Multi-channel parallel film tape identification and distribution method and system

Abstract

The invention relates to the technical field of multi-channel film tape bar code recognition, in particular to a multi-channel parallel film tape recognition and distribution method and system. A dynamic monitoring unit collects the instantaneous tape running speed fluctuation quantity and the real-time tension change value of each channel film tape in real time; the self-adaptive calibration unit generates transverse offset and longitudinal distortion compensation factors based on a pre-training distortion influence model, reduces a feature point matching threshold after superposition and compensates the influence of physical distortion on the recognition precision, and the stability evaluation unit calculates a fluctuation index through a sliding time window. The intelligent distribution decision-making unit analyzes the bar code quality evaluation information, combines the channel stability index and the load, calculates the comprehensive distribution weight by taking the stability as the main weight, selects the channel distribution film tape with the highest weight, and forms closed-loop control; and the multi-channel identification accuracy and the distribution efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of multi-channel film tape barcode recognition technology, and more specifically, to a multi-channel parallel film tape recognition and allocation method and system. Background Technology

[0002] Multi-channel film tape barcode recognition technology is an important technology, specifically applied to the barcode recognition and allocation process of multi-channel parallel film tapes. Its core principle is to achieve precise matching between barcodes and channels by compensating for the impact of physical fluctuations and quantifying channel stability. This meets the core requirements of film tape recognition for efficiency and accuracy. When recognizing film tapes in parallel across multiple channels, the instantaneous speed of the film tape naturally fluctuates, and the real-time tension also changes with the tape's movement. These two physical fluctuations cause lateral offset and longitudinal distortion in barcode sampling, affecting the accuracy of feature point matching and resulting in differences in the recognition stability of each channel. Furthermore, the quality of barcodes on the film tapes to be recognized varies. If channels are allocated based on only one dimension, lower-quality barcodes will be assigned to unstable channels, further reducing the recognition success rate. To address this technical problem, we provide a multi-channel parallel film tape recognition and allocation method and system. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-channel parallel film tape identification and allocation method and system to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, one objective of this invention is to provide a multi-channel parallel film tape identification and allocation method, comprising the following steps: S1. Real-time monitoring of the instantaneous belt speed fluctuation and real-time tension change value of the film belt in each identification channel; S2. Based on the instantaneous belt speed fluctuation and real-time tension change, the processor calibrates the matching threshold of the feature points used for barcode recognition in this channel. The calibration direction of the matching threshold is as follows: The value is adjusted downward as the instantaneous belt speed fluctuation and real-time tension change increase, in order to compensate for the impact of the sampling lateral offset caused by the instantaneous belt speed fluctuation and the barcode longitudinal distortion caused by the real-time tension change on the feature point matching accuracy. S3. Calculate the fluctuation degree index of the instantaneous belt speed fluctuation of each channel and the fluctuation degree index of the real-time tension change value of each channel. Finally, convert the fluctuation degree index of the instantaneous belt speed fluctuation and the fluctuation degree index of the real-time tension change value into a channel stability index that characterizes the channel identification stability. S4. When there is a film strip to be identified that needs to be assigned to the identification channel, obtain the barcode quality assessment information of the film strip to be identified, combine the channel stability index of each channel and the number of film strips to be processed in each channel, perform weighted calculation with the channel stability index as the main weight factor, generate the comprehensive allocation weight of each channel, and select the channel with the highest comprehensive allocation weight to assign the film strip to be identified.

[0005] The second objective of this invention is to provide a system for implementing a multi-channel parallel film tape identification and allocation method as described in any one of the above-mentioned methods, comprising: The dynamic monitoring unit is deployed in the sensor network of each identification channel to collect the instantaneous speed fluctuation and real-time tension change of the film tape in real time. The adaptive calibration unit has a built-in pre-trained distortion effect model and nonlinear mapping algorithm. Based on the instantaneous belt speed fluctuation and real-time tension change, it generates a lateral offset compensation factor and a longitudinal distortion compensation factor. After superposition, it outputs a real-time lowering instruction for the matching threshold, driving the barcode recognition to adjust the matching threshold of the feature points. The stability assessment unit calculates the variance index of instantaneous belt speed fluctuation and the range index of real-time tension change through a sliding time window. After normalization and linear combination of preset weight coefficients, a channel stability index negatively correlated with channel stability is generated. The intelligent allocation decision unit generates a comprehensive quality assessment value by parsing the barcode quality assessment information of the film tape to be identified. It calls the channel stability index and combines it with the current load of each channel. It performs weighted fusion calculation with the stability index as the main weight factor to generate a comprehensive allocation weight. It selects the channel with the highest comprehensive allocation weight and binds the film tape to be identified through a distributed queue to form a closed-loop control link.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses a dynamic monitoring unit to capture the instantaneous speed fluctuations and real-time tension changes of the film tape in each recognition channel in real time, laying a data foundation for subsequent accurate calibration. The adaptive calibration unit relies on a pre-trained distortion impact model to generate lateral offset compensation factors and longitudinal distortion compensation factors, lowering the feature point matching threshold to effectively offset the lateral offset of sampling caused by speed fluctuations and the longitudinal distortion of the barcode caused by tension changes, ensuring stable barcode recognition accuracy under physical fluctuations. The stability evaluation unit uses a sliding time window to statistically analyze fluctuation indicators, and generates a channel stability index through normalization and linear combination, objectively quantifying the recognition stability of each channel. The intelligent allocation decision unit combines the barcode quality assessment information of the film tape to be recognized, the channel stability index, and the current load, and calculates the comprehensive allocation weight with stability as the main weight. When the barcode quality is poor, the stability weight ratio is automatically increased, and the barcode is preferentially allocated to a stable channel, solving the problems of physical fluctuation interference and improper channel allocation, and significantly improving the accuracy and efficiency of multi-channel parallel film tape recognition. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating the overall workflow of the present invention; Figure 2 This is a schematic diagram of the overall structure of the present invention; The meanings of the labels in the diagram are as follows: 1. Dynamic monitoring unit; 2. Adaptive calibration unit; 3. Stability assessment unit; 4. Intelligent allocation decision unit. Detailed Implementation

[0008] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0009] Please see Figure 1 As shown, one of the objectives of this embodiment is to provide a multi-channel parallel film tape identification and allocation method, including the following steps: S1. Real-time monitoring of the instantaneous belt speed fluctuation and real-time tension change value of the film belt in each identification channel; S2. Based on the instantaneous belt speed fluctuation and real-time tension change, the processor calibrates the matching threshold of the feature points used for barcode recognition in this channel. The calibration direction of the matching threshold is as follows: The value is adjusted downward as the instantaneous belt speed fluctuation and real-time tension change increase, in order to compensate for the impact of the sampling lateral offset caused by the instantaneous belt speed fluctuation and the barcode longitudinal distortion caused by the real-time tension change on the feature point matching accuracy. S3. Calculate the fluctuation degree index of the instantaneous belt speed fluctuation of each channel and the fluctuation degree index of the real-time tension change value of each channel. Finally, convert the fluctuation degree index of the instantaneous belt speed fluctuation and the fluctuation degree index of the real-time tension change value into a channel stability index that characterizes the channel identification stability. S4. When there is a film strip to be identified that needs to be assigned to the identification channel, obtain the barcode quality assessment information of the film strip to be identified, combine the channel stability index of each channel and the number of film strips to be processed in each channel, perform weighted calculation with the channel stability index as the main weight factor, generate the comprehensive allocation weight of each channel, and select the channel with the highest comprehensive allocation weight to assign the film strip to be identified.

[0010] The processor's dynamic parameter adjustment module captures continuous sampling data of instantaneous belt speed fluctuations and real-time tension changes in real time. A nonlinear mapping algorithm is used to synchronously associate the two input values ​​with the calibration coefficient of the matching threshold. The processor has a built-in adaptive calibration engine that uses a pre-trained distortion influence model to superimpose the lateral offset compensation factor corresponding to the instantaneous belt speed fluctuations and the longitudinal distortion compensation factor corresponding to the real-time tension changes, generating a real-time downward adjustment instruction for the matching threshold. Finally, the barcode recognition system calls the updated matching threshold to perform feature point matching operations.

[0011] The calibration direction for the matching threshold is limited to: When the instantaneous belt speed fluctuation increases, the adjustment range of the matching threshold is positively correlated with the increment of the instantaneous belt speed fluctuation. When the real-time tension change value increases, the adjustment range of the matching threshold is positively correlated with the increment of the real-time tension change value. Furthermore, the adjustment of the matching threshold by the two is achieved by fusion after the parallel compensation channel calculation of the adaptive calibration engine.

[0012] By narrowing the lateral tolerance range of feature point matching through the lateral offset compensation factor, the pixel displacement error in the direction of film movement is offset. By widening the longitudinal deformation tolerance of feature point matching through the longitudinal distortion compensation factor, the barcode recognition adapts to the elastic stretching deformation of the barcode unit. The compensation effect of the two is adjusted by the weighted output of the parallel compensation channel to adjust the matching threshold, so that the feature point matching accuracy remains stable under physical distortion conditions.

[0013] When calculating the fluctuation index of the instantaneous tape speed fluctuation of each channel, a sliding time window is used to statistically analyze the instantaneous tape speed fluctuation data continuously monitored in each channel, and the variance is extracted as the first fluctuation quantification value. When calculating the fluctuation index of the real-time tension change value of each channel, the range of the real-time tension change value is statistically analyzed using the same time window as the second fluctuation quantification value. The width of the sliding time window is set in relation to the recognition time of a single frame of the film tape.

[0014] After normalizing the first and second fluctuation quantization values ​​into dimensionless parameters, they are input into a preset index fusion model for linear combination. The normalization result of the first fluctuation quantization value is assigned to the first weight coefficient, and the normalization result of the second fluctuation quantization value is assigned to the second weight coefficient. The first weight coefficient is greater than the second weight coefficient. The final output channel stability index is negatively correlated with the linear combination result. The index fusion model is executed by the stability evaluation module built into the processor.

[0015] The barcode quality assessment information includes barcode edge ambiguity score, barcode cell wear level, and feature point contrast. Among them, barcode edge ambiguity score and barcode cell wear level are the main indicators of quality deterioration, and feature point contrast is the auxiliary indicator of quality. The three are weighted to generate a comprehensive quality assessment value.

[0016] When using the channel stability index as the primary weighting factor for weighted calculations, the specific steps include: The channel stability index is converted into the channel stability weight base value by taking the reciprocal, and the number of film strips to be processed in each channel is converted into the channel load coefficient. Finally, the channels are weighted and superimposed according to the preset rules that the stability weight base value accounts for a significant proportion and the channel load coefficient accounts for a secondary proportion to generate the comprehensive allocation weight of each channel. When the comprehensive quality assessment value is lower than the set degradation threshold, the proportion of the channel stability weight base value is automatically increased.

[0017] When selecting the channel with the highest overall allocation weight to assign the film tape to be identified, the distributed task scheduler binds the film tape to be identified to the identification queue of the selected channel. At the same time, it synchronizes the latest instantaneous tape speed fluctuation and real-time tension change value of the channel with the adaptive calibration engine to achieve threshold pre-calibration, requests the stability evaluation module to update the stability index of the channel, and finally sends the channel load increment signal to refresh the overall allocation weight, forming a closed-loop control link.

[0018] Further explanation is needed: to address the issues of low recognition accuracy and unreasonable allocation caused by film tape state fluctuations and barcode quality differences in multi-channel recognition, this method achieves accurate recognition and efficient allocation through a closed-loop logic of real-time monitoring, threshold calibration, stability assessment, and intelligent allocation. The specific implementation method is as follows: The primary step in multi-channel parallel film tape recognition and allocation is real-time monitoring of the instantaneous speed fluctuation and real-time tension change of the film tape in each recognition channel. The instantaneous speed fluctuation refers to the difference between the real-time film tape speed and a preset standard speed, while the real-time tension change refers to the deviation between the current tension on the film tape and a reference tension. Monitoring is achieved through a sensor network deployed in each channel. A laser velocimeter is used to capture the film tape's movement speed in real time, and a contact tension sensor is used to synchronously collect tension data. The raw data collected by the sensors undergoes a low-pass filtering algorithm to remove environmental interference, and then the timestamps are aligned through a data synchronization module to ensure a one-to-one correspondence between speed and tension data at the same moment. This provides accurate input for subsequent threshold calibration. After monitoring the data, based on the instantaneous belt speed fluctuation and real-time tension change, the processor calibrates the matching threshold of the feature points used for barcode recognition in this channel. The feature point matching threshold is the critical grayscale difference value used to determine whether a pixel is a target feature point during barcode recognition. The default value is 30. If the grayscale difference is higher than the threshold, it is determined to be a feature point. The core calibration logic is to dynamically adjust the threshold to offset the impact of physical fluctuations on recognition accuracy. The calibration direction of the matching threshold is clearly to decrease it as the instantaneous belt speed fluctuation and real-time tension change increase. This is because the instantaneous belt speed fluctuation will cause lateral shift in sampling. When the film belt movement speed is unstable, the film belt shifts laterally at the moment of camera sampling, causing the barcode feature points to shift laterally in the image, which needs to be adjusted downwards. The threshold narrows the lateral tolerance range to compensate for the error, while real-time tension changes cause longitudinal distortion of the barcode. When the tension increases, the film stretches, and the barcode unit elastically deforms along the longitudinal direction. The threshold needs to be lowered to relax the longitudinal deformation tolerance to accommodate this deformation. The calibration process is led by the processor's dynamic parameter adjustment module, which captures continuous sampling data of speed and tension in real time. It uses a nonlinear mapping algorithm to synchronously correlate the two to the calibration coefficient of the matching threshold. The processor's built-in adaptive calibration engine generates a lateral offset compensation factor corresponding to speed fluctuations and a longitudinal distortion compensation factor corresponding to tension changes through a pre-trained distortion influence model. The two compensation factors are superimposed to generate a real-time lowering instruction for the matching threshold, which ultimately drives the barcode recognition module to call the update. After the matching threshold is reached, feature point matching is performed to ensure stable recognition accuracy under physical distortion conditions. After threshold calibration, a channel stability index needs to be generated by quantifying the channel fluctuation state to provide a basis for allocation decisions. This involves calculating the fluctuation degree index of the instantaneous tape speed fluctuation of each channel and the fluctuation degree index of the real-time tension change value of each channel. Finally, these two types of indices are converted into a channel stability index that characterizes the recognition stability of the channel. When calculating the fluctuation degree index of the instantaneous tape speed fluctuation, a sliding time window is used to statistically analyze continuously monitored data. The width of the sliding time window is set in relation to the recognition time of a single frame of the film tape. The variance is extracted within each window as the first fluctuation quantization value. The variance reflects the dispersion of the speed fluctuation; the larger the variance, the more unstable the speed.When calculating the fluctuation index of real-time tension change value, the statistical range of a sliding time window of the same width is used as the second fluctuation quantification value. The range is the difference between the maximum and minimum tension values ​​within the window, which can intuitively reflect the fluctuation amplitude of tension. Then, the first and second fluctuation quantification values ​​are normalized to dimensionless parameters with a normalization range of 0-1. The calculation is performed by (original value - minimum value) / (maximum value - minimum value). The maximum and minimum values ​​are determined based on the sensor range and are input into a preset exponential fusion model for linear combination. The model assigns a first weighting coefficient (0.6) to the first fluctuation quantification value that is greater than the second weighting coefficient (0.6) to the second fluctuation quantification value. 4) Because speed fluctuations have a more significant impact on recognition accuracy, the final output channel stability index is negatively correlated with the linear combination result. The smaller the combination result, the smaller the fluctuation, and the higher the stability index. This process is executed by the processor's built-in stability evaluation module to ensure that the index can truly reflect the channel recognition stability. When a film strip to be recognized needs to be allocated to a recognition channel, the intelligent allocation stage is entered. This involves obtaining the barcode quality evaluation information of the film strip to be recognized, combining it with the channel stability index of each channel and the number of film strips to be processed in each channel, using the channel stability index as the main weighting factor for weighted calculation, generating the comprehensive allocation weight for each channel, and selecting the comprehensive allocation weight. The channel with the highest weight is assigned the film strips to be recognized. Barcode quality assessment information includes barcode edge blurriness score, barcode unit wear level, and feature point contrast. Edge blurriness score and wear level are the main indicators of quality degradation, each with a weight of 0.4, while feature point contrast is an auxiliary indicator with a weight of 0.2. The three are weighted to generate a comprehensive quality assessment value, ranging from 0 to 100, with higher values ​​indicating better quality. During the weighted calculation, the reciprocal of the channel stability index is first converted into a channel stability weight base value. Then, the number of film strips currently being processed in each channel is converted into a channel load coefficient. The weighting is calculated according to a preset rule: the stability weight base value accounts for 70%, and the channel load coefficient accounts for 30%. The weighted average is then used to generate a comprehensive allocation weight. If the comprehensive quality assessment value is lower than the set degradation threshold, the stability weight base value is automatically increased to 80%, prioritizing allocation to more stable channels to ensure a higher recognition success rate. After selecting the channel with the highest comprehensive allocation weight, the film strip to be recognized is bound to the recognition queue of that channel through a distributed task scheduler. Simultaneously, the latest speed fluctuation and tension change values ​​of that channel are synchronized with the adaptive calibration engine to achieve threshold pre-calibration. An update to the channel's stability index is requested from the stability assessment module, and a channel load increment signal is sent to refresh the comprehensive allocation weight, forming a closed-loop control link to ensure the real-time performance and accuracy of subsequent allocation decisions.

[0019] After real-time monitoring of the instantaneous belt speed fluctuations and real-time tension changes of the film tape in each recognition channel, the processor needs to accurately process this dynamic data and convert it into calibration instructions for barcode recognition feature point matching thresholds to offset the impact of physical fluctuations on recognition accuracy. The specific implementation method is as follows: First, the processor's built-in dynamic parameter adjustment module captures continuous sampling data of instantaneous belt speed fluctuations and real-time tension changes in real time. This module, a dedicated data processing unit integrated into the processor, possesses high-frequency data reception and caching capabilities. Its acquisition frequency is synchronized with the front-end sensors. The laser velocimeter sensor and tension sensor both operate at 100Hz, so the module receives raw data from the sensor data bus at 100Hz. Simultaneously, it performs real-time preprocessing on the continuous sampling data, removing spike noise caused by mechanical vibration through a 5-point moving average filter. The preprocessed data is then cached in the processor's high-speed cache in the format of timestamp, speed fluctuation, and tension change. The cached data retains the most recent 100 sets to provide continuous data for subsequent correlation calculations. Based on this, a nonlinear mapping algorithm is then used to synchronously correlate the two input values ​​to the calibration coefficient of the matching threshold. The nonlinear mapping algorithm uses an S-shaped function mapping, which can avoid over- or under-calibration caused by linear mapping. Its core is to synchronously map the two physical quantities with different dimensions, velocity fluctuation and tension change, to the dimensionless calibration coefficient range of 0-1. For example, when the velocity fluctuation is 0.2 m / s (corresponding to a mapping coefficient of 0.3) and the tension change is 1N (corresponding to a mapping coefficient of 0.2), the algorithm generates a comprehensive calibration coefficient of 0.3×0.6+0.2×0.4=0.26 through weighted fusion (velocity weight 0.6, tension weight 0.4). This coefficient directly reflects the proportion of the threshold that needs to be adjusted under the current fluctuation state. The larger the coefficient, the more significant the fluctuation and the stronger the need for threshold reduction.

[0020] The processor's built-in adaptive calibration engine handles the core compensation calculation tasks. This engine is a computing unit with model inference capabilities, pre-loaded with a distortion impact model trained through extensive experiments. This model is generated based on 1000 sets of data under different operating conditions. In the experiments, speed fluctuations of 0.1-1 m / s and tension changes of 0.5-5 N were set. The lateral offset (pixel level) and vertical distortion rate (percentage) of the barcode were collected under the corresponding conditions. A model was established to establish the correspondence between speed fluctuations, lateral offset tension changes, and vertical distortion rates. The model output is the lateral offset compensation factor corresponding to the instantaneous conveyor speed fluctuation and the vertical distortion compensation factor corresponding to the real-time tension change. The engine calls this model, first generating two compensation factors separately, and then superimposing them at a ratio of lateral offset compensation factor × 0.6 + vertical distortion compensation factor × 0.4. For example, with a lateral factor of 0.8 and a vertical factor of 1.2, the superimposed result is 0.8 × 0.6 + 1.2 × 0.4 = 0.96. This superimposed result is... As a reference value for the calibration amplitude of the matching threshold, the adaptive calibration engine generates a real-time downward adjustment instruction for the matching threshold based on the superimposed compensation factor. The instruction format is current threshold - (current threshold × calibration amplitude reference value × 0.1), where 0.1 is a preset calibration step size coefficient to avoid misidentification caused by excessive single downward adjustment. For example, if the current matching threshold is 30 and the calibration amplitude reference value is 0.96, the downward adjustment instruction is calculated as 30 - (30 × 0.96 × 0.1) = 30 - 2.88 = 27.12. The integer 27 is taken as the updated matching threshold. Finally, the processor sends the real-time downward adjustment instruction to the barcode recognition module through the internal data bus, driving the barcode recognition module to call the updated matching threshold to perform feature point matching operation. After receiving the instruction, the barcode recognition module immediately updates the threshold parameters of the internal feature point detection algorithm. In subsequent single-frame barcode image recognition, the new threshold is used to determine whether a pixel is a target feature point, ensuring accurate recognition of barcode feature points even under physical distortion caused by speed fluctuations and tension changes.

[0021] To ensure that the calibration of the matching threshold can accurately adapt to different degrees of physical fluctuations, it is necessary to clarify the quantification rules of the calibration direction and achieve the coordination of speed and tension compensation through parallel computing to avoid accuracy deviations caused by single parameter calibration. The specific implementation method is as follows: The calibration direction of the matching threshold is initially limited to a downward trend that is positively correlated with the increment of instantaneous film speed fluctuation. The increment of instantaneous film speed fluctuation refers to the difference between the speed fluctuation in the current sampling period and the fluctuation in the previous sampling period. This increment directly reflects the degree of aggravation of speed fluctuation; the larger the increment, the more unstable the lateral movement of the film strip, and the more significant the lateral offset error during sampling. Therefore, the downward adjustment of the matching threshold needs to be increased synchronously. This positive correlation was calibrated experimentally: when the increment is 0-0.1 m / s, the downward adjustment is 1; when the increment is 0.1-0.2 m / s, the downward adjustment is 2, and the threshold drops to 28. For every 0.1 m / s increase in increment... The adjustment range increases by 1 for every s increment, with a maximum adjustment range not exceeding 8, to avoid misidentification due to excessively low thresholds. For example, when the increment is 0.5 m / s, the adjustment range is 5, ensuring that the adjustment range accurately offsets the lateral offset effect of different increments. Simultaneously, the calibration direction is limited to a downward trend positively correlated with the real-time tension change value increment. The real-time tension change value increment is the difference between the current cycle's tension change value and the previous cycle's. The larger the increment, the more drastic the change in tensile force on the film strip, and the more severe the longitudinal elastic deformation of the barcode unit. A larger threshold adjustment is needed to relax the tolerance for longitudinal deformation. Similarly, the positive correlation is calibrated experimentally, with an increment of 0-0.5 N. The adjustment range is 1. When the increment is 0.5-1N, the adjustment range is 2. For every 0.5N increase in increment, the adjustment range increases by 1, with a maximum adjustment range of 6. For example, when the increment is 2N, the adjustment range is 4. This ensures that it can adapt to barcode distortion of different stretching degrees. The adjustment operation of the matching threshold is achieved by merging the parallel compensation channel calculation of the adaptive calibration engine. The parallel compensation channel consists of two independent calculation units within the adaptive calibration engine: the speed compensation channel and the tension compensation channel. The speed compensation channel only receives the instantaneous belt speed fluctuation increment data and calculates the adjustment range of the speed dimension according to the preset positive correlation rule. The tension compensation channel only... The system receives real-time incremental data of tension changes and calculates the downward adjustment of the tension dimension according to another set of preset positive correlation rules. The two channels are operated synchronously to avoid delays caused by serial operations. After the operation is completed, the two downward adjustment values ​​are sent to the engine's fusion unit. The fusion unit calculates the total downward adjustment value according to the weight ratio of speed downward adjustment value × 0.6 + tension downward adjustment value × 0.4. For example, if the speed is reduced by 3 and the tension is reduced by 3, the total adjustment value is 3 × 0.6 + 3 × 0.4 = 3. Finally, the system generates a downward adjustment instruction to match the threshold based on the total adjustment value, ensuring that the compensation effects of the two physical fluctuations can work together, without omitting the influence of any fluctuation or repeatedly adding calibration values.

[0022] After generating the lateral offset compensation factor and the longitudinal distortion compensation factor, the matching threshold needs to be precisely adjusted through their synergistic effect to offset the influence of lateral displacement error and longitudinal deformation, respectively, to ensure that the feature point matching accuracy remains stable under physical distortion conditions. The specific implementation method is as follows: First, the lateral offset compensation factor is used to narrow the lateral tolerance range of feature point matching to offset pixel displacement errors in the direction of film movement. The lateral tolerance range refers to the maximum allowable positional deviation of barcode feature points in the horizontal direction of the image (direction of film movement), which is set to 5 pixels by default. This means that the actual position of the feature point is less than or equal to 5 pixels from the theoretical position and can still be recognized. The lateral offset compensation factor is a coefficient of 0.5-1.0 generated based on the speed fluctuation. Its function is to reduce the lateral tolerance range proportionally. For example, if the compensation factor is 0.8, the original deviation range of 5 pixels will be reduced to 5 × 0.8 = 4 pixels. This means that feature points will not be recognized if the lateral displacement exceeds 4 pixels, thus accurately filtering out excessive lateral offset pixels caused by speed fluctuations and avoiding mismatch of feature points due to displacement errors. At the same time, the longitudinal distortion compensation factor relaxes the longitudinal deformation tolerance of feature point matching, allowing barcode recognition to adapt to the elastic stretching deformation of barcode units. The longitudinal deformation tolerance refers to the maximum allowable stretching ratio of a barcode unit in the longitudinal direction (perpendicular to the direction of film movement). The default setting is 2%, meaning that a unit length stretch of less than or equal to 2% can still be recognized. The longitudinal distortion compensation factor is a coefficient of 1.0-1.5 generated based on the tension change value. Its function is to relax the longitudinal deformation tolerance proportionally. For example, if the compensation factor is 1.2 and the original tolerance is 2%, the relaxed tolerance is 2% × 1.2 = 2.4%. Even if the barcode unit deforms by 2.3% due to tension stretching, it can still be determined as a valid feature point, avoiding missed feature point recognition due to longitudinal distortion. The compensation effect of both is adjusted through the weighted output of the parallel compensation channel to adjust the matching threshold. The specific adjustment logic is as follows: The fusion unit of the parallel compensation channel first calculates the comprehensive compensation coefficient by weighting the lateral offset compensation factor and the longitudinal distortion compensation factor with a weight of 0.6 for the lateral side and 0.4 for the longitudinal side. Then, it multiplies this comprehensive coefficient by the base adjustment of the current matching threshold to obtain the final threshold adjustment amount (3×0.96≈3). Finally, it subtracts the adjustment amount from the current threshold to obtain the updated matching threshold. For example, the original threshold is 30-3=27. This weighted adjustment method ensures that the compensation effects of lateral offset and longitudinal distortion can be integrated into the threshold adjustment simultaneously. It improves the position matching accuracy by narrowing the lateral deviation range and adapts to deformation by relaxing the longitudinal tolerance. Ultimately, it keeps the feature point matching accuracy stable under the physical distortion conditions of speed fluctuation and tension change. Experiments have verified that this adjustment method can improve the barcode recognition accuracy from 75% without compensation to over 98%.

[0023] After completing the dynamic calibration of the matching threshold, in order to objectively quantify the operational stability of each identification channel, it is necessary to further calculate the fluctuation degree index of instantaneous belt speed fluctuation and real-time tension change value. Only by extracting fluctuation characteristics through standardized statistical methods can the bias of subjective judgment be avoided, providing a reliable quantitative basis for the subsequent generation of channel stability index. The specific implementation method is as follows: When calculating the fluctuation index of instantaneous belt speed fluctuation in each channel, the core method is to use a sliding time window to statistically analyze the instantaneous belt speed fluctuation data continuously monitored in each channel. A sliding time window is a fixed time interval that can move along the time axis. Its characteristic is that it continuously discards the oldest data as new data is acquired, always maintaining statistics on the latest monitoring data, and can reflect short-term fluctuation trends in real time without lag. In specific implementation, the stability assessment module of each channel moves the window at a rhythm synchronized with the sensor sampling frequency, moving by one sampling cycle each time. Continuous instantaneous belt speed fluctuation data is cached in real time within the window. During the statistical process, the module reads the data within the window every 10ms. All data is processed to ensure that the statistical results are synchronized with real-time operating conditions. After acquiring the speed fluctuation data within the window, the variance is extracted as the first fluctuation quantification value. Variance is a statistical measure of the dispersion of data and can accurately reflect the severity of speed fluctuations. The larger the variance, the greater the difference in speed fluctuation at different times within the window, and the more unstable the conveyor belt. During calculation, the average value of all speed fluctuation data within the window is first calculated using the built-in mean calculation unit of the module. Then, the difference between each data point and the average value is calculated one by one. The squares of the differences are summed, and finally divided by the number of data points to obtain the variance. For example, if the data within the window are 0.1, 0.2, 0.1, 0.3, and 0.2 m / s, the average value is 0. 0.18 m / s, with a variance of approximately 0.0008 after calculation, this value is the first fluctuation quantification value, directly related to the stability of the velocity dimension. When calculating the fluctuation index of real-time tension change values ​​for each channel, to ensure the time consistency of data statistics, the range of real-time tension change values ​​calculated using the same sliding time window as the velocity fluctuation calculation is used as the second fluctuation quantification value. The same time window means that the window width and moving frequency of the tension fluctuation statistics are completely consistent with the velocity fluctuation, ensuring that the velocity and tension fluctuation data within the same time period can correspond and match, avoiding stability assessment deviations caused by time misalignment. The range refers to the difference between the maximum and minimum values ​​of the tension change values ​​within the window. Compared to variance, range more directly reflects the amplitude of tension fluctuations and has higher computational efficiency, making it suitable for real-time monitoring scenarios. The core impact of tension fluctuations is the resulting vertical distortion of the barcode, and the degree of distortion is directly related to the extreme differences in tension. Therefore, range can more accurately characterize the interference of tension on recognition stability. This range is the second fluctuation quantification value. The width of the sliding time window is set in association with the recognition time of a single frame of the film strip. This is because the window width needs to cover at least one complete barcode recognition process to ensure that the fluctuation index can reflect the actual tape-moving state during recognition. The recognition time of a single frame of the film strip refers to the time required for the barcode recognition module to process one frame of barcode image. The logic of the association setting is as follows: The window width should be 3-5 times the time taken to recognize a single frame. This avoids insufficient statistical data due to an overly narrow window and data lag due to an overly wide window. Specifically, the processor obtains the time taken to recognize a single frame by reading the status register of the barcode recognition module, automatically calculates the window width, and synchronizes the width parameter to the statistical units of each channel. This ensures that the window settings of all channels are adapted to the current recognition rhythm, and that the fluctuation index can be accurately correlated with physical interference in the recognition process.

[0024] After obtaining the first quantized value of fluctuation (velocity variance) and the second quantized value of fluctuation (tension range), since the two have different units (velocity variance is in m² / s², and tension range is in N), directly combining them will lead to an imbalance in the influence weights of velocity fluctuation and tension fluctuation. Therefore, it is necessary to first eliminate the difference in units through normalization, and then generate an intuitive channel stability index through an index fusion model. The specific implementation method is as follows: First, the first and second fluctuation quantization values ​​are normalized to dimensionless parameters. Normalization refers to mapping raw data from different ranges to a unified interval of 0-1, eliminating the difference between dimensions and numerical ranges. The actual fluctuation range of the first fluctuation quantization value (velocity variance) is calibrated to 0-0.1 m² / s² using sensor range and historical data (exceeding this range is considered abnormal fluctuation). Normalization is calculated as (current first quantization value - first minimum value) / (first maximum value - first minimum value). For example, if the current value is 0.05 m² / s², the normalization result is (0.05 - 0) / (0.1 - 0) = 0.5; the second fluctuation quantization value (tension range)... The calibration range is 0-5N. Normalization is performed using the same logic. For example, if the current value is 2.5N, the result is (2.5-0) / (5-0) = 0.5. The normalization process is executed by the preprocessing unit of the stability assessment module, updating the normalization result every 100ms to ensure data real-time performance and provide a comparable basis for subsequent linear combinations. The two dimensionless parameters after normalization are input into a preset exponential fusion model for linear combination. The exponential fusion model is a structured calculation logic pre-stored in the stability assessment module. Its core function is to fuse the two parameters into a single comprehensive value according to preset weights. Essentially, it reflects the degree of influence of different fluctuations on stability through linear weighting. When the model is called, the module retrieves the parameters from the preprocessing unit. Two normalized parameters are read from the database and automatically substituted into a preset linear combination formula. The comprehensive value = first normalized result × first weight coefficient + second normalized result × second weight coefficient. This ensures that the combination process is fully automated and requires no manual intervention. To highlight the dominant influence of speed fluctuations on recognition stability, the normalized result of the first fluctuation quantization value is assigned to the first weight coefficient, and the normalized result of the second fluctuation quantization value is assigned to the second weight coefficient. The first weight coefficient is greater than the second weight coefficient. The weight coefficients are fixed proportions determined through numerous comparative experiments. The experiments simulated different degrees of speed fluctuations and tension fluctuations, and statistically analyzed their impact on barcode recognition accuracy. It was found that speed fluctuations lead to... The accuracy drop caused by the velocity fluctuation is 1.5 times that of the tension fluctuation. Therefore, the first weighting coefficient is set to 0.6 and the second weighting coefficient is set to 0.4. This ratio ensures that the core impact of velocity fluctuation is fully reflected without completely ignoring the interference of tension fluctuation. The weighting coefficients are pre-stored in the exponential fusion model and can be fine-tuned through the processor's debugging interface, but the default ratio of the first weight to the second weight is maintained. The final output channel stability index is negatively correlated with the linear combination result. This is because the larger the linear combination result, the more severe the velocity and tension fluctuations, and the worse the channel stability. Therefore, a negative correlation mapping is needed to transform the degree of fluctuation into an intuitive stability indicator. The specific mapping logic is as follows: The linear combination result ranges from 0 to 1 (corresponding to no fluctuation to extreme fluctuation), while the channel stability index ranges from 1 to 5, where 1 represents extremely unstable and 5 represents extremely stable. The smaller the combination result, the larger the index. For example, a combination result of 0.2 (small fluctuation) corresponds to an index of 4.2, and a combination result of 0.8 (violent fluctuation) corresponds to an index of 1.8. This mapping relationship allows staff to quickly judge the channel status through the index without interpreting complex fluctuation data. The entire index generation process is executed by the stability assessment module built into the processor. The module first receives the normalized parameters from the preprocessing unit, calls the index fusion model to complete the linear combination, then generates the channel stability index through negative correlation mapping, and finally stores the index in the status database of each channel and displays it synchronously on the monitoring interface, providing a direct basis for the subsequent intelligent allocation decision unit 4 to call the channel stability data.

[0025] After calculating the channel stability index for each channel, the intelligent allocation decision unit 4 also needs to obtain the barcode quality information of the film tape to be identified. Only by combining barcode quality and channel stability can we avoid assigning poor-quality barcodes to unstable channels, which would lead to recognition failure. Therefore, it is necessary to accurately quantify the barcode quality assessment information first. This assessment information includes barcode edge blur score, barcode unit wear level, and feature point contrast. The specific implementation method is as follows: One of the core components of barcode quality assessment information is the barcode edge blur score. This score is a key indicator for measuring the clarity of barcode edges, as the barcode edges are the core area for feature point extraction. Edge blurring directly leads to feature point positioning deviations, severely affecting recognition accuracy. To calculate this score, a 2-megapixel industrial camera deployed at the entrance of the recognition channel first captures an image of the barcode on the film strip to be recognized. The camera lens is kept perpendicular to the film strip surface to ensure no tilt distortion in the image. After capturing the image, the image preprocessing unit converts the RGB image to an 8-bit grayscale image. Then, the Canny edge detection algorithm is used to extract the barcode edges. This algorithm filters out grayscale variations in the image by setting high and low thresholds (high threshold 80, low threshold 40). The process involves dividing pixels to form a continuous barcode edge outline, then calculating the average gradient magnitude of the edges. The gradient magnitude reflects the steepness of the edge; a larger gradient indicates a clearer edge. The gradient magnitude of all edge pixels is iterated and averaged. This average is then mapped to a score of 0-100, resulting in the barcode edge blur score. A score below 50 indicates severe edge blur and a significant decrease in barcode quality. Another core component of barcode quality assessment is the barcode cell wear level. This level assesses the degree of wear or damage to the black and white barcode cells due to prolonged use or improper storage. Wear causes information loss in barcode cells, directly reducing the number of identifiable feature points. To determine this level, the barcode cell wear level is first determined based on the barcode to be identified. The encoding rules are used to locate the boundaries of all barcode units in the image. For example, in Code 128, the widths of the black and white bar units are arranged in a specific ratio. The range of each unit can be accurately divided by the pixel width ratio in the image. Then, the barcode image is binarized, converting the grayscale image to a black-and-white binary image. Black pixels correspond to black bar units, and white pixels correspond to white bar units. Next, the percentage of damaged area in each unit is calculated: for black bar units, the proportion of white pixels (damaged areas caused by wear) to the total area of ​​the unit is calculated; for white bar units, the proportion of black pixels (stains or wear marks) is calculated. The largest percentage of damaged area among all units is taken as the basis for judging the degree of wear, and five wear levels are divided accordingly, with a percentage of damaged area ranging from 0 to... 5% is Level 1 (no obvious wear, barcode units intact), 5%-15% is Level 2 (slight wear, does not affect core features), 15%-30% is Level 3 (moderate wear, some edge units missing), 30%-50% is Level 4 (severe wear, core units missing), and above 50% is Level 5 (extremely severe wear, most units unrecognizable). This level is the barcode unit wear level; the higher the level, the worse the barcode quality. Feature point contrast is an auxiliary indicator for barcode quality assessment. Feature point contrast refers to the grayscale difference between barcode feature points and surrounding background pixels. Low contrast will cause feature points to be submerged by background grayscale, increasing the difficulty of recognition. However, compared to edge blurring and unit wear, its impact can be partially offset by adjusting the matching threshold.Therefore, as an auxiliary indicator, the calculation first involves marking all preset feature points in the image based on the barcode encoding rules. Then, a 3×3 pixel neighborhood window is selected centered on each feature point. The window size is determined experimentally to cover the area surrounding the feature point without including other barcode units. The average difference between the gray value of the feature point pixel within the window and the gray values ​​of the other 8 pixels in the neighborhood is calculated. This difference is the contrast of a single feature point. Finally, the average contrast of all feature points is taken to obtain the feature point contrast of the entire barcode. Among them, the barcode edge blur score and barcode unit wear level are used as the main indicators of quality degradation because they account for more than 70% of the recognition success rate. Edge blur can cause a positioning deviation of 1-2 pixels when extracting feature points, and unit wear can directly reduce 20%-50% of effective feature points, neither of which can be compensated for by subsequent threshold adjustments. Feature point contrast, as an auxiliary quality indicator, can improve the recognition effect by lowering the matching threshold by 10-15 gray units when the contrast is low (average difference of 50). Therefore, it has a lower weight in the comprehensive evaluation. The three factors are weighted to generate the quality assessment. When calculating the overall value, the barcode unit wear level must first be converted into a quantitative score of 0-100 (Level 1 corresponds to 100 points, Level 2 to 80 points, Level 3 to 60 points, Level 4 to 30 points, and Level 5 to 10 points). The feature point contrast (0-255) is then normalized to a dimensionless value of 0-100 by dividing contrast by 255 × 100, ensuring consistency with the numerical range of the barcode edge ambiguity score (0-100). Then, a weighted calculation is performed according to preset weights, assigning a weight of 0.4 to both the barcode edge ambiguity score and the barcode unit wear level quantitative score. The normalized feature point contrast value is assigned a weight of 0.2. The three values ​​(feature point contrast, normalized feature point contrast, and normalized feature point contrast) are multiplied by their respective weights and then summed to obtain the overall quality assessment value. For example, if a barcode on a film strip to be identified has an edge blur score of 70, a wear level of 2 (quantified as 80 points), and a normalized feature point contrast value of 60, its overall quality assessment value = 70 × 0.4 + 80 × 0.4 + 60 × 0.2 = 28 + 32 + 12 = 72. A higher value indicates better barcode quality, providing crucial information for subsequent weighting based on the channel stability index.

[0026] After generating the overall quality assessment value of the film strip to be identified and the channel stability index of each channel, a weighted calculation should be performed with channel stability as the core and load as an auxiliary factor to ensure that the barcode allocation is both adapted to the channel status and avoids overload. The specific implementation method is as follows: When using the channel stability index as the primary weighting factor for weighted calculations, the reciprocal of the channel stability index is first converted into a base value for channel stability weights. This base value is a dimensionless parameter used to quantify the priority of stable channel allocation. Since the channel stability index is negatively correlated with volatility, the reciprocal needs to be calibrated using a mapping table to ensure that the base value for stable channels is higher. In practice, the system calls a pre-stored index-base value mapping table (calibrated based on 500 sets of allocation experiments). For example, channel stability index 5 (extremely stable) corresponds to a base value of 0.8, index 4 (relatively stable) corresponds to 0.7, index 3 (moderately stable) corresponds to 0.5, index 2 (relatively unstable) corresponds to 0.3, and index 1 (extremely unstable) corresponds to 0.1. This mapping table directly relates to the core weight of stable channels in the allocation, avoiding purely mathematical... The reversal of priorities caused by the countdown then converts the number of film strips currently pending processing in each channel into a channel load coefficient. The channel load coefficient is a parameter characterizing the processing pressure of a channel. The more strips there are, the greater the load, and the lower the allocation priority. The conversion is achieved through a hierarchical rule: when there are 0-1 strips pending processing, the coefficient is 1.0 (no load pressure); when there are 2-3 strips, it is 1.3 (light load, sufficient processing capacity); when there are 4-5 strips, it is 1.6 (moderate load, new tasks need to be controlled); and when there are more than 5 strips, it is 2.0 (heavy load, allocation is paused). This rule is based on the maximum processing capacity of 10 strips per minute per channel, ensuring that the coefficient can accurately reflect the impact of the load on recognition efficiency. For example, if channel A currently has 2 strips pending processing, the coefficient is 1.3; if channel B has 4 strips pending processing, the coefficient is 1.6, which intuitively shows that channel B has a greater load pressure.Finally, a weighted average is generated for each channel based on a preset rule: the stability weight base value accounts for a significant proportion, and the channel load coefficient accounts for a secondary proportion. The significant proportion is set at 70%, and the secondary proportion at 30%. This ratio is determined based on the analysis of factors affecting the success rate (channel stability accounts for 75% of the success rate, and load accounts for 25%). For example, channel A has a stability weight base value of 0.8 and a load coefficient of 1.3, so its comprehensive allocation weight is 0.8 × 70% + 1.3 × 30% = 0.56 + 0.39 = 0.95; channel B has a base value of 0.5 and a load coefficient of 1.6, so its weight is 0.5 × 70% + 1.6 × 30% = 0.35 + 0.48 = 0.83. Channel A has a higher weight and is given priority in allocation, ensuring the high priority of stable channels while avoiding over-allocation that could lead to overload. When quality... When the overall evaluation value is lower than the set degradation threshold, the proportion of the channel stability weight base value is automatically increased. The degradation threshold is calibrated to 50 through 1000 sets of barcode recognition experiments. The success rate of barcode recognition below this value is less than 50%. At this time, poor quality barcodes rely more on stable channels to compensate for recognition defects. Therefore, the proportion of the stability weight base value is increased from 70% to 80%, and the proportion of the load coefficient is reduced to 20%. For example, if the overall evaluation value of the film tape to be recognized is 45 (lower than the threshold), the weight of channel A after adjustment is 0.8×80%+1.3×20%=0.64+0.26=0.9, and the weight of channel B after adjustment is 0.5×80%+1.6×20%=0.4+0.32=0.72. This further widens the weight gap between stable and unstable channels, ensuring that poor quality barcodes are preferentially assigned to stable channels and reducing the risk of recognition failure.

[0027] After determining the channel with the highest overall allocation weight, a series of collaborative operations are required to complete the barcode allocation and simultaneously update the system status to form a closed-loop control, avoiding subsequent allocation deviations. The specific implementation method is as follows: When assigning the film strip to be identified to the channel with the highest overall allocation weight, the distributed task scheduler first binds the film strip to the identification queue of the selected channel. The distributed task scheduler is the core module for managing identification tasks of each channel, and has the ability to monitor queues and distribute tasks. During the binding operation, the scheduler first reads the unique identifier of the film strip to be identified and the ID of the selected channel, and then adds the barcode identifier to the identification queue of Channel-05. At the same time, it records the allocation information, including the allocation time, channel ID, and barcode ID, to the system log module for subsequent traceability. If the selected channel queue is full, reaching the 5-channel heavy load threshold, the scheduler automatically selects the second highest weight channel to avoid task blocking. While binding the identification queue, adaptive calibration is also performed. The engine synchronizes the latest instantaneous belt speed fluctuation and real-time tension change values ​​of the channel to achieve threshold pre-calibration. The latest instantaneous belt speed fluctuation and real-time tension change values ​​are obtained from the cache of the dynamic monitoring unit 1 of the channel, and the average of the most recent 10 sets of data is taken. The data is encapsulated into a data packet in the channel ID-speed-tension format through the internal 485 bus of the system and sent to the adaptive calibration engine. After receiving the data, the engine immediately calls the pre-trained distortion influence model to generate the corresponding lateral offset compensation factor and the longitudinal distortion compensation factor. The pre-calibration instruction matching the threshold is generated by superimposing them and stored in the threshold cache area of ​​the channel. This ensures that when the film tape to be identified enters the recognition area, the barcode recognition module can directly call the calibrated threshold without real-time calculation, avoiding recognition delay.

[0028] After synchronization, the scheduler requests an update to the channel stability index from the stability assessment module. The scheduler sends an index update request and a channel ID signal to the stability assessment module. Upon receiving the request, the module immediately recalculates the channel's fluctuation data, using the original sliding time window to recalculate the latest variance of the instantaneous belt speed fluctuation (first fluctuation quantization value) and the range of the real-time tension change value (second fluctuation quantization value). After normalization and linear combination, the updated channel stability index is generated and fed back to the scheduler. Simultaneously, it is synchronized to the index buffer of the intelligent allocation decision unit 4 to ensure that the latest stability data is used in subsequent allocations, avoiding misjudgments based on old data. Finally, a channel load increment signal is sent to refresh the comprehensive allocation weight. The channel load increment signal is generated by the scheduler based on the increment of the number of channels to be processed. The signal format is channel ID and load increment 1. It is sent to the load management module of the intelligent allocation decision unit 4. After receiving the signal, the module immediately updates the channel load coefficient of the channel and recalculates the comprehensive allocation weight according to the original weighting rules, thus completing the real-time refresh of the comprehensive allocation weight. This operation ensures that the system can calculate the weight based on the latest load status during the next allocation, avoiding duplicate allocation due to the load not being updated, forming a closed-loop control link, realizing dynamic synchronization of data of each module, and ensuring the long-term stable operation of the system.

[0029] The second objective of this invention is to provide a system for implementing a multi-channel parallel film tape identification and allocation method including any one of the above-mentioned features, comprising: The dynamic monitoring unit 1 is deployed in the sensor network of each identification channel to collect the instantaneous speed fluctuation of the film tape and the real-time tension change value in real time. The adaptive calibration unit 2 has a built-in pre-trained distortion influence model and nonlinear mapping algorithm. Based on the instantaneous belt speed fluctuation and real-time tension change, it generates a lateral offset compensation factor and a longitudinal distortion compensation factor. After superposition, it outputs a real-time lowering instruction for the matching threshold, driving the barcode recognition to adjust the matching threshold of the feature points. The stability assessment unit 3 calculates the variance index of instantaneous belt speed fluctuation and the range index of real-time tension change through a sliding time window. After normalization and linear combination of preset weight coefficients, a channel stability index negatively correlated with channel stability is generated. The intelligent allocation decision unit 4 generates a comprehensive quality assessment value by parsing the barcode quality assessment information of the film tape to be identified. It calls the channel stability index and combines it with the current load of each channel. It performs weighted fusion calculation with the stability index as the main weight factor to generate a comprehensive allocation weight. It selects the channel with the highest comprehensive allocation weight and binds the film tape to be identified through a distributed queue to form a closed-loop control link.

[0030] In this invention, the dynamic monitoring unit 1 collects the instantaneous belt speed fluctuation and real-time tension change of each channel's film strip in real time. The adaptive calibration unit 2 generates lateral offset and longitudinal distortion compensation factors based on a pre-trained distortion influence model. After superposition, the feature point matching threshold is lowered to compensate for the impact of physical distortion on recognition accuracy. The stability evaluation unit 3 calculates the fluctuation index through a sliding time window, and generates an index negatively correlated with channel stability through normalization and linear combination. The intelligent allocation decision unit 4 analyzes the barcode quality evaluation information, combines the channel stability index and load, calculates the comprehensive allocation weight with stability as the main weight, selects the channel with the highest weight to allocate the film strip, forming a closed-loop control to improve the multi-channel recognition accuracy and allocation efficiency.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-channel parallel film tape identification and allocation method, characterized in that: Includes the following steps: S1. Real-time monitoring of the instantaneous belt speed fluctuation and real-time tension change value of the film belt in each identification channel; S2. Based on the instantaneous belt speed fluctuation and real-time tension change, the processor calibrates the matching threshold of the feature points used for barcode recognition in this channel. The calibration direction of the matching threshold is as follows: The value is adjusted downward as the instantaneous belt speed fluctuation and real-time tension change increase, in order to compensate for the impact of the sampling lateral offset caused by the instantaneous belt speed fluctuation and the barcode longitudinal distortion caused by the real-time tension change on the feature point matching accuracy. S3. Calculate the fluctuation degree index of the instantaneous belt speed fluctuation of each channel and the fluctuation degree index of the real-time tension change value of each channel. Finally, convert the fluctuation degree index of the instantaneous belt speed fluctuation and the fluctuation degree index of the real-time tension change value into a channel stability index that characterizes the channel identification stability. S4. When there is a film strip to be identified that needs to be assigned to the identification channel, obtain the barcode quality assessment information of the film strip to be identified, combine the channel stability index of each channel and the number of film strips to be processed in each channel, perform weighted calculation with the channel stability index as the main weight factor, generate the comprehensive allocation weight of each channel, and select the channel with the highest comprehensive allocation weight to assign the film strip to be identified.

2. The multi-channel parallel film tape identification and allocation method according to claim 1, characterized in that: The processor's dynamic parameter adjustment module captures continuous sampling data of the instantaneous belt speed fluctuation and real-time tension change in real time, and uses a nonlinear mapping algorithm to synchronously associate the two input values ​​with the calibration coefficient of the matching threshold. The processor has a built-in adaptive calibration engine that uses a pre-trained distortion influence model to superimpose the lateral offset compensation factor corresponding to the instantaneous belt speed fluctuation and the longitudinal distortion compensation factor corresponding to the real-time tension change, generating a real-time downward adjustment instruction for the matching threshold. Finally, it drives the barcode recognition to call the updated matching threshold to perform feature point matching operations.

3. The multi-channel parallel film tape identification and allocation method according to claim 2, characterized in that: The calibration direction of the matching threshold is defined as follows: When the instantaneous belt speed fluctuation increases, the adjustment range of the matching threshold is positively correlated with the increment of the instantaneous belt speed fluctuation. When the real-time tension change value increases, the adjustment range of the matching threshold is positively correlated with the increment of the real-time tension change value. Furthermore, the adjustment of the matching threshold by both is achieved through the fusion of the parallel compensation channel calculation of the adaptive calibration engine.

4. The multi-channel parallel film tape identification and allocation method according to claim 2, characterized in that: The lateral offset compensation factor reduces the lateral tolerance range of feature point matching, offsetting pixel displacement errors in the direction of film movement. The longitudinal distortion compensation factor relaxes the longitudinal deformation tolerance of feature point matching, enabling barcode recognition to adapt to the elastic stretching deformation of barcode units. The compensation effect of both is adjusted by the weighted output of the parallel compensation channel to regulate the matching threshold, so that the feature point matching accuracy remains stable under physical distortion conditions.

5. The multi-channel parallel film tape identification and allocation method according to claim 1, characterized in that: When calculating the fluctuation index of the instantaneous tape speed fluctuation of each channel, a sliding time window is used to statistically analyze the instantaneous tape speed fluctuation data continuously monitored in each channel, and the variance is extracted as the first fluctuation quantification value. When calculating the fluctuation index of the real-time tension change value of each channel, the range of the real-time tension change value is statistically analyzed using the same time window as the second fluctuation quantification value. The width of the sliding time window is set in relation to the recognition time of a single frame of the film tape.

6. The multi-channel parallel film tape identification and allocation method according to claim 1, characterized in that: After normalizing the first and second quantized fluctuation values ​​into dimensionless parameters, they are input into a preset exponential fusion model for linear combination. The normalized result of the first quantized fluctuation value is assigned a first weight coefficient, and the normalized result of the second quantized fluctuation value is assigned a second weight coefficient. The first weight coefficient is greater than the second weight coefficient. The final output channel stability index is negatively correlated with the linear combination result. The exponential fusion model is executed by the stability evaluation module built into the processor.

7. The multi-channel parallel film tape identification and allocation method according to claim 1, characterized in that: The barcode quality assessment information includes barcode edge ambiguity score, barcode unit wear level, and feature point contrast. The barcode edge ambiguity score and barcode unit wear level are used as the main indicators of quality deterioration, and the feature point contrast is used as the auxiliary indicator of quality. The three are weighted to generate a comprehensive quality assessment value.

8. The multi-channel parallel film tape identification and allocation method according to claim 7, characterized in that: The weighted calculation using the channel stability index as the main weighting factor specifically includes: The reciprocal of the channel stability index is converted into a channel stability weight base value, and the current number of film strips to be processed in each channel is converted into a channel load coefficient. Finally, the channels are weighted and superimposed according to a preset rule that the stability weight base value accounts for a significant proportion and the channel load coefficient accounts for a secondary proportion to generate a comprehensive allocation weight for each channel. When the comprehensive quality assessment value is lower than the set degradation threshold, the proportion of the channel stability weight base value is automatically increased.

9. The multi-channel parallel film tape identification and allocation method according to claim 7, characterized in that: When selecting the channel with the highest comprehensive allocation weight to assign the film tape to be identified, the distributed task scheduler binds the film tape to be identified to the identification queue of the selected channel. At the same time, it synchronizes the latest instantaneous tape speed fluctuation and real-time tension change value of the channel with the adaptive calibration engine to achieve threshold pre-calibration, requests the stability evaluation module to update the stability index of the channel, and finally sends the channel load increment signal to refresh the comprehensive allocation weight, forming a closed-loop control link.

10. A system for implementing the multi-channel parallel film tape identification and allocation method according to any one of claims 1-9, characterized in that, include: The dynamic monitoring unit (1) is deployed in the sensor network of each identification channel to collect the instantaneous belt speed fluctuation and real-time tension change value of the film tape in real time. The adaptive calibration unit (2) has a built-in pre-trained distortion influence model and nonlinear mapping algorithm. Based on the instantaneous belt speed fluctuation and real-time tension change value, it generates a lateral offset compensation factor and a longitudinal distortion compensation factor. After superposition, it outputs a real-time lowering instruction for the matching threshold, driving the barcode recognition to adjust the matching threshold of the feature points. The stability assessment unit (3) calculates the variance index of instantaneous belt speed fluctuation and the range index of real-time tension change through a sliding time window. After normalization and linear combination of preset weight coefficients, it generates a channel stability index that is negatively correlated with channel stability. The intelligent allocation decision unit (4) generates a comprehensive quality assessment value by parsing the barcode quality assessment information of the film strip to be identified, calls the channel stability index and combines it with the current load quantity of each channel, uses the stability index as the main weight factor to perform weighted fusion calculation to generate a comprehensive allocation weight, selects the channel with the highest comprehensive allocation weight and binds the film strip to be identified through a distributed queue to form a closed-loop control link.

Images