A logistics transportation system with electronic fence and a data calibration method thereof

Abstract

The application relates to a logistics transportation system with an electronic fence and a data calibration method thereof, and relates to the technical field of intelligent workshop logistics transportation calibration. The intelligent workshop logistics transportation system comprises an intelligent fence management terminal, which is used for virtually dividing a visual attention fence area according to an actual conveying line and converting the visual attention fence area into actual attention fence coordinates; a dynamic visual acquisition terminal, which is used for positioning the conveying position of an article, triggering dynamic acquisition of the article in combination with the actual attention fence coordinates, and obtaining an article video stream; a monitoring and identification execution terminal, which is used for cleaning and identifying the article video stream, obtaining an abnormal article, generating an abnormal state code, and completing abnormal removal in combination with an external execution mechanism; and a digital twin optimization terminal, which is used for creating a digital production line according to the production line layout in combination with the visual attention fence area, and simulating and optimizing the fence layout. Through real-time tracking of the position of the article and calculation of the distribution center of the article, an offset correction value is generated, and calibration data is dynamically updated, so that self-calibration and self-adaptive optimization of the electronic fence are realized.

Description

Technical Field

[0001] This application relates to the field of logistics transportation calibration technology, and in particular to a logistics transportation system with an electronic fence and its data calibration method. Background Technology

[0002] During the transportation of goods on logistics conveyor belts, traditional visual recognition systems mostly identify and filter all items within the range that the visual probe can detect. In reality, this is unnecessary because some structures, such as conveyor belts, are constantly in operation and cannot be removed. Continuously having the visual probe analyze data from these structures only leads to a waste of resources and affects efficiency.

[0003] Existing patents disclose a logistics control system and method based on visual recognition. By calculating the material demand and real-time production data of the production line, the system determines the required material delivery needs. It then matches these needs with existing transfer personnel and delivery terminals to obtain the optimal delivery terminal. Simultaneously, during the loading and unloading of materials, the system uses visual recognition to track the loading and unloading time, material information, transfer personnel information, and transportation terminal information, effectively ensuring that the materials are delivered within the specified time. Furthermore, the system monitors and guides the logistics picking, delivery, and assembly processes, effectively reducing the backlog of work-in-process in the workshop, reducing production complexity, improving turnover efficiency, and ensuring the continuous operation of production activities.

[0004] The existing technical solutions mentioned above have the following defects: 1. Data calibration work is mostly based on the actual layout of the logistics line for surveying and mapping, and then the drawings are modified and adjusted proportionally or in the form of estimation, and finally backtracked to the actual logistics line for implementation; in actual operation, there are still many problems that the drawings do not correspond to the actual situation, and frequent modifications and re-layouts are required. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide a logistics transportation system with electronic fences and its data calibration method. By extracting theoretical physical dimensions and establishing a benchmark library through error modeling, and combining coordinate mapping algorithms to generate electronic fences on actual conveyor lines, the calibration data is dynamically corrected using status monitoring and anomaly identification algorithms. This completely solves the difficulties and pain points of existing technologies, achieves high-precision, adaptive intelligent visual calibration, and significantly improves the reliability and maintenance efficiency of logistics sorting systems.

[0006] This was achieved using the following technical solutions: In a first aspect, this application provides a logistics transportation system with an electronic fence, comprising: The intelligent fence management terminal is used to virtually divide the visually concerned fence area according to the actual transmission line and convert it into the actual coordinates of the concerned fence. The dynamic visual acquisition terminal is used to locate the delivery position of the item, and combined with the actual coordinates of the fence of interest, it triggers the dynamic acquisition of the item to obtain the video stream of the item; The monitoring and identification execution terminal is used to clean and identify video streams of items, obtain abnormal items, generate abnormal status codes, and complete the abnormality resolution in conjunction with external execution mechanisms. The digital twin optimization terminal is used to create a digital production line based on the production line layout and visual attention-based fence areas, and to simulate and optimize the fence layout.

[0007] By adopting the above technical solution, the transmission line is virtually partitioned and mapped to actual coordinates through the intelligent fence terminal, which drives the vision terminal to perform positioning trigger acquisition. The video stream is cleaned and analyzed by the recognition terminal to locate anomalies and link the execution mechanism for processing. At the same time, the layout is optimized by the simulation of the digital twin terminal, realizing a closed-loop adaptive management from perception, decision-making to execution, with the advantages of dynamic accuracy, resource efficiency and continuous optimization.

[0008] Furthermore, the intelligent fence management terminal includes: The parameter collection module is used to generate physical structure parameters based on the three-dimensional structural diagram of the logistics conveyor frame and the preset construction error range. The physical structure parameters include frame height, support width, electric roller diameter, and center distance between adjacent rollers. The centerline determination module is used to generate the centerline of the transport path of the item based on the item's dimensions and physical structural parameters. The area division module is used to expand bidirectionally based on the center line of the goods transportation path and the actual conveyor line to obtain the visual attention fence area and determine the visual neglect area. The coordinate mapping module is used to generate a mapping transformation matrix based on the structure pixel size and physical structure parameters, and to perform coordinate transformation on the visual attention fence area to calculate the actual attention fence coordinates.

[0009] By adopting the above technical solution, the three-dimensional structural parameters of the conveyor frame are extracted through the parameter collection module, and the center line of the transportation path is generated based on the size of the items. Then, the visual attention area and the neglect area are divided. Finally, the virtual fence is converted into actual coordinates through the coordinate mapping matrix, realizing the automated and high-precision calibration from physical structure to digital fence, which significantly improves the positioning accuracy and environmental adaptability of the vision system.

[0010] Furthermore, the dynamic visual acquisition terminal includes: The location triggering module is used to locate the transmission position of the items based on the transportation speed of the logistics conveyor, and generate a collection trigger command in combination with the actual coordinates of the fence of interest. The data acquisition module is used to collect data on the transportation process of goods according to the acquisition trigger command, and obtain a global video stream; The blurring and segmentation module is used to blur the carrier and segment the target in the global video stream based on the visually ignored area to obtain the object video stream.

[0011] By adopting the above technical solution, the position triggering module based on motion prediction algorithms (such as Kalman filtering) accurately locates the object and generates acquisition instructions. Then, the background segmentation algorithm (such as Mask R-CNN) is used to blur the visually ignored areas and extract the target in the global video stream, thereby obtaining a clean object video stream. This achieves high-efficiency, low-redundancy dynamic visual acquisition and intelligent preprocessing.

[0012] Furthermore, the monitoring and identification execution terminal includes: The filtering and noise reduction module is used to filter noise from the video stream of the object to obtain a noise-free video stream; The image framing module is used to decompose the noiseless video stream into frames to obtain several frames of object images; The feature extraction module is used to perform edge detection and information extraction on the object image to obtain the object skeleton outline and transportation target information; The anomaly detection module is used to match the skeleton outline of an item with the transportation target information based on a preset item feature information database, and calculate the corresponding feature anomaly value. If all the abnormal feature values ​​are greater than the preset abnormality judgment threshold, then the appearance features and transportation information of the current item are determined to be abnormal. If any outlier value exceeds the anomaly detection threshold, the current item is determined to have either an abnormal appearance or abnormal transportation information. The status determination module is used to perform temporal optical flow tracking on the object image and mark the object's movement trajectory; if the object's movement trajectory does not change, the current object's transmission status is determined to be abnormal. The early warning and notification module is used to collect information on abnormal items based on the type of abnormality and generate an abnormality status code. The anomaly resolution module is used to generate corresponding mechanism control instructions based on the anomaly status code, and to call external actuators to control the anomaly items for sorting, counting, or defect identification.

[0013] By adopting the above technical solution, a digital twin model synchronized with the physical production line is constructed through 3D modeling and real-time data fusion algorithms. Simulation optimization algorithms are then used to simulate and adjust the fence layout and evaluate its effects, achieving low-cost and high-efficiency production line layout optimization and decision support in a virtual environment.

[0014] Secondly, this application also provides a data calibration method based on a logistics transportation system, which adopts the following technical solution; A data labeling method based on a logistics transportation system includes: Extract the theoretical physical dimensions of the component's 3D drawing and combine them with a preset construction error range to generate physical structural parameters and establish a component dimension reference library; By analyzing the component size reference library and combining it with the actual conveyor line, the center line of the goods transportation path is determined, the electronic fence area is generated, and the mapping calibration data is extracted. The mapping calibration data includes the fitting function of the center line of the goods transportation path, the spatial feature anchor point set of the visual attention fence area, and the coordinate transformation matrix of the electronic fence. Monitor the transportation status of items within the electronic fence area, trigger image acquisition and anomaly recognition of the items, correct the mapping calibration data, and obtain corrected calibration data.

[0015] By adopting the above technical solution, a benchmark library is established through theoretical physical dimension extraction and error modeling. An electronic fence is generated on the actual conveyor line by combining coordinate mapping algorithm. The calibration data is dynamically corrected by status monitoring and anomaly recognition algorithm, which realizes high-precision and adaptive intelligent visual calibration, significantly improving the reliability and maintenance efficiency of the logistics sorting system.

[0016] Furthermore, by analyzing the component size reference library and combining it with the actual conveyor line, the centerline of the goods transportation path is determined, an electronic fence area is generated, and mapping calibration data is extracted, including: A global scan of the actual transmission line is performed according to the preset resolution to obtain a diagram of the transmission line components and to calculate the pixel parameters of the components. Based on the component size reference library, perform entity matching on the conveyor line component drawings to extract physical structural parameters; The center anchor point is determined based on the support width and the center distance between adjacent rollers in the physical structure parameters. Based on the center anchor point and the size of the item, the actual conveyor line is smoothly fitted to determine the center line of the item transportation path. Based on the preset effective recognition width, the center line of the item transportation path is offset and closed on both sides to form an electronic fence area; Based on the component pixel parameters and physical structure parameters, the electronic fence area is mapped and transformed, and the mapping calibration data is extracted.

[0017] By adopting the above technical solution, physical parameters are extracted from the component diagram of the conveyor line through global image scanning and feature matching algorithms. The center line of the transport path of the goods is determined by the center anchor point positioning and curve fitting algorithm. Then, the electronic fence and its calibration data are generated by the geometric offset and coordinate mapping algorithm. This realizes automated and high-precision visual calibration based on the actual structure, which significantly improves the system deployment efficiency and environmental adaptability.

[0018] Furthermore, the system monitors the transport status of goods within the electronic fence area, triggers image acquisition and anomaly identification of the goods, corrects the mapping calibration data, and obtains corrected calibration data, including: The electronic fence area is divided according to the production line layout, and visually attention-focused fence areas are selected. Monitor the transportation status of items within the electronic fence area, determine the real-time location of the items, and calculate the coordinates of the item's center point; If the real-time location of an item is within the visual attention fence area, then cluster analysis is performed on the coordinates of the item's center point to obtain the item's distribution center, and the difference between the center point and the center line of the item's transportation path is calculated to obtain the center deviation value. If the center deviation value is greater than the preset center deviation amount, the item will be image acquired and anomaly identified to generate the item's transportation trajectory. The actual transport fence area is obtained by fitting the transport trajectory of the goods, and the deviation is calculated between it and the visually concerned fence area to obtain the offset correction value. The mapping calibration data is corrected based on the offset correction value, and the corrected calibration data is calculated.

[0019] By adopting the above technical solution, the location of items is tracked in real time and the distribution center is calculated through status monitoring and cluster analysis algorithms. When the deviation from the theoretical path exceeds the limit, image acquisition and trajectory fitting are triggered. Then, the offset correction value is generated through geometric deviation calculation and the calibration data is dynamically updated. This realizes the self-calibration and adaptive optimization of the electronic fence based on actual operation data, which significantly improves the long-term accuracy and robustness of the system.

[0020] Thirdly, this application also provides a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the data calibration method based on the logistics transportation system as described above.

[0021] In summary, the beneficial technical effects of this application are as follows: By tracking the location of items in real time and calculating their distribution center, when the deviation from the theoretical path exceeds the limit, image acquisition and trajectory fitting are triggered. Then, offset correction values ​​are generated through geometric deviation calculation and calibration data is dynamically updated. This achieves self-calibration and adaptive optimization of the electronic fence based on actual operating data, which significantly improves the long-term accuracy and robustness of the system. By extracting theoretical physical dimensions and establishing a benchmark library through error modeling, and generating electronic fences on actual conveyor lines using coordinate mapping algorithms, and dynamically correcting calibration data using status monitoring and anomaly recognition algorithms, high-precision, adaptive intelligent visual calibration is achieved, significantly improving the reliability and maintenance efficiency of the logistics sorting system. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the logistics and transportation system in this application; Figure 2This is a schematic diagram of the structure of the intelligent fence management terminal in this application; Figure 3 This is a schematic diagram of the structure of the monitoring and identification execution terminal in this application; Figure 4 This is a flowchart illustrating the data calibration method used in this application. Detailed Implementation

[0023] The present application will be further described in detail below with reference to the accompanying drawings.

[0024] On the one hand, refer to Figure 1 This application discloses a logistics transportation system with an electronic fence, comprising: The intelligent fence management terminal is used to virtually divide the visually concerned fence area according to the actual transmission line and convert it into the actual coordinates of the concerned fence. The dynamic visual acquisition terminal is used to locate the delivery position of the item, and combined with the actual coordinates of the fence of interest, it triggers the dynamic acquisition of the item to obtain the video stream of the item; The monitoring and identification execution terminal is used to clean and identify video streams of items, obtain abnormal items, generate abnormal status codes, and complete the abnormality resolution in conjunction with external execution mechanisms. The digital twin optimization terminal is used to create a digital production line based on the production line layout and visual attention-based fence areas, and to simulate and optimize the fence layout.

[0025] The implementation principle of this embodiment is as follows: The intelligent fence management terminal uses computer vision and geometric mapping algorithms to virtually divide the visual attention area based on the actual conveyor line structure and generate actual coordinates; the dynamic visual acquisition terminal uses target tracking and area triggering algorithms to accurately locate the position of the item and start acquisition to obtain a high-quality video stream of the item; the monitoring and identification execution terminal uses image processing and machine learning algorithms to clean and analyze the video stream to identify anomalies and generate status codes, and links external execution mechanisms to handle anomalies; the digital twin optimization terminal uses simulation modeling and optimization algorithms to build a digital twin model of the production line and simulate and optimize the fence layout, forming a full-link intelligent closed loop of perception, decision-making, execution and continuous optimization.

[0026] Preferred, refer to Figure 2 The intelligent fence management terminal includes: The parameter collection module is used to generate physical structure parameters based on the three-dimensional structural diagram of the logistics conveyor frame and the preset construction error range. The physical structure parameters include frame height, support width, electric roller diameter, and center distance between adjacent rollers. In this embodiment, by analyzing the point cloud data of the three-dimensional structure diagram and combining it with the measured points of laser scanning, the physical structure parameters are generated by fitting using the least squares method, and a tolerance analysis algorithm is introduced to dynamically update the construction error range, thereby establishing an adaptive benchmark library that is resistant to environmental disturbances.

[0027] The centerline determination module is used to generate the centerline of the transport path of the item based on the item's dimensions and physical structural parameters. In this embodiment, a transport line dynamics model is constructed based on physical structure parameters. By combining a path planning algorithm (such as a variant of the A* algorithm) with item size constraints, an optimal transport path centerline that balances efficiency and stability is generated, supporting templated configuration for multiple item sizes.

[0028] The area division module is used to expand bidirectionally based on the center line of the goods transportation path and the actual conveyor line to obtain the visual attention fence area and determine the visual neglect area. In this embodiment, a morphological image processing method is used to perform adaptive bidirectional erosion and dilation operations based on the center line. Combined with a heat map of the historical object location distribution, the boundary of the area of ​​interest is dynamically adjusted, and semantic segmentation technology is used to automatically identify visually ignored areas.

[0029] The coordinate mapping module is used to generate a mapping transformation matrix based on the structure pixel size and physical structure parameters, and to perform coordinate transformation on the visual attention fence area to calculate the actual attention fence coordinates. In this embodiment, a high-precision mapping transformation matrix is ​​solved by combining the nine-point calibration method with a nonlinear optimization algorithm, and an online recalibration mechanism is integrated to compensate for coordinate drift caused by equipment vibration or thermal deformation in real time.

[0030] Preferably, the dynamic visual acquisition terminal includes: The location triggering module is used to locate the transmission position of the items based on the transportation speed of the logistics conveyor, and generate a collection trigger command in combination with the actual coordinates of the fence of interest. In this embodiment, encoder pulse signals and visual tracking algorithms (such as KCF trackers) are integrated to construct an object motion state prediction model. When the object enters the preset look-ahead interval of the fence of interest coordinates, a pre-trigger command is generated to achieve zero-delay acquisition.

[0031] The data acquisition module is used to collect data on the transportation process of goods according to the acquisition trigger command, and obtain a global video stream; In this embodiment, a programmable exposure strategy and multi-frame HDR fusion technology are used to adaptively cope with high-speed motion blur and sudden changes in lighting, and timestamps and location metadata are embedded simultaneously to generate a structured global video stream.

[0032] The blurring and segmentation module is used to blur the carrier and segment the target in the global video stream based on the visually ignored area to obtain the object video stream; In this embodiment, an attention-based U-Net network model is applied to perform real-time semantic segmentation on the global video stream, accurately separating dynamic objects from the background of visually ignored areas, and generating a visually harmonious video stream of objects through Gaussian blurring and edge fusion techniques.

[0033] Preferred, refer to Figure 3 The monitoring and identification execution terminals include: The filtering and noise reduction module is used to filter noise from the video stream of the object to obtain a noise-free video stream; In this embodiment, a combined denoising algorithm of temporal nonlocal mean filtering and spatial wavelet transform is used to effectively suppress random noise and periodic interference under complex working conditions while preserving the edge details of the object.

[0034] The image framing module is used to decompose the noiseless video stream into frames to obtain several frames of object images; The feature extraction module is used to perform edge detection and information extraction on the object image to obtain the object skeleton outline and transportation target information; In this embodiment, a lightweight convolutional neural network (such as MobileNet-V3) is deployed to perform multi-scale feature extraction, and a traditional edge detection algorithm (an optimized version of the Canny operator) is run in parallel to obtain sub-pixel-level skeleton contours. An enhanced item descriptor is output through a feature fusion network.

[0035] The anomaly detection module is used to match the skeleton outline of an item with the transportation target information based on a preset item feature information database, and calculate the corresponding feature anomaly value. If all the abnormal feature values ​​are greater than the preset abnormality judgment threshold, then the appearance features and transportation information of the current item are determined to be abnormal. If any outlier value exceeds the anomaly detection threshold, the current item is determined to have either an abnormal appearance or abnormal transportation information. In this embodiment, a feature matching engine based on metric learning is constructed. The extracted features are compared with the information database to measure similarity. An adaptive threshold mechanism is used to calculate feature outliers. A graph neural network is introduced to model the relationship between features in order to identify compound anomaly patterns.

[0036] The status determination module is used to perform temporal optical flow tracking on the object image and mark the object's movement trajectory; if the object's movement trajectory does not change, the current object's transmission status is determined to be abnormal. In this embodiment, the dense optical flow algorithm (an improved version of the Farneback algorithm) is used to reconstruct the motion trajectory, and the trajectory state transition probability is analyzed by combining the Markov chain model to accurately identify transmission anomalies such as stagnation and backflow.

[0037] The early warning and notification module is used to collect information on abnormal items based on the type of abnormality and generate an abnormality status code. In this embodiment, a hierarchical early warning system is established, and an abnormality type is mapped to a standardized abnormality status code through a rule engine, and a multi-channel notification strategy (audio-visual alarm, SMS, industrial bus message) is integrated.

[0038] The anomaly resolution module is used to generate corresponding mechanism control instructions based on the anomaly status code, and to call external actuators to control the anomaly items for sorting, counting, or defect identification.

[0039] In this embodiment, an instruction orchestration engine is developed to generate composite mechanism control instructions, including path planning and torque control parameters, based on abnormal status codes, supporting real-time collaborative control with actuators such as robots and sorting machines.

[0040] Preferably, the digital twin optimized terminal includes: The twin construction module is used to perform twin simulation of the electronic fence and logistics conveyor based on the production line layout, and generate an initial twin conveyor line; In this embodiment, a physics-based modeling method is adopted, and a dynamic twin of the transmission line with deformation simulation capability is constructed by combining finite element analysis, so as to realize the deep coupling modeling of the electronic fence and the mechanical structure.

[0041] The data synchronization module is used to synchronize information of the initial twin conveyor line based on the visual attention fence area and the item transmission status to obtain a digital production line. In this embodiment, millisecond-level state synchronization is achieved through edge computing nodes, and a virtual-real data consistency verification mechanism is established to ensure the spatiotemporal alignment of the twin and the physical production line.

[0042] The simulation optimization module is used to simulate and adjust the fence layout of the digital production line and generate a layout change evaluation table. In this embodiment, based on a multi-objective optimization algorithm (such as NSGA-II), the fence layout is parametrically scanned and analyzed in a twin environment, automatically generating a layout modification evaluation table that includes dimensions such as efficiency improvement rate, false detection rate change, and return on investment, and supporting human-computer interactive optimization simulation.

[0043] The implementation principle of this embodiment is as follows: Through the intelligent fence management terminal, physical structure parameters are extracted based on the three-dimensional structure diagram analysis and error modeling algorithm, and the path planning algorithm is used to generate the center line of the item transportation path. Then, the virtual visual attention fence area is converted into actual coordinates through the geometric mapping algorithm. The dynamic visual acquisition terminal integrates encoder signals and visual tracking algorithms to predict the position of objects, initiates adaptive exposure acquisition within the trigger area, and uses a semantic segmentation network to separate objects from the background to obtain a high-quality video stream of objects; the identification and monitoring execution terminal uses temporal-spatial combined filtering, lightweight convolutional networks, and optical flow analysis algorithms to denoise, extract features, and identify anomalies in the video stream, and then generates an anomaly status code through a rule engine and links the execution mechanism to complete the handling. Meanwhile, the digital twin optimization terminal constructs a transmission line dynamic twin based on physical modeling, achieves virtual-real synchronization through a high-throughput data bus, and uses a multi-objective optimization algorithm to perform parametric scanning and deduction of the fence layout in a simulation environment. Example

[0044] Draw some electronic fences in the video to isolate and identify the area, which will facilitate image tracking and processing; The electronic fence uses a coordinate data format for marking, and the area to be identified is located within the coordinate system, similar to a box selection. Based on this, the data is formed visually, regardless of the size of the drawing, only the quantity and the size of the center line are considered, so that all data is processed in the concept of size, and the feasible route for transporting goods is marked within the electronic fence.

[0045] For example, the dimensions of the electric rollers on the logistics conveyor belt, and the distance between adjacent rollers, are all presented in a digitized form. Using this data as a reference, a specific portion of the image captured by the vision probe is designated as the area requiring identification and data analysis; this portion is the electronic fence. The area within the electronic fence is the effective area that needs to be identified.

[0046] The corresponding data calibration is processed using the same standardized data collection method. In the design and layout of the logistics line, only the quantity and the size and location of the center line of the operating area need to be considered, which can achieve relative uniformity between online and offline surveying and effectively improve the accuracy and efficiency of data calibration.

[0047] The data processing can create a database in the local area network of the logistics line, and all data is entered into the database and verified by AI. As the data continues to accumulate, AI is used to verify the authenticity and rationality of the newly entered data, further improving the effectiveness and accuracy of electronic fences and data labeling.

[0048] On the other hand, refer to Figure 4 This application discloses a data calibration method for a logistics transportation system, comprising: S1: Extract the theoretical physical dimensions of the component's 3D drawing, combine them with the preset construction error range, generate physical structural parameters, and establish a component dimension reference library; S2: Combine the component size reference library with the actual conveyor line to determine the center line of the goods transportation path, generate the electronic fence area, and extract the mapping calibration data; the mapping calibration data includes the fitting function of the center line of the goods transportation path, the spatial feature anchor point set of the visual attention fence area, and the coordinate transformation matrix of the electronic fence; S3: Monitor the transportation status of items within the electronic fence area, trigger image acquisition and anomaly recognition of items, correct the mapping calibration data, and obtain corrected calibration data.

[0049] In this embodiment, the data is automatically extracted using a calibrated 3D vision system to eliminate human error. Secondly, it is determined whether the entered dimensions are sufficient to unambiguously define the spatial structure of the logistics line. For example, the "roller diameter" alone is insufficient for positioning; it is necessary to combine the "center distance between adjacent rollers" and the "offset of the starting position of the roller array" to construct a complete coordinate system. Internal consistency verification is performed by calculating the geometric constraints between these dimensional parameters (e.g., total length = single pitch × number). If a logical conflict is found (e.g., the calculated total length does not match the actual measured total length), the benchmark database data is deemed abnormal, requiring remeasurement to ensure a solid foundation for accurate calibration from the outset.

[0050] First, it's necessary to determine if the visual image can stably identify the reference feature (such as the edge of the roller). If it can, then based on the preset pixel-to-physical size conversion relationship, the user-specified "centerline physical position" and "effective physical width" are converted into specific areas in the image. The key judgment point lies in handling changes in camera perspective: continuously monitoring the position of the reference feature in the image, if the feature pixel coordinates shift due to vibration, it is determined as a slight change in camera pose, and the pixel coordinates of the fence are automatically recalculated and adjusted according to the new feature position, rather than triggering an alarm. This logic ensures that the electronic fence can "follow" the equipment's movement, maintaining the effectiveness of the calibration.

[0051] When new calibration data (such as a new fence definition) is generated, the AI ​​verification module is triggered. Its judgment logic is based on difference comparison: the AI ​​calls up the current visual image, remeasures key physical dimensions (such as the actual roller spacing), and compares this measurement with the corresponding standard value in the "benchmark library" in the database. The system presets an acceptable error threshold (e.g., ±0.5%). If the difference is within the threshold, the new calibration data is deemed valid and allowed to be added to the database. If the difference exceeds the limit, a root cause judgment is triggered: Is it an installation error in the benchmark equipment itself? Or is it uncorrected camera lens distortion? Or is it a recognition error caused by drastic changes in lighting? According to preset rules, an alarm is triggered for suspected hardware problems (such as continuous deviation), and a retry is prompted for instantaneous environmental interference.

[0052] The system tracks targets in real time (through simple moving object detection or sensor signals), calculates the coordinates of the center point of its circumscribed rectangle, and determines whether this coordinate point is located within any "electronic fence" polygonal area. If it "enters," a high-precision identification process is immediately initiated for that fenced area; if it "leaves," the computing power is released. Simultaneously, when targets enter multiple fences at the same time, computing power is dynamically allocated according to preset rules (such as upstream workstation priority, high-speed production line priority) to ensure no delay in identifying critical nodes. This logic ensures that valuable computing resources are 100% applied to the effective area.

[0053] The system identifies the center points of all successfully grabbed items within a defined fence and calculates the standard deviation of their distribution perpendicular to the "operation centerline." If the distribution range (e.g., ±3σ) is consistently and significantly smaller than the currently set "effective width," the fence definition is considered too lenient and has room for optimization. The system will generate a suggested parameter to reduce the width, but this will not be applied automatically. This suggestion, along with a confidence metric (e.g., stable results after 1000 consecutive statistical iterations), will be submitted to the administrator for confirmation or tested in a more rigorous "sandbox" mode.

[0054] The implementation principle of this embodiment is as follows: a component size reference library is established by extracting theoretical physical dimensions and error modeling algorithms. Electronic fences and calibration data are generated on the actual transmission line by combining coordinate mapping algorithms. The calibration parameters are dynamically corrected by state monitoring and adaptive correction algorithms, forming a complete adaptive visual calibration closed loop from theoretical modeling, actual calibration to continuous optimization.

[0055] Preferably, step S2 includes: The actual conveyor line is scanned globally at a preset resolution to obtain images of the conveyor line components and to calculate the pixel-level parameters of the components. Based on the component size reference library, entity matching is performed on the images of the conveyor line components to extract the corresponding physical structure parameters; Based on the support width and the center distance between adjacent rollers in the physical structural parameters, a stable spatial feature anchor point is calibrated. Based on the item size information and the already marked center anchor point, a smooth curve is fitted to determine the centerline of the item transportation path. Based on the preset effective recognition width, the electronic fence area is generated by shifting to both sides and closing along the center line of the item transportation path. Based on the component pixel parameters and physical structure parameters, the electronic fence area is mapped to the actual coordinate space, and structured mapping calibration data is extracted.

[0056] In this embodiment, the entire actual conveyor line is scanned, and a pattern recognition algorithm automatically searches for physical structures that match the feature descriptions (such as roller diameter D and spacing L) in the "component size reference library." The system calculates the error between the actual identified dimensions and the reference library values. If the error is within the allowable threshold (e.g., ±0.5%), the match is considered successful, and these identified structures (e.g., the centers of rollers 1, 5, and 10) are established as stable "spatial feature anchor points." If the match fails, a calibration alarm is triggered, prompting a need to verify the physical reference or vision system parameters. Based on the anchored feature points, the system determines the "centerline of the goods transportation path" according to logistics process requirements or by manual specification by the operator in the augmented reality interface. Mathematically, this centerline can be defined as a smooth spline curve or straight line passing through a series of path points. The system uses the absolute physical coordinates of the anchor points and relative dimensional relationships in a reference library (e.g., the centerline is parallel to the roller array and offset by a certain distance) to accurately calculate the complete equation of the centerline in the real-world coordinate system and stores it as core calibration data. Using the calculated "centerline" as a reference, and based on the preset "effective recognition width" parameter (which is usually determined by the maximum item size plus a safety margin), the system automatically shifts to both sides of the centerline to generate a closed, strip-shaped polygonal area, which is the geometric range of the "electronic fence." The entire fence is not stored as a fixed set of pixel coordinates, but is described digitally as: "a strip-shaped area of ​​width W along the centerline path, with the line connecting [feature anchor point A] and [feature anchor point B] as the reference." This description binds the fence to a physical reference, rather than to volatile image pixels. The complete calibration data package is extracted in a structured manner and stored in the local database. This data package must contain at least: 1) the referenced benchmark library version ID; 2) a list of spatial feature anchor points used and their coordinates; 3) the mathematical description parameters of the centerline; and 4) a data-driven description of the generated electronic fence. Before being stored in the database, a rapid positive verification is required: the system redraws the electronic fence on the screen in real time based on the newly added calibration data and compares it with the operator's intuitive judgment or historical successful data to ensure logical consistency.

[0057] Preferably, step S3 includes: The electronic fence area is divided according to the production line layout, and visually attention-focused fence areas are selected. Monitor the transportation status of items within the electronic fence area, determine the real-time location of the items, and calculate the coordinates of the item's center point; If the real-time location of an item is within the visual attention fence area, then cluster analysis is performed on the coordinates of the item's center point to obtain the item's distribution center, and the difference between the center point and the center line of the item's transportation path is calculated to obtain the center deviation value. If the center deviation value is greater than the preset center deviation amount, the item will be image acquired and anomaly identified to generate the item's transportation trajectory. The actual transport fence area is obtained by fitting the transport trajectory of the goods, and the deviation is calculated between it and the visually concerned fence area to obtain the offset correction value. The mapping calibration data is corrected based on the offset correction value, and the corrected calibration data is calculated.

[0058] In this embodiment, the real-time status within the electronic fence area is continuously monitored. The judgment logic is based on two types of inputs: 1) determining whether an object has entered the area using photoelectric sensors, encoders, or background subtraction visual algorithms; and 2) determining whether the object is within the valid recognition area using an object tracking algorithm. The system determines a "valid trigger event" only when both "object exists" and "is within the valid area" are simultaneously met, and immediately sends a high-priority image acquisition and processing command to the industrial camera and edge computing unit. This judgment logic avoids false triggers caused by objects briefly lingering at the fence edge or vibrating. An AI model is run on the acquired local high-definition images to identify the category, quantity, posture, damage, and other preset defects of the items; at the same time, the actual location distribution of the items within the fence is analyzed. The judgment logic is as follows: the center point coordinates of all successfully processed items within a period of time (e.g., 1000 times) are clustered and statistically analyzed. If a significant and systematic deviation is found between the distribution center and the preset "transportation path centerline" (e.g., the center point of 50 consecutive batches of items is biased to one side of the fence, and the deviation exceeds the threshold δ), a "calibration offset alarm" is triggered, indicating that changes in the environment or equipment have caused the original calibration data to be inaccurate with the actual physical layout.

[0059] Once the "calibration offset alarm" is confirmed, the system calculates a new "actual centerline" and / or a suggested "effective width" based on the statistically analyzed actual path distribution of items, and generates a set of "corrected calibration data proposals". First, in a sandbox environment or a separate test channel, simulate or briefly apply the new parameters to observe their logical consistency and recognition performance. Key judgment logic includes: Can the new fence still completely encompass all normal items? Can it still effectively exclude fixed structures? Is the correction range within the historical normal fluctuation range? If the verification passes, the system will mark the proposal as high confidence and it can be automatically implemented or submitted to the administrator for one-click confirmation; if the verification fails (e.g., new parameters cause overlap with fixed structural areas), it will be judged as a complex anomaly (may involve mechanical failure), the correction proposal will be rejected, and it will be upgraded to an equipment maintenance alarm.

[0060] The implementation principle of this embodiment is as follows: physical parameters are extracted from the conveyor line component diagram and the center anchor point is determined by global scanning and feature matching algorithms. Curve fitting is used to generate the center line of the transportation path and the electronic fence. Then, based on cluster analysis, the position deviation of the items is monitored in real time. When the deviation exceeds the limit, image acquisition and trajectory fitting are triggered. Finally, the calibration data is dynamically corrected through a closed-loop feedback algorithm to achieve adaptive high-precision visual calibration.

[0061] This application discloses a storage medium storing at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the data calibration method based on the logistics transportation system as described above.

[0062] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A logistics transportation system with an electronic fence, characterized in that, include: The intelligent fence management terminal is used to virtually divide the visually concerned fence area according to the actual transmission line and convert it into the actual coordinates of the concerned fence. A dynamic visual acquisition terminal is used to locate the delivery position of an item, and triggers dynamic acquisition of the item by combining the actual focus fence coordinates to obtain an item video stream; The monitoring and identification execution terminal is used to clean and identify video streams of items, obtain abnormal items, generate abnormal status codes, and complete the abnormality resolution in conjunction with external execution mechanisms. The digital twin optimization terminal is used to create a digital production line based on the production line layout and the visual attention fence area, and to simulate and optimize the fence layout.

2. The logistics transportation system according to claim 1, characterized in that, The intelligent fence management terminal includes: The parameter collection module is used to generate physical structure parameters based on the three-dimensional structural diagram of the logistics conveyor frame and a preset construction error range; the physical structure parameters include frame height, support width, electric roller diameter, and center distance between adjacent rollers. The centerline determination module is used to generate the centerline of the transport path of the item based on the item size and the physical structure parameters. The area division module is used to expand bidirectionally based on the center line of the item transportation path and the actual conveyor line to obtain the visual attention fence area and determine the visual neglect area. The coordinate mapping module is used to generate a mapping transformation matrix based on the structure pixel size and the physical structure parameters, and to perform coordinate transformation on the visual attention fence area to calculate the actual attention fence coordinates.

3. The logistics transportation system according to claim 1, characterized in that, The dynamic visual acquisition terminal includes: The location triggering module is used to locate the transmission position of the items based on the transportation speed of the logistics conveyor, and generate a collection trigger command in combination with the actual coordinates of the fence of interest. The data acquisition module is used to collect data on the transportation process of the goods according to the acquisition trigger command, and obtain a global video stream; The blurring and segmentation module is used to blur the carrier and segment the target in the global video stream according to the visual neglect area to obtain the object video stream.

4. The logistics transportation system according to claim 1, characterized in that, The monitoring and identification execution terminal includes: The filtering and noise reduction module is used to filter noise from the video stream of the object to obtain a noise-free video stream; The image framing module is used to decompose the noiseless video stream into frames to obtain several frames of object images; The feature extraction module is used to perform edge detection and information extraction on the item image to obtain the item skeleton outline and transportation target information; The anomaly detection module is used to perform association matching between the skeleton outline of the item and the transportation target information based on a preset item feature information database, and calculate the corresponding feature anomaly value; If all the abnormal feature values ​​are greater than the preset abnormality judgment threshold, then the appearance features and transportation information of the current item are determined to be abnormal. If any of the abnormal feature values ​​is greater than the abnormality determination threshold, then the appearance feature of the current item or the transportation information is determined to be abnormal. The status determination module is used to perform temporal optical flow tracking on the object image and mark the object's movement trajectory; if the object's movement trajectory does not change, the current object's transmission status is determined to be abnormal. The early warning and notification module is used to collect information on abnormal items based on the type of abnormality and generate an abnormality status code. The anomaly resolution module is used to generate corresponding mechanism control instructions based on the anomaly status code, and to call external actuators to control the anomaly items for sorting, counting, or defect identification.

5. The logistics transportation system according to claim 1, characterized in that, The digital twin optimized terminal includes: The twin construction module is used to perform twin simulation of the electronic fence and logistics conveyor based on the production line layout, and generate an initial twin conveyor line; The data synchronization module is used to synchronize information of the initial twin conveyor line based on the visual attention fence area and the item transmission status to obtain the production line data requirements. The simulation optimization module is used to simulate and adjust the fence layout of the digital production line and generate a layout change evaluation table.

6. A data calibration method based on a logistics transportation system, applied to the system described in any one of claims 1-5, characterized in that, include: Extract the theoretical physical dimensions of the component's 3D model, combine them with a preset construction error range, generate physical structural parameters, and establish a component dimension reference library; By analyzing the component size reference library and combining it with the actual conveyor line, the center line of the item transportation path is determined, an electronic fence area is generated, and mapping calibration application data is extracted. Monitor the transport status of items within the electronic fence area, trigger image acquisition and anomaly recognition of the items, correct the mapping calibration data, and obtain corrected calibration application data.

7. The logistics transportation system according to claim 6, characterized in that, The process of analyzing the component size reference library and combining it with the actual conveyor line to determine the centerline of the goods transportation path, generating an electronic fence area, and extracting mapping calibration data includes: The actual conveyor line is scanned globally at a preset resolution to obtain images of the conveyor line components and to calculate the pixel-level parameters of the components. Based on the component size reference library, the physical structure parameters of the conveyor line component are extracted by matching the physical structure with the component image. Based on the support width and the center distance between adjacent rollers in the physical structural parameters, the center anchor point is calibrated. Based on the item size information, a smooth curve is fitted to the center anchor point to determine the centerline of the item transportation path; Based on the preset effective recognition width, the electronic fence area is generated by shifting to both sides and closing along the center line of the item transportation path. Based on the component pixel parameters and physical structure parameters, the electronic fence area is mapped to the actual coordinate space, and structured mapping calibration data is extracted.

8. The logistics transportation system according to claim 6, characterized in that, The monitoring of the transport status of goods within the electronic fence area triggers image acquisition and anomaly identification of the goods, corrects the mapping calibration data, and obtains corrected calibration data, including: The electronic fence area is divided according to the production line layout, and visually attention-focused fence areas are selected. Monitor the transportation status of items within the electronic fence area, determine the real-time location of the items, and calculate the coordinates of the item's center point; If the real-time location of the item is within the visual attention fence area, then cluster analysis is performed on the coordinates of the item's center point to obtain the item's distribution center, and the difference is calculated between the center point and the center line of the item's transportation path to obtain the center deviation value. If the center deviation value is greater than the preset center deviation amount, then the item is image acquired and anomaly is identified to generate the item transportation trajectory; The transport trajectory of the goods is fitted to a region to obtain the actual transport fence area, and the deviation is calculated between the actual transport fence area and the visual attention fence area to obtain the offset correction value. The mapping calibration data is corrected based on the offset correction value, and the corrected calibration data is calculated.

9. The data calibration method for a logistics transportation system according to claim 6, characterized in that, The mapping calibration data includes a fitting function for the centerline of the goods transportation path, a set of spatial feature anchor points for the visual attention fence area, and a coordinate transformation matrix for the electronic fence.

10. A storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the data calibration method for a logistics transportation system as described in any one of claims 6 to 9.

Images