Method for detecting an electronic tag

Abstract

The present application relates to the technical field of electronic tag detection, and discloses a kind of detection method of electronic tag, comprising: obtaining displacement pulse construction system coordinate system and extracting dynamic correction quantity;Extract label visual feature to screen defect and calculate radio frequency before and after offset;Combining reference counting, correction quantity and before and after offset, calculate radio frequency pulse target point;Dynamic demarcate constrained reading window, extract spatial response center and combine multidimensional cost function to determine label attribution;According to lateral offset and rotation angle, calculate compensation threshold to generate determination result;When comprehensive determination is unqualified, forcibly remove before and after offset to calculate elimination pulse trigger separation action.The present application solves the problem of high-speed string reading of continuous dense label by incorporating multidimensional physical deviation into pulse mapping and combining space demarcation and cost function;According to actual attachment posture, adaptively adjust determination standard to reduce rejection rate, and decouple coordinate system reference in separation stage, to ensure the timing accuracy of elimination action.

Description

Technical Field

[0001] This invention relates to the field of electronic tag detection technology, specifically to a method for detecting electronic tags. Background Technology

[0002] Electronic tags are typically manufactured as continuous, high-speed conveyor belts, requiring online inspection by automated equipment. During high-speed conveyor belt operation, tension causes mechanical deformation, and combined with communication delays in the system hardware, this leads to dynamic drift in the actual physical position of the electronic tag when it reaches the theoretical inspection station. Most existing inspection systems rely on fixed time intervals or trigger signals from a single sensor to perform RFID readings, lacking adaptive compensation mechanisms for dynamic position deviations. This results in the RFID antenna's reading timing failing to synchronize with the tag's actual physical coordinates.

[0003] Because the electronic tags on the tape are densely packed, the radiation range of the radio frequency antenna exhibits physical diffusion and sidelobe effects. When detecting a target tag, the equipment can easily receive radio frequency signals returned by adjacent tags in the queue simultaneously. Current technologies, when processing such dense radio frequency data streams, typically rely solely on the strength of the received signal to determine tag affixation, lacking multi-dimensional cross-verification considering spatial distance and signal phase. This singular evaluation dimension makes the detection equipment prone to cross-reading and misreading in complex radio frequency environments, making it difficult to effectively separate the target tag's true electromagnetic feedback from surrounding interference signals.

[0004] Furthermore, electronic tags inevitably undergo a certain degree of rotation, tilting, or lateral displacement during the processing and attachment stage. These physical attitude changes alter the coupling state between antennas, resulting in an objective attenuation of the radio frequency (RF) signal strength received by the device. Traditional judgment mechanisms often use fixed RF strength thresholds for a one-size-fits-all judgment, ignoring the objective impact of the tag's physical orientation on the electromagnetic signal. This can easily lead to misjudging tags with normal RF functionality but slight attachment deviations as defective products. Simultaneously, when the system triggers subsequent waste separation actions, it often confuses the virtual offset introduced in the earlier RF detection stage with the physical coordinates of the mechanical mechanism. This causes the timing of the rejection mechanism to deviate from the actual target, increasing the risk of mistakenly rejecting adjacent qualified products. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for detecting electronic tags, which solves the problems of position drift caused by belt deformation, signal crosstalk caused by radio frequency interference, and inaccurate system misjudgment and rejection positioning due to tag attachment deviation when detecting continuously densely packed electronic tags at high speed.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting electronic tags includes the following steps: Obtain the displacement pulse sequence to construct the system coordinate system, and obtain the dynamic correction value; When the electronic tag arrives at the visual inspection station, the reference count is latched, a visual image is acquired, and the feature parameters of the visual image are extracted. The radio frequency offset is calculated based on the longitudinal offset contained in the feature parameters. The radio frequency pulse target point is calculated based on the reference count, the dynamic correction amount, the radio frequency offset, the calibration distance, and the system delay. A constrained reading window is defined based on the radio frequency pulse target point and radiation width. Discrete data tuples are obtained within the constrained reading window, and the spatial response center is extracted based on the discrete data tuples. The electronic tag affiliation is determined by combining the cost function. A compensation threshold is calculated based on the electronic tag affiliation and the lateral offset and rotation angle included in the feature parameters. A judgment result is generated by combining the compensation threshold. When the determination result is unqualified, the rejection pulse is calculated using the reference count and the rejection station calibration distance, and a separation action is triggered when the displacement pulse sequence reaches the rejection pulse. The radio frequency offset is removed when the rejection pulse is calculated.

[0007] By extracting visual image feature parameters, the multidimensional deviations generated by physical assembly are reduced in dimension and decoupled, and then applied to the spatial alignment mechanism and the radio frequency (RF) determination mechanism respectively. For the longitudinal offset in the conveyor belt direction, it is converted into a one-dimensional RF offset. Combined with a high-speed displacement pulse sequence, a constrained reading window is dynamically defined. An isolation boundary is established in the spatial domain using the pulse count value. The RF signal reception is forcibly truncated by physical pulses to eliminate cross-response interference caused by the small physical distance between adjacent electronic tags.

[0008] To address energy loss caused by lateral offset and rotation angle, a bounded truncation function is introduced to calculate the compensation threshold, separating electromagnetic attenuation caused by assembly tolerances from chip physical damage to prevent accidental rejection. In subsequent rejection processes, independent physical reference decoupling logic is executed, discarding the RF offset representing the displacement of the electromagnetic response center, allowing the rejection process to re-anchor the actual physical contour boundary of the product, ensuring the spatial accuracy of mechanical separation operations under high-speed conditions.

[0009] Preferably, the step of acquiring the displacement pulse sequence to construct the system coordinate system and acquiring the dynamic correction amount includes: The pulse count value of the encoder is cleared to zero, and the continuously accumulated pulse count values ​​after clearing to zero constitute the system coordinate system; When an interrupt signal is received from the registration marker sensor, the actual pulse count value at the current moment is latched. Extract the difference between the actual pulse count values ​​corresponding to two adjacent interrupt signals, calculate the deviation between the difference between the actual pulse count values ​​and the preset theoretical pulse number, and generate the dynamic correction amount based on the deviation.

[0010] Preferably, the step of calculating the radio frequency offset based on the longitudinal offset included in the feature parameter includes: The longitudinal offset included in the feature parameter is multiplied by the electromagnetic position mapping coefficient, and the product result is superimposed with the one-dimensional electromagnetic center offset constant to calculate the radio frequency offset.

[0011] Preferably, the step of calculating the radio frequency pulse target point based on the reference count, the dynamic correction amount, the radio frequency offset, the calibration distance, and the system delay includes: The reference span pulse number corresponding to the calibration distance is obtained by dynamic belt transport measurement; The pulse output frequency per unit time is multiplied by the system delay to convert it into the number of delay compensation pulses; Divide the radio frequency offset by the pulse equivalent to convert it into the number of offset pulses; The radio frequency pulse target is obtained by adding the reference count, the reference span pulse count, and the delay compensation pulse count to the reference count, and then subtracting the offset pulse count and the dynamic correction amount.

[0012] Preferably, the step of defining the constrained readout window based on the radio frequency pulse target and radiation width includes: The physical width of the constrained reading window is set according to the radiation width and position uncertainty, and the physical width is divided by the pulse equivalent to convert it into a window span pulse; Using the radio frequency pulse target as the spatial center point, half of the window span pulse is symmetrically distributed to the front and back sides to generate the start pulse and end pulse of the constrained reading window, respectively.

[0013] Preferably, the step of acquiring discrete data tuples within the constrained reading window and extracting spatial response centers based on the discrete data tuples includes: The unique identifier, spatial physical location, received signal strength indication value and signal phase angle corresponding to a single read event within the constrained read window are packaged to generate the discrete data tuple, which together constitute the original radio frequency dataset. Extract the global maximum value of the received signal strength indication value, subtract the set attenuation tolerance to generate the sidelobe filtering threshold, remove discrete data tuples in the original radio frequency data set that are lower than the sidelobe filtering threshold, and retain the data to form the main peak effective set; The received signal strength indication value of each discrete data tuple in the effective set of the main peak is converted into the actual radio frequency power value using an exponential transformation relationship. The actual radio frequency power value is used as a weighting coefficient to calculate the weighted average of the spatial physical locations in the effective set of the main peak, thereby obtaining the spatial response center.

[0014] Preferably, the determination of electronic tag affixing using the cost function includes: Construct a multidimensional cost function that integrates spatial distance differences, phase position verification, and envelope fitting errors; For each candidate electronic tag identification code appearing within the constrained reading window, the spatial response center, phase stationary point, and root mean square error of the quadratic curve fitting of the corresponding main peak effective set are extracted and substituted into the multidimensional cost function to calculate the cost evaluation value. Extract the first candidate cost and the second candidate cost, and subtract the first candidate cost from the second candidate cost to obtain the confidence evaluation difference; When the confidence assessment difference is greater than the confidence judgment threshold, it is confirmed that the electronic tag identification code corresponding to the first candidate cost belongs to the electronic tag currently in the queue.

[0015] Preferably, the step of calculating a compensation threshold based on the electronic tag attribution and the lateral offset and rotation angle included in the feature parameters, and generating a judgment result by combining the compensation threshold, includes: The lateral offset and rotation angle of the confirmed current electronic tag are multiplied by the lateral electromagnetic attenuation coefficient and the angular electromagnetic attenuation coefficient, respectively, and substituted into the bounded truncation function. The compensation threshold is calculated in combination with the reference radio frequency qualification threshold. The bounded truncation function is used to limit unlimited compensation when the physical deviation is too large. Extract and confirm the main peak signal strength of the current electronic tag. If the main peak signal strength is greater than or equal to the compensation threshold, the current electronic tag is determined to be radio frequency qualified. If the main peak signal strength is less than the compensation threshold, the current electronic tag radio frequency is determined to be unqualified, and the determination result is generated.

[0016] Preferably, the determination result includes a visual determination result and a radio frequency determination result; When the determination result is unqualified, it also includes: A comprehensive fault-tolerant processing matrix is ​​constructed by fusing the visual determination results with the radio frequency determination results; If either the visual judgment result or the radio frequency judgment result is unqualified, the current electronic tag is determined to be a product that is unqualified in all aspects, marked as an object to be removed, and the separation action is triggered. When the number of continuously identified objects to be removed reaches the continuous scrap threshold, a shutdown alarm mechanism is triggered.

[0017] Preferably, the step of calculating the rejection pulse using the reference count and the calibrated distance of the rejection station includes: Divide the calibrated distance of the rejection station by the pulse equivalent to convert it into the number of rejection span pulses, and multiply the mechanical response time by the real-time output frequency of the encoder to convert it into the number of delayed pulses. The rejection pulse is calculated by adding the base count to the rejection span pulse count, subtracting the delay pulse count and the dynamic correction amount obtained for the segment where the corresponding tag is located.

[0018] This invention provides a method for detecting electronic tags. It has the following beneficial effects: 1. This invention constructs a system coordinate system by acquiring displacement pulse sequences, and extracts visual image features and dynamic correction values ​​when the electronic tag arrives at the detection station. It combines the calibration distance and system delay to comprehensively calculate the radio frequency pulse target point. The label position deviation captured by visual detection, the mechanical deformation and stretching generated during the material belt operation, and the communication processing delay of the hardware system are all incorporated into the algebraic calculation of the position pulse. This effectively eliminates the detection position drift caused by the superposition of multiple factors and improves the synchronous correspondence between the reading and writing actions and the actual physical position of the tag under high-speed belt conveyor conditions.

[0019] 2. This invention defines a constrained reading window around the radio frequency pulse target point, extracts the maximum value of the received signal strength within this interval, filters out side lobes to generate the effective set of the main peak, and combines a multidimensional cost function that integrates spatial distance, phase verification, and envelope fitting error to determine the true affiliation of the electronic tag. This effectively filters out electromagnetic interference and pseudo signals generated by adjacent queued tags, solves the problem of cross-reading and misreading that easily occurs in the radio frequency reading stage of continuously densely arranged electronic tags, and improves the reliability of single tag identification.

[0020] 3. This invention calculates the RF pass / fail compensation threshold by substituting the extracted lateral offset and rotation angle of the electronic tag into a bounded truncation function. When a non-compliance is detected and mechanical separation is triggered, the RF offset introduced in the previous step is removed from the calculation process of the rejection pulse. This allows the system to dynamically adjust the RF intensity evaluation criteria according to the actual attachment posture of the tag, avoiding misjudgment of defective products due to slight physical deviation. At the same time, the virtual coordinate offset caused by the electromagnetic environment is removed in the subsequent defect separation stage, ensuring the accuracy of the mechanical action triggering timing and reducing the false rejection rate and the missed rejection rate. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall physical structure and control architecture of an electronic tag detection system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the overall process of an electronic tag detection method provided in an embodiment of the present invention; Figure 3 This is a sub-flowchart of the dynamic radio frequency pulse target spatial alignment mechanism in an embodiment of the present invention; Figure 4 This is a sub-flowchart illustrating the dynamic delineation of constrained reading windows to establish spatiotemporal isolation boundaries in an embodiment of the present invention. Figure 5 This is a sub-flowchart illustrating the execution of the underlying synchronization logic and signal cleaning algorithm in an embodiment of the present invention; Figure 6 This is a sub-flowchart of the anti-cross-read attribution determination and radio frequency status determination logic in an embodiment of the present invention; Figure 7 This is a schematic diagram comparing the main peak waveform with the traditional unwindowed waveform in the spatial domain in an embodiment of the present invention; Figure 8 This is a schematic diagram comparing the effects of dynamic compensation threshold and traditional fixed judgment threshold under different physical offset parameters in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See attached document Figure 1 The present invention provides an electronic tag detection system, which includes a feeding mechanism, an encoder, a vision inspection station, an RFID inspection station, a rejection station, and a control unit.

[0024] The feeding mechanism is used to drive a continuous strip containing multiple electronic tags to move along a preset direction.

[0025] The encoder is coaxially connected to the feeding mechanism and is used to output a displacement pulse sequence during continuous material belt movement.

[0026] The visual inspection station is located upstream of the continuous conveyor belt movement path and includes an industrial camera and a light source.

[0027] The RF inspection station is located at a predetermined distance downstream of the visual inspection station and includes an RF reader and a near-field RF antenna. The near-field RF antenna is a polarization-confined microstrip antenna or a magnetically coupled antenna. The radiation width of the near-field RF antenna is strictly less than the physical distance between two adjacent electronic tags on the continuous strip.

[0028] The rejection station is located downstream of the RF testing station and includes a pneumatic blowing valve or a mechanical punching device.

[0029] The control unit is communicatively connected to the encoder, vision inspection station, RF inspection station, and rejection station. The control unit includes a low-level hardware processing module and a high-level computing module. The low-level hardware processing module is electrically connected to the encoder's pulse output and the RF reader's general-purpose input / output pins, used to execute microsecond-level interrupt responses and latch displacement pulse sequences. The high-level computing module is used to perform visual image analysis, matrix operations, and comprehensive logic decisions.

[0030] See attached document Figure 2 This invention provides a method for detecting electronic tags, comprising the following steps: S100, the control unit acquires the displacement pulse sequence collected in real time by the encoder to establish the system coordinate system, and obtains the dynamic correction amount according to the registration mark sensor to compensate for the continuous material belt slippage deviation, and establishes an independent data record for each electronic tag entering the detection area; S200, based on the established system coordinate system, when the electronic tag arrives at the visual inspection station, the underlying hardware processing module latches the current encoder's baseline count, the upper computing module obtains the visual image of the electronic tag to screen for physical defects and extract feature parameters, strips the longitudinal offset data from the feature parameters, and calculates the front and rear offset of the electronic tag's predicted radio frequency coupling center through the mapping model. S300, based on the calculated forward and backward offset of the predicted RF coupling center, the preset calibration distance between the visual inspection station and the RF inspection station, and the preset system delay, the control unit calculates the RF pulse target point of the corresponding electronic tag. S400 dynamically defines a constrained reading window based on the calculated RF pulse target point and the preset radiation width of the near-field RF antenna. Within the constrained reading window, it performs main peak clustering and spatial response center extraction based on the data tuples latched by the underlying hardware processing module. It determines the electronic tag affiliation by combining the cost function and calculates the compensation threshold based on the electronic tag affiliation and the lateral offset and rotation angle included in the feature parameters to complete the RF status determination. S500 generates a comprehensive judgment result by combining the visual judgment result. When the visual judgment result or radio frequency judgment result of the electronic tag is unqualified, the control unit calculates the rejection pulse based on the latched reference count and the preset rejection station calibration distance, and triggers the rejection station to perform the separation action when the equipment runs to the rejection pulse. The calculation of the rejection pulse forcibly removes the forward and backward offset of the predicted radio frequency coupling center.

[0031] Based on the method flow described above, this embodiment provides a detailed explanation of the unified coordinate system establishment and sliding compensation mechanism involved in step S100, which is implemented through the following sub-steps: S110, the control unit acquires the displacement pulse sequence collected in real time by the encoder to establish the system coordinate system. In this embodiment, the underlying hardware processing module in the control unit acquires the orthogonal pulse signal output by the incremental photoelectric encoder in real time through a high-speed counting pin. The incremental photoelectric encoder is installed on the end of the measuring wheel shaft of the feeding mechanism, and the measuring wheel is in frictional contact with the surface of the continuous material belt. The movement of the continuous material belt drives the measuring wheel to rotate, and the incremental photoelectric encoder continuously outputs a displacement pulse sequence. In actual physical transmission, the rotation angle of the encoder has a linear mapping relationship with the linear physical displacement of the continuous material belt. The system characterizes the physical position of the material belt in real time by recording the accumulated number of continuous pulses.

[0032] At the start of each production batch, the control unit receives a trigger signal from the origin sensor and resets the pulse count value inside the underlying hardware processing module to zero. The continuously accumulated pulse count value after being reset constitutes a unified position reference scale for the entire system. The control unit calculates the pulse equivalent based on the outer diameter of the measuring wheel and the resolution of the incremental photoelectric encoder. The specific calculation relationship is as follows: assuming the physical diameter of the measuring wheel is... The nominal number of pulses output per revolution of an incremental photoelectric encoder is Then the pulse equivalent Physical diameter With nominal pulse number These are all inherent physical hardware parameters of the equipment, obtained from the equipment's manufacturer's specifications or calibrated on-site using standard gauge blocks. The system coordinate system is a discrete spatial coordinate system with pulse accumulation values ​​as one-dimensional coordinate axes.

[0033] S120, the control unit uses a registration mark sensor to obtain dynamic correction values ​​to compensate for continuous strip slippage. In this embodiment, equipment operation and roll tension fluctuations can easily cause mechanical slippage between the continuous strip and the measuring wheel. In addition, the continuous strip will undergo elastic deformation under different winding and unwinding tensions. This deformation will accumulate errors as the strip advances, causing the physical position of the downstream station to deviate from the theoretical count value of the encoder.

[0034] Therefore, an independent external sensor is introduced for real-time position correction. The registration mark sensor is mounted above the continuous conveyor belt path and employs an industrial color mark sensor or a high-frequency photoelectric switch. The registration mark sensor detects the physical substrate edge of pre-printed positioning eye markers or electronic tags on the continuous conveyor belt and outputs a hardware interrupt signal to the underlying hardware processing module when it detects a sudden change in optical contrast.

[0035] Upon receiving a hardware interrupt signal, the underlying hardware processing module immediately latches the current actual pulse count value. The upper-level calculation module extracts the difference between the actual pulse count values ​​corresponding to two adjacent interrupt signals and compares it with the preset theoretical pulse count. The control unit calculates the difference to generate a dynamic correction amount. Let the theoretical pulse count between two adjacent positioning eye markers be... The difference between two adjacent interrupt latch pulses measured by the underlying hardware processing module is: Then, for the first segment within this section The dynamic correction amount of each electronic tag is Theoretical pulse number Divide the physical design distance between two adjacent positioning eye markers by the pulse equivalent. The resulting dynamic correction is used to overlay the calculation of the execution position of subsequent physical workstations.

[0036] S130, the control unit establishes an independent data record for each electronic tag entering the detection area. In this embodiment, due to the existence of multiple physical workstations and the high speed of the conveyor belt, multiple electronic tags may be distributed in the spatial area between different workstations at the same physical moment. The system manages the concurrent detection data in the form of a queue. Whenever the registration tag sensor detects the arrival signal of an electronic tag, the host computing module dynamically allocates a structure space in the system memory as the independent data record for the current electronic tag.

[0037] Each independent data record contains a unique identifier, an initial tracking pulse, and status flags for each workstation. The unique identifier is generated using an incrementing integer serial number mechanism. The initial tracking pulse is the current pulse count value read in the system coordinate system at the instant the arrival signal is detected. The status flags for each workstation include a visual status register for recording results from visual inspection and an RF status register for recording results from RF inspection; both are initially configured to a non-detected state. Throughout the entire equipment operation cycle, the independent data record is data-bound to its corresponding physical electronic tag to prevent data misalignment during subsequent asynchronous multi-workstation inspections.

[0038] In this embodiment, in conjunction with the aforementioned process of establishing a unified position coordinate system, the visual detection and dimensionality reduction mapping prediction mechanism involved in step S200 specifically includes the following sub-steps: S210: When the electronic tag moves with the continuous conveyor belt to the vision inspection station, the control unit outputs a trigger level to the industrial camera. At the instant the rising edge of the trigger level is generated, the underlying hardware processing module synchronously responds to a hardware interrupt, saving the real-time displacement pulses accumulated by the encoder to the corresponding electronic tag's independent data record, forming a baseline count. Baseline count Representing the The absolute spatial coordinates of the electronic tag at the moment of exposure imaging overcome the delay error caused by the transmission of image data to the upper computing module through network communication.

[0039] S220, the host computer module receives the visual image captured by the industrial camera, performs grayscale conversion and filtering on the image, and performs physical defect screening. In this embodiment, the host computer module uses morphological analysis to detect the chip area and antenna trace connectivity. When a chip is missing or the antenna is disconnected, it determines that the current electronic tag has a fatal defect, writes a visual non-compliance flag into the status flag bit, and stops the subsequent visual feature calculation of the current electronic tag.

[0040] The system sets the input visual tolerance range based on the product process specifications. Visual tolerance range This includes the maximum tolerance for lateral offset and the maximum tolerance for vertical offset. If the current electronic tag does not have a fatal defect, but the physical offset extracted from the image exceeds the visually permissible tolerance range... Similarly, products with severe physical deviations are also judged as visually unqualified, preventing them from entering the subsequent RF compensation judgment stage and avoiding the system from releasing poorly assembled products by lowering the RF judgment threshold.

[0041] S230, for electronic tags that pass the defect screening, the host computing module further extracts visual feature parameters. In this embodiment, the host computing module extracts the preset antenna outline boundary in the image using a contour recognition algorithm, and calculates the diagonal intersection point as the tag reference center; it also locates the physical boundary of the chip die using an edge detection algorithm, and calculates the geometric centroid as the actual center of the chip.

[0042] The host computing module, in conjunction with the camera's pixel physical equivalents, calculates the longitudinal offset of the chip's actual center relative to the tag's reference center. With lateral offset Vertical offset With lateral offset All units are in mm. The host computer module obtains the rotation angle by measuring the angle between the chip edge and the reference horizontal line. The above three parameters are combined to score the image quality. Together they constitute visual feature parameters Image quality score It is calculated by weighting the contrast and sharpness of the image, and the value range is set from 0 to 100.

[0043] S240, the host computing module processes visual feature parameters. Perform dimensionality reduction and decoupling calculations. In actual assembly, changes in chip position cause a physical spatial shift in the energy coupling point of the RF antenna. Lateral offset. With rotation angle The main change is altering the effective overlap area of ​​the electromagnetic coupling, leading to attenuation of the radio frequency signal amplitude; while the longitudinal offset... Parallel to the direction of the material strip movement, the relative timing and spatial position of the electronic tag entering the downstream near-field RF antenna energy radiation zone are altered. Furthermore, due to the asymmetrical distribution of metal wire grids inside the chip, there is a positional deviation between the physical geometric center of the chip's exterior and the actual RF energy coupling center.

[0044] Therefore, the system removes the longitudinal offset from the visual feature parameters. , will vertical offset Substitute the values ​​into the pre-calibrated mapping model to calculate the corresponding radio frequency offset. In this embodiment, the mapping model is calculated using the following formula: ; In the formula, This is the radio frequency offset. The longitudinal offset extracted from visual feature parameters; Preset electromagnetic position mapping coefficients; This is a preset one-dimensional electromagnetic center offset constant caused by the internal structure of the chip. With preset All data were obtained through static scanning calibration of standard samples at the RF station. This dimensionality reduction mapping calculation transforms multidimensional visual-physical biases into single-directional electromagnetic spatial variables, providing a one-dimensional compensation basis for calculating the spatial coordinates of target points at downstream stations.

[0045] See attached document Figure 3 In this embodiment, after the aforementioned visual feature extraction and dimensionality reduction mapping prediction are completed, the dynamic radio frequency pulse target spatial alignment mechanism involved in step S300 specifically includes the following sub-steps: S310, the host computer module obtains the preset calibration distance and preset system delay between the visual inspection station and the radio frequency inspection station. In this embodiment, the calibration distance is defined as the fixed spatial distance between the center of the industrial camera's field of view and the physical center of the near-field radio frequency antenna along the direction of continuous material strip movement. To eliminate assembly errors of mechanical equipment, the calibration distance is obtained by dynamically measuring the material strip by setting a standard reflective target on the continuous material strip.

[0046] Specifically, a continuous strip with a standard reflective target attached is made to run at a fixed speed. The underlying hardware processing module records the pulse count values ​​corresponding to hardware interruptions triggered when the target center passes through the field of view of the industrial camera and the physical center of the near-field radio frequency antenna, and extracts the pulse difference between two adjacent interruptions as the baseline span pulse count. System latency This includes the optocoupler isolation turn-on time, interrupt response time, and internal bus communication link transmission time of the underlying hardware processing module.

[0047] In a continuous conveyor belt operating at high speed, a fixed time delay translates into a physical displacement error in space. The host computer module reads the pulse output frequency of the encoder per unit time. Directly output pulse frequency With system delay time Multiplying them together quantifies the time error into the number of delay compensation pulses. The conversion relationship is as follows: .

[0048] S320, the control unit uses the reference count, the reference span pulse count, the delay compensation pulse count, and the RF offset to perform algebraic superposition calculations to generate the RF pulse target point corresponding to the electronic tag. In this embodiment, the RF offset calculated in step S200... The unit of measurement is mm. The host computer calculation module will calculate the radio frequency offset. Divide by the preset pulse equivalent This is converted into the number of offset pulses in a discrete spatial coordinate system. This is then combined with the dynamic correction amount obtained from the registration marker sensor in step S100. The host computing module constructs the radio frequency pulse target calculation logic for the corresponding electronic tag.

[0049] In this calculation logic, the positive direction along the continuous strip is defined as the positive direction of the spatial coordinate system. When the radio frequency offset... When the algebraic sign is positive, it indicates that the internal chip is offset downstream of the tape relative to the tag reference center, and the physical electronic tag arrives at the RF detection station earlier than the theoretical position. Therefore, the corresponding offset pulse number needs to be subtracted from the total pulse span. In this embodiment, the RF pulse target point is calculated using the following formula: ; In the formula, For the first The radio frequency pulse target point corresponding to each electronic tag; For the first The baseline count of each electronic tag latched at the visual inspection station; The reference span pulse number; The number of pulses is used to compensate for the delay. This is the radio frequency offset. It is the pulse equivalent; In order to target the The dynamic correction amount obtained from the segment where each electronic tag is located.

[0050] S330, the host computing module will calculate the radio frequency pulse target point The data is written into the independent data record of the corresponding electronic tag. In this embodiment, the radio frequency pulse target point is calculated through multi-dimensional algebraic superposition. Representing the The theoretical absolute pulse count value output by the encoder in the system coordinate system is the physical value when the actual RF coupling center of the chip inside the electronic tag is directly below the near-field RF antenna at the RF detection station. In high-speed, continuous industrial scenarios, the physical spacing between adjacent electronic tags on the continuous material strip is extremely small. The energy field of the near-field RF antenna can simultaneously cover and activate multiple electronic tags, thus triggering data crosstalk.

[0051] The control unit converts the physical assembly deviation of a single product identified by the vision station into the position advance or delay of the RF detection station in the spatial coordinate system, and assigns a dynamically changing RF trigger position anchor point to each continuously arranged electronic tag. The spatial alignment mechanism, through hardware-level data binding, provides spatial reference support for subsequently defining constrained isolation reading windows in the time and spatial domains, eliminating electromagnetic interference from adjacent tags, and performing anti-cross-read determination.

[0052] See attached document Figure 4 In this embodiment, based on the radio frequency pulse target points determined in the aforementioned steps, the hardware-level spatiotemporal binding and anti-cross-read attribution determination mechanism involved in step S400 dynamically delineates the constrained reading window through the following sub-steps to establish a spatiotemporal isolation boundary: S410, the host computer module obtains the preset radiation width and position uncertainty of the near-field radio frequency antenna. In this embodiment, the radiation width is defined as the unidirectional physical area spanned by the near-field radio frequency antenna on the continuous strip running plane, enabling the radio frequency reader to successfully demodulate the echo signal of the electronic tag, in mm. The radiation width is determined by the magnetic field distribution characteristics of the antenna design and is measured through static frequency sweep testing of standard electronic tags. The position uncertainty originates from the mechanical vibration of the feeding mechanism, the random minute deformation caused by tension fluctuations, and the discrete sampling error of the incremental photoelectric encoder.

[0053] In this embodiment, the positional uncertainty is calibrated by capturing the continuous material strip's trajectory offline using a high-speed camera and extracting the positional limit deviation. The value range in this embodiment is set to 1mm to 3mm. To establish physical isolation boundaries and prevent overlapping of adjacent tag signals, the constrained reading window's setting range is limited by the physical distance between two adjacent electronic tags on the continuous material strip.

[0054] S420, the host computer module calculates the window span pulse based on the minimum sampling number. In this embodiment, the RFID reader transmits a continuous wave signal at a fixed polling period during continuous reading. To ensure that the complete signal main peak envelope curve can be extracted for spatial response center comparison in subsequent signal processing, the number of times the electronic tag is successfully demodulated within the radiation width cannot be less than the minimum sampling number. The minimum sampling number is set according to the data point threshold required for the clustering algorithm to converge. The host computer module sets the physical width of the constrained reading window based on the radiation width and position uncertainty, divides it by a preset pulse equivalent to convert it into a window span pulse, and uses the following formula to convert it into a window span pulse: ; In the formula, For window span pulses; The width of the radiation; For positional uncertainty; This is the pulse equivalent.

[0055] Simultaneously, the host computing module performs boundary condition checks on the window span pulse. Since the electronic tag needs to be repeatedly charged and demodulated by continuous electromagnetic waves within its radiation width, the physical time corresponding to the window span pulse must cover the total time required for the RF reader to complete the minimum number of sampling polls. On the other hand, to avoid incorrectly including adjacent tags in the current window, the window span pulse must be less than the theoretical distance between adjacent tags minus the antenna radiation margin. In this embodiment, the boundary condition check satisfies the following inequality: ; In the formula, Minimum number of samples; This is the inherent single polling cycle of an RF reader / writer; The real-time running speed of the continuous material belt is dynamically calculated by multiplying the number of pulses accumulated by the encoder per unit time by the pulse equivalent. This represents the theoretical number of pulses between two adjacent electronic tags.

[0056] If the window span pulse exceeds the limit, the system records the abnormality in the alarm register and triggers the equipment to stop.

[0057] S430, the control unit dynamically defines the start and end boundaries of the constrained reading window based on the calculated window span pulse and RF pulse target point. In this embodiment, the host computing module uses the RF pulse target point of the corresponding electronic tag generated in the aforementioned steps as the spatial center point, and symmetrically distributes half of the window span pulse to the front and back sides to generate the start pulse and end pulse of the constrained reading window respectively, thereby constructing a one-dimensional spatiotemporal isolation boundary.

[0058] After calculation, the constrained read window, consisting of the start and end pulses, is written into the comparison register of the underlying hardware processing module as an independent digital isolation interval. When the encoder's real-time displacement pulse enters the digital isolation interval, the underlying hardware processing module enables the reception and latching of signals from the RF reader's general-purpose input / output pins; when the real-time displacement pulse leaves the digital isolation interval, the system shields the corresponding RF read event response. Based on the mechanism of physically forcibly truncating RF signal reception, cross-response interference caused by the extremely small physical distance between adjacent electronic tags is eliminated at the hardware level, completing the spatiotemporal binding of a single electronic tag in a dense array.

[0059] See attached document Figure 5 In this embodiment, after the dynamic delineation of the constrained reading window is completed, the hardware-level spatiotemporal binding and anti-cross-read attribution determination mechanism involved in step S400 also executes the underlying synchronization logic and signal cleaning algorithm through the following sub-steps: S440, within the constrained reading window where the underlying hardware processing module allows open RF signal reception, the system executes underlying synchronization logic to generate discrete data tuples. In this embodiment, during continuous tape operation, the RF reader continuously transmits continuous wave RF energy through a near-field RF antenna. When the electronic tag enters the effective energy radiation area, the chip inside the electronic tag is activated and backscatters the carrier signal carrying the unique identifier.

[0060] Each time the RF reader successfully demodulates the echo signal, it outputs a hardware pulse signal to the underlying hardware processing module via a general-purpose input / output interface. Upon capturing the rising edge of the hardware pulse signal, the underlying hardware processing module immediately triggers a microsecond-level interrupt, latching the pulse count value output by the incremental photoelectric encoder at the current moment as the current spatial physical position. Simultaneously, the host computer module reads the received signal strength indicator and signal phase angle from the RF reader's internal registers via the communication interface. The unit of the received signal strength indicator is dBm, and the unit of the signal phase angle is °.

[0061] The host computing module packages the unique identifier, spatial physical location, received signal strength indicator value, and signal phase angle corresponding to a single read event into a discrete data tuple. Within the entire constrained read window's open period, the same electronic tag generates multiple independent discrete data tuples as it moves physically, collectively forming the current electronic tag's original RF dataset. At the hardware level, electromagnetic demodulation events are directly bound to the encoder's physical space, eliminating time delay errors caused by software communication polling.

[0062] S450, the host computing module sets the sidelobe filtering threshold based on the discrete data tuples and extracts the effective set of the main peak. In this embodiment, due to the radio frequency multipath effect caused by metal reflection in the industrial environment and the inherent sidelobe radiation phenomenon of near-field radio frequency antennas, the original radio frequency data usually contains fluctuation noise or spurious response peaks.

[0063] The host computing module traverses the original RF dataset and extracts the global maximum value of the received signal strength indication (RSI). The host computing module then sets a sidelobe filtering threshold based on the ambient electromagnetic background noise level. The sidelobe filtering threshold is generated by subtracting a preset attenuation tolerance from the global maximum value; in this embodiment, the preset attenuation tolerance ranges from 3dBm to 6dBm. The host computing module removes discrete data tuples from the original RF dataset whose RSI values ​​are below the sidelobe filtering threshold, retaining the data that constitutes the effective set of the main peak. By setting the sidelobe filtering threshold, the system eliminates edge-change data in the antenna sidelobe region, retaining the core response data with concentrated energy distribution and stable physical location within the main lobe region.

[0064] S460, the host computing module performs power weighting calculations within the main peak effective set to extract the spatial response center and phase stationary point. In this embodiment, since the received signal strength indication value is a logarithmic power metric, directly using it for spatial location weighting would introduce nonlinear errors.

[0065] The host computing module uses a preset exponential conversion relationship to convert the received signal strength indication values ​​of each discrete data element within the main peak effective set into actual radio frequency power values ​​with a linear scale. Specifically, the conversion logic involves dividing the received signal strength indication value by 10 to obtain the exponent with a base of 10, thus restoring it to a linear power dimension. The host computing module then uses the converted actual radio frequency power value as a weighting coefficient to perform a weighted average calculation of the spatial physical locations within the main peak effective set to obtain the spatial response center. In this embodiment, the spatial response center is calculated using the following formula: ; In the formula, This represents the pulse count value corresponding to the spatial response center. The total number of discrete data tuples within the effective set of the main peak; For the first The actual RF power value after conversion of discrete data tuples; For the first The spatial physical location pulse count value corresponding to each discrete data tuple.

[0066] Simultaneously, the upper-level computing module performs quadratic curve fitting on the signal phase angle within the effective set of the main peak, extracting the phase stationary points and root mean square error generated during the fitting process. The quadratic curve fitting uses the least squares method to perform quadratic polynomial curve fitting between the spatial physical location within the effective set of the main peak and the signal phase angle. The derivative of the fitted quadratic curve function is calculated and set to zero; the x-coordinates of the calculated extreme points are the position coordinates corresponding to the phase stationary points, and the root mean square error of this fitting process is extracted simultaneously.

[0067] By extracting key nodes where the rate of change reverses within the peak range of physical energy, the system eliminates random deviations caused by electromagnetic fluctuations at single discrete sampling points, obtaining state parameters that truly reflect the physical center of electromagnetic coupling within the electronic tag's chip. The extracted spatial response center, phase stationary point, and root mean square error are simultaneously written into the corresponding independent data record of the electronic tag, providing data support for subsequent multi-station data alignment and comprehensive product qualification assessment.

[0068] See attached document Figure 6 In this embodiment, after extracting the spatial response center and phase stationary point, to prevent sidelobe radiation from causing adjacent tags to cross the boundary and enter the current constrained reading window, step S400 also includes the following sub-steps to confirm the unique ownership of the electronic tag and complete the radio frequency status determination: In S470, the host computing module constructs a multi-dimensional cost function that integrates spatial distance differences, phase position verification, and envelope fitting errors. By comparing the cost and confidence difference of the first candidate, it executes anti-cross-read decision logic. For each candidate electronic tag identification code appearing within the constrained reading window, the host computing module extracts the spatial response center, phase stationary point, and root mean square error of the quadratic curve fitting of the corresponding main peak effective set.

[0069] The multidimensional cost function is composed of a weighted average of the deviations of the above three parameters relative to the current radio frequency pulse target point. In this embodiment, the multidimensional cost function unifies the discrete physical displacement, phase frequency characteristics, and mathematical fit under the same dimension, and is used to objectively evaluate the degree of agreement between the read signal characteristics and the expected theoretical target point.

[0070] The cost assessment value is calculated using the following formula: ; In the formula, The cost assessment value for candidate electronic tag identification codes; The pulse count value corresponding to the spatial response center of the candidate electronic tag identification code; The target point for the radio frequency pulse calculated for the current physical tag; The number of phase stationary pulses for the candidate electronic tag identification code; To smooth the root mean square error during the fitting process; , and These are the maximum tolerance for spatial distance, the maximum tolerance for phase offset, and the maximum tolerance for fitting error preset by the system. By dividing each value by the maximum tolerance of the corresponding term, the values ​​of each term are normalized to the dimensionless range of 0 to 1, thereby giving the allocation of weight coefficients practical mathematical meaning. Weights are based on spatial distance differences. Weights for phase position verification; The weights represent the envelope fitting error. The three weighting coefficients are determined experimentally and are set to values ​​of 0.5 to 0.7, 0.2 to 0.3, and 0.1 to 0.2, respectively, with the sum of the three always being 1.

[0071] After the calculation is completed, the upper-level calculation module sorts the cost evaluation values ​​of all candidate electronic tag identification codes in the window in ascending order, extracts the cost evaluation value ranked first as the first candidate cost, and extracts the cost evaluation value ranked second as the second candidate cost. The upper-level calculation module subtracts the first candidate cost from the second candidate cost to obtain the confidence evaluation difference, and compares the calculated confidence evaluation difference with the preset confidence judgment threshold.

[0072] The confidence threshold is calculated by multiplying the theoretical pulse distance between adjacent physical tags by a fixed safety margin coefficient, reflecting the critical pulse difference that the system can distinguish between similar tags at the current operating speed. When the confidence assessment difference is strictly greater than the confidence threshold, the electronic tag identification code corresponding to the first candidate cost is confirmed to uniquely belong to the currently queued physical electronic tag, completing the anti-cross-read decision; when the confidence assessment difference is less than or equal to the confidence threshold, a severe signal collision that cannot be decoupled has occurred within the current constrained reading window, and the current tag's radio frequency is directly marked as unqualified in the independent data record.

[0073] After completing the anti-interference reading attribution determination, the upper-level computing module of S480 executes electromagnetic decoupling logic on the echo signal confirmed as the current electronic tag to complete the radio frequency status determination. The longitudinal offset in the visual feature parameters has been converted into a one-dimensional electromagnetic spatial variable. With pre-compensation completed, the lateral offset and rotation angle, which are not involved in spatial alignment, will not cause a shift in the signal response center, but they change the effective electromagnetic overlap area between the near-field radio frequency antenna and the internal grid of the electronic tag chip, resulting in electromagnetic energy attenuation. If physical attenuation factors are not eliminated, the system may easily misclassify a qualified product with slight assembly offset but good radio frequency performance as a defective product.

[0074] The host computer module uses only the retained lateral offset and rotation angle, substituting them into a preset bounded truncation function, to calculate the compensation threshold for electromagnetic attenuation. In this embodiment, the preset bounded truncation function satisfies the following calculation logic: ; In the formula, The compensation threshold is expressed in dBm. The preset reference RF pass threshold is the threshold value under ideal, unbiased conditions. This is a horizontal offset; The rotation angle; The preset transverse electromagnetic attenuation coefficient; The preset angular electromagnetic attenuation coefficient; To allow for a maximum compensation limit. The preset lateral electromagnetic attenuation coefficient and angular electromagnetic attenuation coefficient are obtained by changing the lateral and angular orientation of a standard electronic tag in offline mode, measuring the corresponding signal strength attenuation, and performing linear fitting. This is a bounding function that returns the smaller of the two values ​​within the parentheses.

[0075] The boundary comparison operation inside the bounded truncation function restricts unlimited compensation when physical deviations are too large, preventing severely defective products that deviate significantly from specifications from being deemed acceptable. The upper limit of allowable compensation is determined based on the permissible mismatch loss in the antenna design drawings, with a value range set from 1dBm to 3dBm.

[0076] The host computer module extracts and confirms the peak received signal strength indication value of the effective peak of the current electronic tag, defining it as the peak signal strength. The host computer module compares the peak signal strength with a compensation threshold to determine the radio frequency (RF) status: if the peak signal strength is greater than or equal to the compensation threshold, the current electronic tag is deemed RF qualified; if the peak signal strength is less than the compensation threshold, the current electronic tag is deemed RF unqualified, and an RF determination result is generated. The determination result is synchronously updated to the corresponding electronic tag's independent data record. Algebraic logic is used to separate the read / write window misalignment caused by position from electromagnetic parameter attenuation, establishing a multi-dimensional dimensionality reduction compensation mechanism to improve the reliability of determination under complex conveyor conditions.

[0077] In this embodiment, after completing the pre-processing multi-station detection and status determination, step S500 involves an independent physical benchmark decoupling and comprehensive elimination mechanism, which specifically includes the following sub-steps: S510, the host computing module combines the visual judgment result to generate a comprehensive judgment result, specifically including: fusing the visual judgment result and the radio frequency judgment result to construct a comprehensive fault-tolerant processing matrix and executing a shutdown alarm fallback mechanism. In this embodiment, the comprehensive fault-tolerant processing matrix follows a judgment logic of rejection if any one of the criteria is not met: when both the visual judgment result and the radio frequency judgment result are qualified, the current electronic tag is determined to be a comprehensively qualified product and is allowed to be released; when either the visual judgment result or the radio frequency judgment result is unqualified, the current electronic tag is determined to be a comprehensively unqualified product, marked as an object to be rejected, and a separation action is triggered.

[0078] In this embodiment, to prevent the front-end equipment from continuously generating defective products, the upper-level computing module sets a continuous defect counter. When the number of continuously rejected items reaches the continuous defect threshold, the system triggers a shutdown alarm mechanism and sends an emergency stop command to the main control logic controller. In this embodiment, the preset continuous defect threshold is set to 3 to 5.

[0079] S520, the host computer calculation module, based on the visual reference count and the preset rejection station calibration distance, executes the physical reference decoupling logic for discarding RF offsets and calculates the rejection pulse. For electronic tags marked as objects to be rejected, defective products are physically peeled off or destroyed from the continuous conveyor belt at the downstream separation station. In this embodiment, the rejection station calibration distance is defined as the fixed spatial distance between the center of the visual inspection station's field of view and the center of the downstream rejection execution mechanism along the conveyor belt's running direction. To avoid the separation action deviating from the cutting substrate, the rejection pulse calculation is based on the reference count captured by visual inspection and the rejection station calibration distance.

[0080] During the calculation of the rejection pulse, the upper-level calculation module discards the radio frequency offset introduced in the preceding steps. The principle of adopting an independent physical reference decoupling mechanism is as follows: the forward and backward offsets introduced in the preceding steps reflect the displacement of the electromagnetic response center of the chip inside the electronic tag in the antenna field of the RF reader, which is used to align the electromagnetic reading window; while the downstream rejection action is a mechanical cutting or pneumatic blowing action, which acts on the physical outline boundary of the electronic tag and the continuous material strip substrate. If the radio frequency response offset of the internal chip is superimposed on the mechanical action, it will cause the cutter or nozzle to deviate from the physical spacing center between adjacent tags, thereby cutting the adjacent tags or damaging the material strip substrate.

[0081] The system ensures that the mechanical rejection action is always anchored to the physical boundary of the product by separating electromagnetic coordinates and mechanical coordinates. Because the rejection actuator has a mechanical response delay, a level signal must be triggered in advance to compensate for the mechanical lag. Therefore, the number of delayed pulses is subtracted from the total span in the algebraic calculation. In this embodiment, the rejection pulse is calculated using the following formula: ; In the formula, This is the removal pulse corresponding to the tag to be removed; This serves as the baseline count for the corresponding label latched at the visual inspection station; The number of span pulses to be eliminated during the conversion of the station calibration distance is obtained by dividing the elimination station calibration distance by the preset pulse equivalent. To eliminate the number of delayed pulses due to the actuator response time conversion, the delay pulses are obtained by multiplying the preset mechanical response time of the actuator by the real-time output frequency corresponding to the displacement pulse sequence. The preset mechanical response time is calibrated according to the hardware parameters of the pneumatic valve or electromagnetic relay. In this embodiment, the value range can be set from 10ms to 30ms. This is the dynamic correction amount obtained for the segment where the corresponding label is located.

[0082] S530, the underlying hardware processing module compares the real-time displacement pulse with the rejection pulse, triggering the downstream actuator to complete the timing control process. In this embodiment, the upper-level calculation module sends the calculated rejection pulse to the action comparison register of the underlying hardware processing module. As the continuous material belt runs, the incremental photoelectric encoder continuously outputs real-time displacement pulses. The underlying hardware processing module compares the real-time displacement pulse count value with the rejection pulse in the action comparison register at the hardware circuit level.

[0083] When the encoder's real-time displacement pulse reaches the calculated rejection pulse, the general-purpose output pin of the underlying hardware processing module outputs a level-to-flip signal. This level-to-flip signal drives an external high-speed solid-state relay to close, thereby triggering a downstream pneumatic valve or punching device to perform a stripping or breaking action. Because the triggering process is independently completed by the underlying hardware comparator, the task scheduling delay caused by communication between the operating system and the software bus is eliminated, ensuring the consistency of the spatial position of mechanical movements under high-speed conveyor conditions.

[0084] Specific application examples: The operating parameters of a high-speed electronic tag detection device are set as follows: continuous conveyor belt speed. m / s, measuring wheel diameter mm, incremental photoelectric encoder resolution PPR. The pulse equivalent can be calculated. mm / pulse, current encoder output frequency Hz. The theoretical physical spacing between adjacent tags is 20mm, which translates to a theoretical pulse count. pulse.

[0085] The registration mark sensor detects adjacent positioning eye markers on a continuous strip, and the pulse difference is latched by the underlying hardware. pulse.

[0086] Extract the dynamic correction amount based on the formula: Substitute the values: pulse. When the When an electronic tag arrives at the visual inspection station, the camera triggers and latches the baseline count. pulse.

[0087] Visual image feature extraction parameter: vertical offset mm, lateral offset mm, rotation angle According to offline calibration, the electromagnetic position mapping coefficient One-dimensional electromagnetic center offset constant mm. Calculate the radio frequency offset using the formula: Substitute the values: mm.

[0088] The calibrated distance between the vision station and the RF station is 100mm, corresponding to the reference span pulse count. pulse. System hardware communication and interrupt latency. ms, number of delay compensation pulses Pulse. Calculate the radio frequency pulse target point according to the formula: Substitute the values: pulse.

[0089] Calibrate the radiation width of the near-field radio frequency antenna mm, position uncertainty Calculate the window span pulse according to the formula: Substitute the values: pulse.

[0090] The start and end boundaries of the constrained reading window are: [13302-286, 13302+286], that is, the interval [13016, 13588].

[0091] Upon entering this window, the system extracts the main peak signal. The currently measured peak received signal strength is the main peak signal strength -42.5 dBm. Set the baseline acceptable threshold. dBm, attenuation coefficient dBm / mm, dBm / ∘, maximum allowable value dBm.

[0092] Calculate the compensation threshold using the formula: Substituting the values: the attenuation is 1.5 × |0.5| + 0.5 × |2| = 1.75 dBm, which is less than the upper limit of 2 dBm. Compensation threshold. dBm.

[0093] Because the actual read intensity was -42.5dBm < -41.75dBm, the tag was deemed to have RF non-compliance and was marked as a rejection item. The calibration distance between the vision station and the rejection station was 300mm, and the span pulse... Pulse. The pneumatic actuator has a response time of 20ms, with a delayed pulse. pulse.

[0094] Calculate the rejection pulse according to the formula: Substitute the values: pulse.

[0095] Experimental verification and effect comparison: To verify the engineering effectiveness of the proposed anti-interference and dynamic dimensionality reduction compensation mechanisms, the detection data of the traditional fixed-time-window method and the multi-dimensional pulse mapping method of this invention were compared under the same hardware platform and high-speed conveyor conditions. The test samples used continuously arranged electronic tags, and random physical position assembly deviations were artificially introduced.

[0096] In the radio frequency signal acquisition stage, traditional methods employ broad continuous readings, which easily capture sidelobe signals from adjacent tags. Combined with... Figure 7 It is known that when the physical distance between adjacent tags is small, traditional reading methods are prone to data overlap and cross-reading, resulting in the fusion of signal envelopes from two adjacent reading events. This invention transforms visual offset into spatial target points and constrains start and stop pulses, resulting in a steep waveform truncation of the measured discrete data tuples in the spatial domain. The energy concentration of the main peak effective set is improved compared to traditional algorithms, reducing the probability of cross-reading.

[0097] When analyzing the impact of assembly deviations on radio frequency (RF) strength, using a traditional fixed strength threshold for judgment can easily lead to labels with large lateral offsets, even when the RF chip is actually functioning correctly, being misjudged as defective, resulting in a high rejection rate. For example... Figure 8 As shown, after the present invention introduces a bounded truncation function to calculate the adaptive compensation threshold, it determines that the standard surface exhibits a step-like dynamic downward trend, decoupling the electromagnetic attenuation caused by assembly tolerances from the actual physical damage to the chip, thereby effectively reducing the overall failure rate of the system without increasing the missed detection rate.

[0098] For pneumatic rejection triggered by labels marked as non-compliant, the rejection calculation mechanism of this invention eliminates the electromagnetic position offset during the RF reading stage and directly anchors to the initial physical reference of visual inspection. The actual positional deviation between the action center of the rejection actuator and the physical center of the conveyor belt is controlled within a limited range. Compared to traditional systems without coordinate system decoupling control, which are prone to cutting adjacent compliant labels, this solution maintains mechanical synchronization stability under high-speed conditions.

Claims

1. A method for detecting electronic tags, characterized in that, Includes the following steps: Obtain the displacement pulse sequence to construct the system coordinate system, and obtain the dynamic correction value; When the electronic tag arrives at the visual inspection station, the reference count is latched, a visual image is acquired, and the feature parameters of the visual image are extracted. The radio frequency offset is calculated based on the longitudinal offset contained in the feature parameters. The radio frequency pulse target point is calculated based on the reference count, the dynamic correction amount, the radio frequency offset, the preset calibration distance, and the preset system delay. A constrained reading window is defined based on the radio frequency pulse target and the preset radiation width. Discrete data tuples are obtained within the constrained reading window, and the spatial response center is extracted based on the discrete data tuples. The electronic tag affiliation is determined by combining the multidimensional cost function. A compensation threshold is calculated based on the electronic tag affiliation and the lateral offset and rotation angle included in the feature parameters. A radio frequency judgment result is generated by combining the compensation threshold, and a comprehensive judgment result is generated by combining the visual judgment result. When the overall judgment result is unqualified, the rejection pulse is calculated using the reference count and the preset rejection station calibration distance, and the separation action is triggered when the displacement pulse sequence reaches the rejection pulse.

2. The method for detecting an electronic tag according to claim 1, characterized in that, The process of acquiring the displacement pulse sequence to construct the system coordinate system and obtaining the dynamic correction amount includes: The pulse count value of the encoder is cleared to zero, and the continuously accumulated pulse count values ​​after clearing to zero constitute the system coordinate system; When an interrupt signal is received from the registration marker sensor, the actual pulse count value at the current moment is latched. Extract the difference between the actual pulse count values ​​corresponding to two adjacent interrupt signals, calculate the deviation between the difference between the actual pulse count values ​​and the preset theoretical pulse number, and generate the dynamic correction amount based on the deviation.

3. The method for detecting an electronic tag according to claim 1, characterized in that, The step of calculating the radio frequency offset based on the longitudinal offset included in the feature parameters includes: The longitudinal offset included in the feature parameter is multiplied by a preset electromagnetic position mapping coefficient, and the product result is superimposed with a preset one-dimensional electromagnetic center offset constant to calculate the radio frequency offset.

4. The method for detecting an electronic tag according to claim 1, characterized in that, The step of calculating the radio frequency pulse target point based on the reference count, the dynamic correction amount, the radio frequency offset, the preset calibration distance, and the preset system delay includes: The reference span pulse number corresponding to the calibration distance is obtained by dynamic belt transport measurement; The pulse output frequency per unit time is multiplied by the system delay to convert it into the number of delay compensation pulses; Divide the radio frequency offset by a preset pulse equivalent to convert it into the number of offset pulses; The radio frequency pulse target is obtained by adding the reference count, the reference span pulse count, and the delay compensation pulse count to the reference count, and then subtracting the offset pulse count and the dynamic correction amount.

5. The method for detecting an electronic tag according to claim 1, characterized in that, The step of defining the constrained readout window based on the radio frequency pulse target and the preset radiation width includes: The physical width of the constrained reading window is set according to the radiation width and the preset position uncertainty, and the physical width is divided by the preset pulse equivalent to convert it into a window span pulse; Using the radio frequency pulse target as the spatial center point, half of the window span pulse is symmetrically distributed to the front and back sides to generate the start pulse and end pulse of the constrained reading window, respectively.

6. The method for detecting an electronic tag according to claim 1, characterized in that, The step of acquiring discrete data tuples within the constrained reading window and extracting spatial response centers based on the discrete data tuples includes: The unique identifier, spatial physical location, received signal strength indication value and signal phase angle corresponding to a single read event within the constrained read window are packaged to generate the discrete data tuple, which together constitute the original radio frequency dataset. Extract the global maximum value of the received signal strength indication value, subtract the preset attenuation tolerance to generate the sidelobe filtering threshold, remove discrete data tuples in the original radio frequency data set that are lower than the sidelobe filtering threshold, and retain the data to form the main peak effective set; The received signal strength indication value of each discrete data element in the effective set of the main peak is converted into the actual radio frequency power value using a preset exponential conversion relationship. The actual radio frequency power value is used as a weighting coefficient to perform a weighted average calculation on the spatial physical location in the effective set of the main peak to obtain the spatial response center.

7. The method for detecting an electronic tag according to claim 6, characterized in that, The method of determining electronic tag attribution by combining multidimensional cost functions includes: Construct a multidimensional cost function that integrates spatial distance differences, phase position verification, and envelope fitting errors; For each candidate electronic tag identification code appearing in the constrained reading window, the spatial response center of the corresponding main peak effective set is extracted, and the signal phase angle in the main peak effective set is subjected to quadratic curve fitting. The phase stationary point and root mean square error generated by the fitting are extracted, and the spatial response center, the phase stationary point and the root mean square error are substituted into the multidimensional cost function to calculate the cost evaluation value. The cost evaluation values ​​are sorted, the first candidate cost and the second candidate cost are extracted, and the confidence evaluation difference is obtained by subtracting the first candidate cost from the second candidate cost. When the confidence assessment difference is greater than the preset confidence judgment threshold, it is confirmed that the electronic tag identification code corresponding to the first candidate cost belongs to the electronic tag currently in the queue.

8. The method for detecting an electronic tag according to claim 7, characterized in that, The step of calculating a compensation threshold based on the electronic tag attribution and the lateral offset and rotation angle included in the feature parameters, and generating an RF determination result by combining the compensation threshold, includes: The lateral offset and rotation angle of the confirmed current electronic tag are multiplied by a preset lateral electromagnetic attenuation coefficient and a preset angular electromagnetic attenuation coefficient, respectively, and substituted into a preset bounded truncation function. The compensation threshold is calculated by combining the preset reference radio frequency qualification threshold. The bounded truncation function is used to limit unlimited compensation when the physical deviation is too large. Extract and confirm the main peak signal strength of the current electronic tag. If the main peak signal strength is greater than or equal to the compensation threshold, the current electronic tag is determined to be radio frequency qualified. If the main peak signal strength is less than the compensation threshold, the current electronic tag is determined to be unqualified in radio frequency, and the radio frequency determination result is generated.

9. The method for detecting an electronic tag according to claim 1, characterized in that, The process of generating a comprehensive judgment result by combining the visual judgment results includes: A comprehensive fault-tolerant processing matrix is ​​constructed by fusing the visual judgment results with the radio frequency judgment results; If either the visual judgment result or the radio frequency judgment result is unqualified, the current electronic tag is determined to be a product that is unqualified in all aspects, marked as an object to be removed, and the separation action is triggered. When the number of continuously identified objects to be rejected reaches the preset continuous scrap threshold, a shutdown alarm mechanism is triggered.

10. The method for detecting an electronic tag according to claim 1, characterized in that, The calculation of the rejection pulse using the reference count and the preset rejection station calibration distance includes: Divide the calibration distance of the rejection station by the preset pulse equivalent to convert it into the rejection span pulse number, and multiply the preset mechanical response time by the real-time output frequency corresponding to the displacement pulse sequence to convert it into the delay pulse number. The rejection pulse is calculated by adding the base count to the rejection span pulse count, subtracting the delay pulse count and the dynamic correction amount obtained for the segment where the corresponding tag is located.

Images