the original preset positioning time in the next printing cycle, forming a closed-loop compensation link for the circumferential position deviation of the printing plate.

[0105] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for collecting and compensating for positioning information of a printing roller, characterized in that, The method includes: The printing roller diameter, surface roughness, and rotation direction are obtained, and the circumferential positioning origin of the roller surface is established using the encoder zero-position pulse; The first photoelectric signal is continuously acquired in the color mark area, and the rotation speed, time index and encoder angle index are recorded simultaneously. The signal is decomposed into specular reflection component and diffuse reflection component, mapped to the actual arc position and the length of the light spot is determined. When the length of the light spot exceeds the preset length threshold, the photoelectric signal of the corresponding arc segment is intercepted, the adjacent sampled values ​​are differentiated and the extreme position of the rate of change is extracted to generate a second photoelectric signal with an edge position index. The edge fuzzy width is adaptively clustered and divided according to the curvature of the arc segment corresponding to the second photoelectric signal, and a dynamic evaluation threshold is set for each width variation interval in combination with the surface roughness. By using the edge position index to select the dynamic evaluation threshold, outlier removal and sliding window smoothing are performed on the second photoelectric signal to obtain a third photoelectric signal that retains the edge transition. The third photoelectric signal is compared with the dynamic evaluation threshold to identify the first set of sampling points that continuously cross the threshold. The starting point time index is used as the target jump time. The target rotation angle is determined according to the encoder angle index. The time offset is obtained by subtracting the preset positioning time. Based on the time offset and the rotation direction, the circumferential compensation distance is determined, and a circumferential position deviation compensation instruction for the printing plate is generated.

2. The method for collecting and compensating for positioning information of a printing roller according to claim 1, characterized in that, The process of acquiring the printing roller diameter, surface roughness, and rotation direction, and establishing the circumferential positioning origin of the roller surface using the encoder zero-position pulse, includes: The roller diameter is obtained by a contact measuring device, the roller surface roughness parameters are collected by a surface profilometer, and the rotation direction is obtained by reading the feedback signal of the drive motor. The roller surface angle position corresponding to the rising edge of the zero pulse is used as the circumferential positioning origin, and the roller surface arc length is angularly indexed and encoded.

3. The method for collecting and compensating for positioning information of a printing roller according to claim 1, characterized in that, The process of continuously acquiring the first photoelectric signal in the color mark area, simultaneously recording the rotation speed, time index, and encoder angle index, decomposing the signal into specular reflection and diffuse reflection components, mapping it to the actual arc segment position, and determining the elongation length of the light spot includes: Photoelectric probes are deployed in the color mark area to collect the intensity of reflected light to form the first photoelectric signal, and the rotation speed, timestamp and angle index are read simultaneously; the specular reflection component and diffuse reflection component are separated based on the waveform difference between narrow peak high amplitude and wide peak low amplitude and mapped to the actual arc position; the first sampling point and the last sampling point of the switching of the dominant relationship between the two components of specular reflection component and diffuse reflection component are calibrated to determine the length of the light spot elongation.

4. The method for collecting and compensating for positioning information of a printing roller according to claim 3, characterized in that, The first and last sampling points for calibrating the switching of the dominance relationship between the two components are used to determine the beam elongation length, including: The dominance relationship between the specular reflection component and the diffuse reflection component is determined point by point along the rotation direction of the printing roller. Sampling points where the amplitude of the specular reflection component is greater than that of the diffuse reflection component are marked as specular dominance, and sampling points where the amplitude of the diffuse reflection component is greater than that of the specular reflection component are marked as diffuse reflection dominance. The sampling point that switches from specular dominance to diffuse reflection dominance is captured as the first sampling point, and the sampling point that switches back from diffuse reflection dominance to specular dominance is captured as the last sampling point. By tracing back the actual arc position of the first and last sampling points, the elongation length of the light spot is determined.

5. The method for acquiring and compensating for positioning information of a printing roller according to claim 1, characterized in that, When the light spot elongates beyond a preset length threshold, the photoelectric signal of the corresponding arc segment is intercepted, the difference between adjacent sampled values ​​is calculated, and the extreme value position of the rate of change is extracted to generate a second photoelectric signal with an edge position index, including: The length of the light spot is compared with a preset length threshold. If it is greater than the threshold, the circumferential arc range is defined based on the arc positions of the first and last sampling points. The set of sampling points falling within the range is extracted from the first photoelectric signal. The first-order difference is performed on the extracted sampling points to obtain the rate of change sequence. The position of the sampling point with the largest absolute value of the rate of change is captured as the extreme position of the rate of change. The angle index of the extreme position of the rate of change and the actual arc position are recorded as the edge position index.

6. The method for acquiring and compensating for positioning information of a printing roller according to claim 1, characterized in that, The step of adaptively clustering the edge fuzziness width according to the curvature of the arc segment corresponding to the second photoelectric signal, and setting a dynamic evaluation threshold for each width variation interval in combination with surface roughness, includes: The actual arc position and edge ambiguity width of each sampling point are extracted from the second photoelectric signal to form a two-dimensional feature vector. K-means clustering is used to divide the interval along the ambiguity width dimension and the mean value is taken as the cluster center width. The threshold bias is calculated according to the surface roughness and the bias is superimposed on the cluster center width to obtain the dynamic evaluation threshold.

7. The method for acquiring and compensating for positioning information of a printing roller according to claim 1, characterized in that, The process of selecting a dynamic evaluation threshold using edge position indexing, performing outlier removal and sliding window smoothing on the second photoelectric signal, and obtaining a third photoelectric signal that retains edge transitions includes: Read the edge fuzzy width bound to the edge position index sampling point, classify it into the corresponding width change interval according to the upper and lower boundary classification, retrieve the dynamic evaluation threshold of the corresponding width change interval and convert it into the amplitude judgment threshold; traverse the second photoelectric signal along the rotation direction, when the absolute value of the difference between the amplitude of the sampling point and the median of the adjacent sampling points exceeds the threshold, it is judged as an outlier and replaced by the median; perform sliding window smoothing on the signal after removal, the window width is divided by the width of the cluster center of the interval and rounded down, and the original amplitude of the edge position index sampling point is not included in the mean replacement.

8. The method for collecting and compensating for positioning information of a printing roller according to claim 1, characterized in that, The step of comparing the third photoelectric signal with the dynamic evaluation threshold, identifying the first set of sampling points that continuously cross the threshold, using the starting point time index as the target jump time, determining the target rotation angle based on the encoder angle index, and subtracting the preset positioning time to obtain the time offset includes: The smoothed amplitude is compared with the amplitude judgment threshold of the interval and marked as having crossed the threshold. The sampling point sequence with the crossing mark is traversed, and the first set of three or more consecutive sampling points that have crossed the threshold is captured. The first sampling point is taken as the starting point, and the starting point time index and angle index are read to calculate the target rotation angle relative to the circumferential positioning origin.

9. The method for acquiring and compensating for positioning information of a printing roller according to claim 1, characterized in that, The step of determining the circumferential compensation distance based on the time offset and the rotation direction, and generating the circumferential position deviation compensation instruction for the printing plate, includes: The current instantaneous rotation speed of the printing roller is read from the servo driver, the roller diameter is retrieved from the initial parameter set of the roller surface, and the circumferential linear velocity of the roller surface is calculated by multiplying the roller diameter by half the angular velocity corresponding to the instantaneous rotation speed. The time offset is multiplied by the circumferential linear velocity to obtain the circumferential overprinting position offset of the color mark edge from the preset positioning position.

10. The method for acquiring and compensating for positioning information of a printing roller according to claim 9, characterized in that, After multiplying the time offset by the circumferential linear velocity to obtain the circumferential overprint position offset of the color mark edge from the preset positioning position, the process includes: The inverse of the circumferential overprint position offset is used as the current deviation input for the PID controller. The PID compensation output value is obtained by multiplying the current deviation, the cumulative deviation, and the change in deviation by the proportional, integral, and derivative coefficients, respectively. The circumferential compensation distance is determined according to the rotation direction. A compensation command is generated based on the circumferential compensation distance and sent to the servo control interface. The target rotation angle is superimposed with the angle increment, and the target jump moment is superimposed with the time increment, and then written to the positioning reference register of the next cycle.

Images