An unmanned intelligent cargo handling system and method based on a portal crane

Abstract

The present application relates to the technical field of port logistics and cargo handling automation, in particular to an unmanned intelligent cargo handling system and method based on a portal crane. The system comprises a multi-source fusion positioning sensing and processing unit, a container identity recognition and box number analysis unit, a full-process cargo handling operation management and control unit, and an edge computing and cloud collaborative control unit. The present application provides accurate triggering basis for box number collection based on multi-source fusion positioning, adapts to the dynamic operation scene of the portal crane, and reduces the interference of the operation environment on box number collection and identification. By constructing a multi-level container identity verification mechanism, the dynamic binding and consistency verification of the box number, trailer identification and terminal operation plan are realized, meeting the high-precision identity verification requirements in dynamic scenarios. At the same time, the edge computing and cloud collaborative architecture is adopted to ensure the stable operation of the system, combined with remote monitoring, data tracing and self-learning optimization design, improving the adaptability and practicality of cargo handling automation.

Description

Technical Field

[0001] This invention relates to the field of automated cargo handling technology in port logistics, and more specifically, to an unmanned intelligent cargo handling system and method based on a gantry crane. Background Technology

[0002] Currently, container tallying operations at ports are a core component ensuring smooth logistics flow and traceable operational data. Gantry cranes, as the core equipment for container loading and unloading at ports, play a crucial role in container identification and data recording during their operation. With the development of port automation, visual recognition technology is gradually being applied to container number collection and verification. By deploying visual acquisition equipment within the gantry crane's operational environment, attempts are being made to replace manual labor in tasks such as container number identification and data recording, thereby improving the efficiency and standardization of tallying operations.

[0003] However, in the dynamic operation scenario of gantry cranes, container number recognition and identity verification still have significant adaptability deficiencies: container number acquisition relies heavily on single-view vision equipment, which is easily affected by complex environmental interferences such as changes in lighting, partial occlusion of the container body, and viewpoint shifts during dynamic operations such as lifting and rotating spreaders, making it difficult to guarantee the accuracy of container number recognition. Furthermore, container number recognition lacks a multi-level verification mechanism that integrates multi-view temporal data fusion, container number validity verification, and cross-verification of business data. It cannot achieve dynamic binding and consistency comparison with trailer identification and terminal operation plans, making it difficult to meet the high-precision container identity verification requirements of gantry crane dynamic operation scenarios, thus hindering further improvements in the automation level of tallying operations. Therefore, we propose an unmanned intelligent tallying system and method based on gantry cranes. Summary of the Invention

[0004] The purpose of this invention is to provide an unmanned intelligent cargo handling system and method based on a gantry crane, so as to solve the problem of insufficient adaptability of container number recognition and identity verification in dynamic operation scenarios of gantry cranes as mentioned in the background art.

[0005] To address the aforementioned technical problems, one objective of this invention is to provide an unmanned intelligent cargo handling system based on a gantry crane, comprising:

[0006] The multi-source fusion positioning perception and processing unit uses a dual-axis digital compass sensor installed on the gantry crane spreader to collect the absolute azimuth angle data of the gantry crane spreader, reuses the encoder of the gantry crane's own mechanism to collect the relative displacement data of each mechanism, and fuses and corrects the absolute direction reference and relative displacement data through a data fusion algorithm to output the real-time position and angle data of the gantry crane spreader.

[0007] The container identification and container number parsing unit uses a container number visual acquisition component and a vehicle number identification acquisition component installed on the gantry crane spreader to collect multi-view image data of the container and trailer identification data, respectively. The collected multi-view image data of the container is processed by character recognition to generate several candidate complete container numbers. The standard coding rules of container numbers are introduced to conduct a preliminary screening of the legality of the candidate complete container numbers. At the same time, the corresponding container number information and vehicle number information are collected at key nodes of container operation and key nodes of trailer operation. A dynamic binding relationship between operation cycle, trailer identification and container number is established. The candidate complete container numbers after preliminary screening are sorted and confirmed in combination with terminal operation plan data and vehicle number binding results. The container number recognition results are cross-validated in combination with the bound vehicle number information and terminal operation plan data.

[0008] The full-process cargo handling operation control unit automatically synchronizes the shift operation plan data from the terminal operating system and provides it to the container identification and container number parsing unit as terminal operation plan data. Based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit, the container number visual acquisition component of the container identification and container number parsing unit is triggered to perform multi-view image data acquisition of the container. At the container landing and gantry crane spreader unlocking nodes, the vehicle number identification acquisition component of the container identification and container number parsing unit is triggered to collect trailer identification data. The cross-verification logic of the container identification and container number parsing unit is executed, and the manual review results are fed back to the system as training data for the self-learning model.

[0009] The edge computing and cloud collaborative control unit deploys an edge computing industrial control computer at the gantry crane end, runs the data fusion algorithm of the multi-source fusion positioning perception and processing unit, the character recognition processing, container number parsing and visual recognition algorithm of the container identification and container number parsing unit, deploys a system server in the cloud to perform data interaction with the terminal operating system, receives the tallying business scheduling instructions of the full-process tallying operation control unit, pushes abnormal alarm information of the container identification and container number parsing unit, and stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit, the container identification and container number parsing unit, and the full-process tallying operation control unit;

[0010] The remote interaction and operation data traceability unit displays the cargo handling operation status of the full-process cargo handling operation control unit and the abnormal alarm information of the container identification and container number parsing unit through the remote cargo handling center monitoring screen and mobile application. It receives remote manual review instructions and feeds them back to the full-process cargo handling operation control unit. It encrypts and stores the full-link operation data stored in the edge computing and cloud collaborative control unit, and supports the retrieval of operation chain data records for any container.

[0011] As a further improvement to this technical solution, the multi-source fusion positioning sensing and processing unit includes an azimuth data acquisition module, a relative displacement data acquisition module, a data fusion correction module, and a positioning data output module, wherein:

[0012] The azimuth data acquisition module uses a dual-axis digital compass sensor installed on the gantry crane spreader to acquire the absolute azimuth data of the gantry crane spreader and transmit it to the data fusion and correction module.

[0013] The relative displacement data acquisition module reuses the encoders of the hoisting mechanism, the luffing mechanism, and the slewing mechanism of the gantry crane to collect relative displacement data of the hoisting height of the gantry crane spreader, relative displacement data of the boom pitch angle of the gantry crane, and relative displacement data of the slewing angle of the gantry crane, respectively, and transmits them to the data fusion and correction module.

[0014] The data fusion correction module receives absolute azimuth data transmitted by the azimuth data acquisition module, relative displacement data of lifting height, relative displacement data of boom pitch angle, and relative displacement data of slewing angle transmitted by the relative displacement data acquisition module. It then uses a data fusion algorithm to fuse and correct the absolute direction reference corresponding to the absolute azimuth data with the three types of relative displacement data, and transmits the corrected positioning data to the positioning data output module.

[0015] The positioning data output module receives the corrected positioning data transmitted by the data fusion and correction module, and outputs the real-time position and angle data of the gantry crane lifting device.

[0016] As a further improvement to this technical solution, the container identification and container number parsing unit includes a multi-source visual acquisition module, a container number character parsing module, a container number validity verification module, an identity binding and association module, and a container number identity verification module, wherein:

[0017] The multi-source visual acquisition module uses a container number visual acquisition component and a vehicle number identification acquisition component installed on the gantry crane spreader to acquire container multi-view image data and trailer identification data respectively. The container multi-view image data is transmitted to the container number character parsing module, and the trailer identification data is transmitted to the identity binding association module.

[0018] The container number character parsing module receives container multi-view image data transmitted by the multi-source visual acquisition module, performs character recognition processing on the container multi-view image data to generate several candidate complete container numbers, and transmits them to the container number legality verification module.

[0019] The container number validity verification module receives the candidate complete container number transmitted by the container number character parsing module, introduces the standard encoding rules of container numbers to perform a preliminary screening of the candidate complete container number for validity, and transmits the preliminary screened valid candidate container number to the container number identity verification module.

[0020] The identity binding and association module synchronously collects corresponding container number information and vehicle number information at key nodes of container operation and key nodes of trailer operation, establishes a dynamic binding relationship between operation cycle, trailer identification and container number and transmits it to the container number identity verification module.

[0021] The container number identity verification module receives the legitimate candidate container numbers transmitted by the container number legality verification module and the dynamic binding relationship transmitted by the identity binding association module. It sorts and confirms the legitimate candidate container numbers after initial screening in combination with the terminal operation plan data and vehicle number binding results. It also performs cross-verification of the container number identification results in combination with the bound vehicle number information and the terminal operation plan data.

[0022] As a further improvement to this technical solution, the multi-source visual acquisition module includes a container number image acquisition submodule, a container number image preprocessing submodule, a trailer identification acquisition submodule, and a trailer identification preprocessing submodule, wherein:

[0023] The container number image acquisition submodule is used to continuously acquire container number image data from multiple perspectives of the container at a preset acquisition frequency during the operation of lifting and rotating containers by the gantry crane spreader. The preset acquisition frequency is adapted and adjusted according to the operating speed of the gantry crane, and the value range is 5-15 frames / second. The multiple perspectives include at least two opposing perspectives of the container.

[0024] The container number image preprocessing submodule is used to sequentially perform Gaussian filtering noise reduction and contrast-limited adaptive histogram equalization contrast enhancement operations on the single frame container number image data acquired by the container number image acquisition submodule, and transmit the processed container number image data to the container number character parsing module.

[0025] The trailer identification acquisition submodule is used to acquire trailer identification image data at the container loading node and the gantry crane spreader unlocking node.

[0026] The trailer identification preprocessing submodule is used to sequentially perform Canny edge detection contour extraction and projection character segmentation operations on the trailer identification image data collected by the trailer identification acquisition submodule, and transmit the processed trailer identification data to the identity binding association module.

[0027] As a further improvement to this technical solution, the process by which the container number character parsing module performs character recognition processing on multi-view image data of containers to generate several candidate complete container numbers includes the following steps:

[0028] S22.1. According to the ISO 6346 standard, the container number identification target is decomposed into the owner code segment, equipment category identifier segment, serial number segment, and check code segment. The first segment in a single-view, single-frame image is recorded. The recognition result of each character segment is ;

[0029] S22.2. Locate the local character segments in each frame of the image and calculate the segment-level confidence score for each character segment. Filter out valid character segments whose confidence level meets the preset threshold;

[0030] S22.3. Perform time-sequential splicing of valid character segments from different frames to generate several candidate complete box numbers. And calculate the overall confidence of the candidate complete box number based on the fragment-level confidence. .

[0031] As a further improvement to this technical solution, the process of the container number validity verification module performing preliminary screening of candidate complete container numbers includes the following steps:

[0032] S23.1, For candidate complete container numbers Perform format compliance verification and eliminate candidate box numbers that do not conform to the ISO6346 standard format;

[0033] S23.2 Perform a checksum operation on the candidate complete container numbers that meet the format requirements, and convert the first 10 characters of the container number into a numeric value. Based on numerical values Calculate the check code ;

[0034] S23.3. Compare the calculated check code with the check code segment of the candidate complete box number. If the comparison is consistent, it is determined to be a valid candidate box number.

[0035] As a further improvement to this technical solution, the container number identity verification module sorts and confirms legitimate candidate container numbers in conjunction with terminal operation plan data and vehicle number binding results, and completes cross-verification, including the following steps:

[0036] S25.1 Set three weighted dimensions: overall confidence level, planned character matching degree, and vehicle number association matching degree, and assign weights to each weighted dimension;

[0037] S25.2 Overall confidence level based on candidate complete container numbers Calculate the complete box number of the legal candidate based on the matching degree of each dimension. Overall score ;

[0038] S25.3, Based on the overall score The valid candidate container numbers are sorted in descending order, and the candidate container number with the highest score is selected as the target container number. The final selection is based on the overall score. Execute the corresponding judgment operation;

[0039] S25.4. Perform a consistency comparison between the target container number and the terminal operation plan data and the planned trailer information corresponding to the bound vehicle number to complete the cross-verification of the container number identity.

[0040] As a further improvement to this technical solution, the full-process cargo handling operation control unit includes an operation plan synchronization module, a container number acquisition trigger module, a vehicle number acquisition trigger module, a verification logic execution module, and a verification data feedback module, wherein:

[0041] The operation plan synchronization module automatically synchronizes the shift operation plan data from the terminal operating system and provides the shift operation plan data to the container identification and container number parsing unit;

[0042] The container number acquisition triggering module triggers the container number visual acquisition component of the container identification and container number parsing unit to perform multi-view image data acquisition of the container based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit.

[0043] The vehicle number acquisition triggering module triggers the vehicle number identification acquisition component of the container identification and container number parsing unit to collect trailer identification data at the container loading and gantry crane spreader unlocking nodes.

[0044] The verification logic execution module executes the cross-verification logic of the container identification and container number parsing unit;

[0045] The verification data feedback module feeds back the manual verification results to the edge computing and cloud collaborative control unit as training data for the self-learning model.

[0046] As a further improvement to this technical solution, the edge computing and cloud collaborative control unit includes an edge computing deployment module, a cloud interaction module, a scheduling instruction receiving module, an anomaly alarm push module, and a full-link data storage module, wherein:

[0047] The edge computing deployment module deploys an edge computing industrial control computer at the end of the gantry crane, running the data fusion algorithm of the multi-source fusion positioning perception and processing unit, and the character recognition processing, container number parsing and visual recognition algorithm of the container identification and container number parsing unit.

[0048] The cloud interaction module deploys a system server in the cloud and performs data interaction with the dock operating system.

[0049] The scheduling instruction receiving module receives the cargo handling business scheduling instructions from the full-process cargo handling operation control unit;

[0050] The abnormal alarm push module pushes abnormal alarm information from the container identification and container number parsing unit;

[0051] The full-link data storage module stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit, the container identification and container number parsing unit, and the full-process cargo handling operation control unit.

[0052] The second objective of this invention is to provide an unmanned intelligent cargo handling method based on a gantry crane. The unmanned intelligent cargo handling system based on the aforementioned gantry crane includes the following steps:

[0053] S1. The absolute azimuth angle data of the lifting device is collected by the dual-axis digital compass sensor installed on the lifting device of the gantry crane. The relative displacement data of each mechanism is collected by the encoder of the gantry crane's own mechanism. After the absolute direction reference and relative displacement data are fused and corrected by the data fusion algorithm, the real-time position and angle data of the lifting device of the gantry crane are output.

[0054] S2. Automatically synchronize the shift operation plan data from the terminal operating system, and use the shift operation plan data as the terminal operation plan data for subsequent container number identification and verification.

[0055] S3. Based on the real-time position and angle of the gantry crane spreader output in step S1, trigger the acquisition of multi-view image data of the container. Perform character recognition processing on the acquired multi-view image data of the container to generate several candidate complete container numbers. Introduce the standard coding rules for container numbers to perform a preliminary screening of the legality of the candidate complete container numbers and obtain legal candidate container numbers.

[0056] S4. At the container loading and gantry crane spreader unlocking nodes, trigger the collection of trailer identification data, and simultaneously collect the corresponding container number information and vehicle number information at the key nodes of container operation and trailer operation, and establish a dynamic binding relationship between operation cycle, trailer identification and container number.

[0057] S5. The legal candidate container numbers obtained in step S3 are combined with the terminal operation plan data and vehicle number binding results to perform a weighted comprehensive score sorting to confirm the target container number. The target container number is then compared with the terminal operation plan data and the planned trailer information corresponding to the bound vehicle number to complete the cross-verification of the container number identity.

[0058] S6. Runs data fusion algorithms, character recognition processing, container number parsing and visual recognition algorithms at the gantry crane end, performs data interaction with the terminal operating system in the cloud, receives tallying business scheduling instructions, pushes abnormal alarm information in the container number recognition and verification process, and stores the full-link operation data generated in the entire tallying operation process.

[0059] S7. Displays the status of cargo handling operations and abnormal alarm information for container number identification and verification through a remote cargo handling center monitoring screen and mobile application. Receives remote manual review instructions and feeds them back into the cargo handling process. Encrypts and stores data across the entire operation chain and supports retrieving operation chain data records for any container.

[0060] S8. Feed back the results of remote manual review as training data for the self-learning model of container number character recognition, container number parsing and identity cross-verification.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] This invention relies on multi-source fusion positioning to provide accurate triggering basis for container number acquisition, adapting to the dynamic operation scenarios of gantry cranes and reducing interference from the operating environment on container number acquisition and recognition. By constructing a multi-level container identity verification mechanism, it realizes dynamic binding and consistency verification of container numbers with trailer identification and terminal operation plans, meeting the high-precision identity verification requirements in dynamic operation scenarios. At the same time, combined with the architecture design of edge computing and cloud collaboration, it ensures the stable operation of the system in dynamic operation scenarios. With the assistance of remote monitoring, data traceability and self-learning optimization design, the container number recognition and identity verification links are more adapted to the operating characteristics of gantry cranes, improving the adaptability and practicality of automated cargo handling operations. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the system framework of the present invention;

[0064] The meanings of the labels in the diagram are as follows:

[0065] 1. Multi-source fusion positioning sensing and processing unit; 11. Azimuth angle data acquisition module; 12. Relative displacement data acquisition module; 13. Data fusion correction module; 14. Positioning data output module;

[0066] 2. Container identification and container number parsing unit; 21. Multi-source visual acquisition module; 22. Container number character parsing module; 23. Container number validity verification module; 24. Identity binding and association module; 25. Container number identity verification module;

[0067] 3. Full-process cargo handling operation control unit; 31. Operation plan synchronization module; 32. Container number acquisition trigger module; 33. Vehicle number acquisition trigger module; 34. Verification logic execution module; 35. Review data feedback module;

[0068] 4. Edge computing and cloud-based collaborative control unit; 41. Edge computing deployment module; 42. Cloud-based interaction module; 43. Scheduling instruction receiving module; 44. Anomaly alarm push module; 45. End-to-end data storage module;

[0069] 5. Remote interaction and operation data traceability unit. Detailed Implementation

[0070] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0071] like Figure 1 As shown, this embodiment provides an unmanned intelligent cargo handling system based on a gantry crane, including:

[0072] The multi-source fusion positioning sensing and processing unit 1 uses a dual-axis digital compass sensor installed on the gantry crane's lifting device to collect the absolute azimuth data of the gantry crane's lifting device, reuses the encoder of the gantry crane's own mechanism to collect the relative displacement data of each mechanism, and fuses and corrects the absolute direction reference and relative displacement data through a data fusion algorithm to output the real-time position and angle data of the gantry crane's lifting device. The multi-source fusion positioning sensing and processing unit 1 includes an azimuth data acquisition module 11, a relative displacement data acquisition module 12, a data fusion correction module 13, and a positioning data output module 14, wherein:

[0073] In this embodiment, the azimuth data acquisition module 11 uses a dual-axis digital compass sensor installed on the gantry crane's lifting device to acquire the absolute azimuth data of the gantry crane's lifting device and transmit it to the data fusion correction module 13. This provides absolute direction reference data for the multi-source fusion positioning perception and processing unit 1, eliminates the cumulative errors that are easily generated by relative displacement data, and ensures the real-time performance and effectiveness of azimuth data acquisition. The specific implementation is as follows:

[0074] The dual-axis digital compass sensor is fixedly installed at the center of the top of the lifting device, where there is no strong electromagnetic interference and no mechanical obstruction. The azimuth data acquisition module 11 establishes a data connection with the data fusion correction module 13 using the RS485 industrial serial communication protocol. It continuously acquires the horizontal azimuth and pitch azimuth data of the lifting device at a frequency of 10Hz. Each set of acquired data synchronously generates a standardized timestamp. The timestamp and azimuth data form a data group and are transmitted together to the data fusion correction module 13.

[0075] In this embodiment, the relative displacement data acquisition module 12 reuses the encoders of the hoisting mechanism, luffing mechanism, and slewing mechanism of the gantry crane to collect relative displacement data of the lifting height of the gantry crane's spreader, relative displacement data of the boom pitch angle, and relative displacement data of the slewing angle of the gantry crane, respectively, and transmits them to the data fusion and correction module 13. This fully utilizes the existing hardware of the crane, avoiding the cost of adding additional displacement sensors. At the same time, it collects relative displacement data reflecting the multi-dimensional motion state of the spreader, providing complete motion parameters for data fusion and correction. The specific implementation is as follows:

[0076] The relative displacement data acquisition module 12 directly connects to the PROFIBUS industrial bus of the original control system of the gantry crane, retrieves the real-time pulse signals of the encoders of the hoisting mechanism, the luffing mechanism, and the slewing mechanism, and converts the pulse signals into relative displacement data of hoisting height, relative displacement data of boom pitch angle, and relative displacement data of slewing angle based on the transmission ratio and mechanical geometric parameters of each mechanism. The acquisition frequency is consistent with that of the azimuth angle data acquisition module 11 at 10Hz. Each set of data generates a timestamp with the same format as the azimuth angle data and transmits it to the data fusion and correction module 13 through the PROFIBUS bus.

[0077] In this embodiment, the data fusion correction module 13 receives absolute azimuth data transmitted by the azimuth data acquisition module 11, and relative displacement data of lifting height, boom pitch angle, and slewing angle transmitted by the relative displacement data acquisition module 12. It then uses a data fusion algorithm to fuse and correct the absolute direction reference corresponding to the absolute azimuth data with the three types of relative displacement data. The corrected positioning data is then transmitted to the positioning data output module 14, achieving normalization processing and error complementarity of multi-source positioning data. The algorithmic fusion eliminates the deviation problem of single data points, outputting accurate raw positioning data for the lifting device. The specific implementation is as follows:

[0078] The data fusion and correction module 13 is deployed on the edge computing industrial control computer at the gantry crane end. First, it performs time axis alignment processing on all received data based on the timestamp, eliminating asynchronous data caused by transmission delay. Then, it establishes a local operation coordinate system adapted to the dock operation with the gantry crane rotation center as the coordinate origin. The absolute azimuth data and the three types of relative displacement data are mapped to this coordinate system through coordinate transformation formulas to complete the coordinate system unification of multi-source data. Finally, the Kalman filter algorithm is used to iteratively fuse the absolute azimuth data as the observation and the three types of relative displacement data as the prediction. After completing the data correction, the corrected positioning data is transmitted to the positioning data output module 14.

[0079] In this embodiment, the positioning data output module 14 receives the corrected positioning data transmitted by the data fusion and correction module 13, outputs the real-time position and angle data of the gantry crane spreader, and performs standardization and validity processing on the corrected positioning data to ensure the compatibility of the output data so that it can be accurately identified and called by the full-process cargo handling operation control unit 3. At the same time, it realizes the synchronous storage of positioning data, which is specifically implemented as follows:

[0080] The positioning data output module 14 first performs integrity verification on the received calibrated positioning data, eliminating invalid data caused by transmission or calculation anomalies, and abnormal data exceeding the operating range of the gantry crane; then it converts the valid data into a system-unified 32-bit floating-point standardized format, the data content of which includes the three-dimensional spatial position of the spreader in the local operating coordinate system, the spreader rotation angle, and the spreader pitch angle; finally, it transmits the standardized real-time position and angle data synchronously to the full-process cargo handling operation control unit 3 and the edge computing and cloud collaborative control unit 4 via the industrial Ethernet TCP / IP protocol at an update frequency of 10Hz.

[0081] The container identification and container number parsing unit 2 uses a container number visual acquisition component and a vehicle number identification acquisition component installed on the gantry crane spreader to collect multi-view image data of containers and trailer identification data, respectively. It performs character recognition processing on the collected multi-view image data to generate several candidate complete container numbers. It introduces standard container number coding rules to perform initial screening for the legality of the candidate complete container numbers. Simultaneously, it collects corresponding container number information and vehicle number information at key nodes of container operations and key nodes of trailer operations, establishing a dynamic binding relationship between operation cycles, trailer identification, and container numbers. It sorts and confirms the candidate complete container numbers after initial screening by combining them with terminal operation plan data and vehicle number binding results. It cross-validates the container number recognition results by combining the bound vehicle number information with terminal operation plan data. The container identification and container number parsing unit 2 includes a multi-source visual acquisition module 21, a container number character parsing module 22, a container number legality verification module 23, an identity binding association module 24, and a container number identity verification module 25.

[0082] Specifically, the core hardware of the container identification and container number parsing unit 2 includes a container number visual acquisition component and a vehicle number identification acquisition component deployed on the gantry crane spreader, as well as an edge computing industrial control computer deployed on the gantry crane end. The container number visual acquisition component and the vehicle number identification acquisition component are both industrial-grade high-definition visual acquisition devices, equipped with anti-glare, dustproof, and vibration-resistant protective accessories and dedicated fixed-focus lenses, and are connected to the edge computing industrial control computer through waterproof aviation connectors and shielded data cables. The multi-source visual acquisition module 21, container number character parsing module 22, container number legality verification module 23, identity binding association module 24, and container number identity verification module 25 of the container identification and container number parsing unit 2 are all deployed on the edge computing industrial control computer in the form of software algorithms, relying on the computing power of the industrial control computer to complete the entire process of data processing and calculation.

[0083] Furthermore, the container identification and container number parsing unit 2, the multi-source fusion positioning perception and processing unit 1, the full-process tallying operation control unit 3, and the edge computing and cloud collaborative control unit 4 achieve bidirectional data interaction through the industrial Ethernet TCP / IP protocol. It receives terminal operation plan data and collection trigger instructions issued by the full-process tallying operation control unit 3, receives real-time position and angle data of the spreader output by the multi-source fusion positioning perception and processing unit 1, feeds back the container number identification and cross-verification results to the full-process tallying operation control unit 3, and synchronizes the full-process operation data and abnormal alarm information to the edge computing and cloud collaborative control unit 4.

[0084] In addition, the multi-source visual acquisition module 21, the container number character parsing module 22, the container number legality verification module 23, the identity binding association module 24, and the container number identity verification module 25 achieve directional one-way data transmission through the local high-speed PCIe data interface of the industrial control computer. The data transmission end and the receiving end correspond one-to-one, forming a serial data link of "acquisition-parsing-verification-binding-final verification". All internal and external transmitted data carry millisecond-level unified format timestamps and are attached with spreader position angle tags and operation node identifiers to ensure the time sequence consistency and scenario relevance of the data throughout the process. Each module is triggered to work according to the unified scheduling instructions of the full-process cargo handling operation control unit 3. When there are no instructions, it is in a low-power standby state to adapt to the intermittent operation rhythm of the gantry crane.

[0085] In this embodiment, the multi-source visual acquisition module 21 uses a container number visual acquisition component and a vehicle number identification acquisition component installed on the gantry crane spreader to acquire container multi-view image data and trailer identification data, respectively. The container multi-view image data is transmitted to the container number character parsing module 22, and the trailer identification data is transmitted to the identity binding and association module 24. The multi-source visual acquisition module 21 includes a container number image acquisition submodule, a container number image preprocessing submodule, a trailer identification acquisition submodule, and a trailer identification preprocessing submodule, wherein:

[0086] The container number image acquisition submodule is used to continuously acquire container number image data from multiple perspectives of the container at a preset acquisition frequency during the lifting and slewing operations of the gantry crane spreader. The preset acquisition frequency is adjusted according to the operating speed of the gantry crane, with a value range of 5-15 frames / second. The multiple perspectives include at least two opposing perspectives of the container, providing high-quality raw image data that covers multiple perspectives of the container and is sequentially continuous for subsequent container number character parsing. By adapting the acquisition frequency to the crane's operating speed, character recognition errors caused by motion blur are avoided. At the same time, the acquisition of opposing perspectives eliminates the problem of single-view container number occlusion or incomplete recognition. The specific implementation is as follows:

[0087] Specifically, the hardware carrier of the container number image acquisition submodule is the container number visual acquisition component. This component can use a 2-megapixel industrial-grade high-definition CMOS camera, equipped with an 8mm fixed-focus lens and anti-glare and dustproof optical filters. It is installed on the side of the gantry crane spreader in a position where there is no mechanical obstruction and no operational interference. The angle between the lens optical axis and the container number printing plane is controlled between 30° and 60° to ensure that the container number characters are completely captured in the image.

[0088] Meanwhile, the acquisition trigger signal of the container number image acquisition submodule is issued by the full-process cargo handling operation control unit 3. The trigger condition is that the spreader position data output by the multi-source fusion positioning perception and processing unit 1 meets the condition that "the spreader lifting height is ≥2m and the rotation angle is ≥5°", that is, the acquisition is started after entering the container lifting or rotation operation stage.

[0089] Furthermore, the adaptive adjustment rule for the preset sampling frequency is as follows:

[0090] When the gantry crane operates at speed At m / s, the sampling frequency Frames per second;

[0091] when At m / s, the sampling frequency Frames per second;

[0092] when At m / s, the sampling frequency Frames per second;

[0093] Work speed The displacement data of the lifting device output by the multi-source fusion positioning sensing and processing unit 1 is calculated using the timestamp. The calculation formula is as follows:

[0094] ;

[0095] in:

[0096] express The three-dimensional coordinates of the lifting device in the local operating coordinate system at all times; and the local operating coordinate system is established uniformly by the multi-source fusion positioning perception and processing unit 1 with the rotation center of the gantry crane as the origin;

[0097] express The three-dimensional coordinates of the spreader in the local operating coordinate system at any given time;

[0098] The timestamp represents two consecutive location data points, in seconds.

[0099] In addition, during the acquisition process, the container number image acquisition submodule synchronously reads the spreader rotation angle data output by the multi-source fusion positioning perception and processing unit 1. When the spreader rotation angle difference is ≥90°, it determines the opposing view of the container and ensures that at least two sets of container number images, such as the front and back or the left and right sides of the container, are acquired. Each frame of the acquired container number image data carries a timestamp, spreader 3D coordinates, spreader rotation angle, and view type label. It is stored in JPEG format with a resolution of 1920×1080 and a bit depth of 8-bit grayscale image. It is transmitted to the container number image preprocessing submodule through the local high-speed PCIe interface.

[0100] The container number image preprocessing submodule performs Gaussian filtering for noise reduction and contrast-limited adaptive histogram equalization for contrast enhancement on the single-frame container number image data acquired by the container number image acquisition submodule. The processed container number image data is then transmitted to the container number character parsing module 22. This process eliminates image noise and uneven lighting in the port's outdoor environment, enhances the edge and texture features of the container number characters, improves the accuracy and robustness of subsequent OCR character recognition, and provides the container number character parsing module 22 with clear and recognizable container number image data. The specific implementation is as follows:

[0101] Specifically, the container number image preprocessing submodule first performs Gaussian filtering for noise reduction. A 5×5 Gaussian kernel is used to convolve the single-frame grayscale image of the container number to eliminate salt-and-pepper noise and Gaussian noise in the port's outdoor environment. The formula for generating the Gaussian kernel is:

[0102] ;

[0103] in:

[0104] The coordinates in the Gaussian kernel are... The weight value of the location;

[0105] This represents the standard deviation of the Gaussian kernel, with a value of 1.2.

[0106] , The coordinates of the center of the Gaussian kernel are represented as follows: (Corresponding to the center position of the 5×5 cores);

[0107] Represents the pixel coordinates within the Gaussian kernel, with a value range of 0-4;

[0108] During convolution, the Gaussian kernel is multiplied pixel-by-pixel with the image and the results are summed to obtain the grayscale values ​​of the filtered image. The calculation formula is:

[0109] ;

[0110] in:

[0111] This represents the gray value at coordinates (x, y) in the filtered image;

[0112] This represents the grayscale value at the corresponding location in the original image;

[0113] Represents the coordinates in the Gaussian kernel The weight value;

[0114] Simultaneously, the box number image preprocessing submodule performs contrast-limited adaptive histogram equalization (CLAHE) to divide the filtered image into 8×8 non-overlapping sub-blocks. The histogram equalization operation for each sub-block is constrained by a contrast-limited threshold to avoid local overexposure or underexposure. First, the histogram of each sub-block is calculated. The portion of the histogram exceeding the threshold is cropped and evenly distributed to other gray levels. Then, equalization is performed on the cropped histogram. (Contrast-limited threshold...) The calculation formula is:

[0115] ;

[0116] in:

[0117] This represents the total number of pixels in a single sub-block, i.e., 8×8=64;

[0118] This represents the restriction factor, with a value of 40;

[0119] This represents the total number of gray levels in the image, with a value of 256.

[0120] Furthermore, after equalization, the equalization results of each sub-block are stitched together into a complete image through bilinear interpolation to obtain enhanced container number image data. This data maintains the same resolution and format as the original image, carries the original timestamp and label information, and is transmitted to the container number character parsing module 22.

[0121] The trailer identification acquisition submodule is used to acquire trailer identification image data at the container loading node and the gantry crane spreader unlocking node. This provides raw visual data for the subsequent identity binding association module 24 to establish a dynamic binding relationship between the operation cycle, trailer identification, and container number. Data is only acquired at key nodes to reduce data redundancy and improve processing efficiency. The specific implementation is as follows:

[0122] Specifically, the hardware carrier of the trailer identification acquisition submodule is the vehicle number identification acquisition component, which is a 1.3-megapixel industrial-grade high-definition CMOS camera equipped with a 6mm fixed-focus lens and a dustproof and vibration-resistant protective shell. It is installed on the side of the gantry crane spreader near the trailer. The angle between the lens optical axis and the trailer identification affixation plane is controlled between 45° and 60° to ensure that the trailer identification is fully captured in the image.

[0123] Meanwhile, the acquisition trigger signal of the trailer identification acquisition submodule is provided by the limit switch and spreader unlocking sensor of the gantry crane. When the signal of "spreader in place" or "spreader unlocking action completed" is detected, an acquisition is immediately triggered. Only one valid image is acquired at each key node, and the acquisition frequency is fixed at 1 frame / time.

[0124] In addition, the collected trailer identification image data carries a timestamp, node type label ("carrying" or "unlocked"), and spreader location label. It is stored in JPEG format with a resolution of 1280×720 and a bit depth of 8-bit grayscale image. It is transmitted to the trailer identification preprocessing submodule through the local high-speed PCIe interface.

[0125] The trailer identification preprocessing submodule performs Canny edge detection contour extraction and projection character segmentation operations sequentially on the trailer identification image data acquired by the trailer identification acquisition submodule. It then transmits the processed trailer identification data to the identity binding and association module 24. This process highlights the contour features of the trailer identification characters, separates individual character regions, and eliminates background interference, providing structured character data for subsequent trailer identification character recognition and identity binding. This improves the accuracy and efficiency of identity binding. The specific implementation is as follows:

[0126] Specifically, the trailer marking preprocessing submodule first performs Canny edge detection contour extraction, and then uses a 3×3 Sobel operator to calculate the horizontal gradient of the image. with vertical gradient The gradient calculation formula is:

[0127] ;

[0128] ;

[0129] in:

[0130] This represents the gradient value in the horizontal direction;

[0131] This represents the gradient value in the vertical direction;

[0132] Represents the coordinates in the original trailer identification image. grayscale value;

[0133] This represents the convolution operation;

[0134] Simultaneously, the gradient magnitude is calculated based on the horizontal and vertical gradients. With gradient direction The calculation formula is:

[0135] ;

[0136] ;

[0137] in:

[0138] Indicates the gradient direction, with a value range of 1. ;

[0139] Then, non-maximum suppression is performed, retaining only pixels with local maxima along the gradient direction; finally, a double thresholding method is used to filter edges, with the higher threshold... low threshold The gradient magnitude is greater than The pixels are identified as strong edges, and the gradient magnitude is within and Pixels that are connected to strong edges are identified as weak edges, and the remaining pixels are identified as non-edges, thus obtaining the edge contour image of the trailer logo;

[0140] Furthermore, the trailer identification preprocessing submodule performs projection character segmentation, performs vertical projection on the edge contour image, and calculates the sum of the number of edge points in each column of pixels. The calculation formula is:

[0141] ;

[0142] in:

[0143] Indicates the first The vertical projection value of the column;

[0144] Indicates the height of the image;

[0145] Represents the coordinates in the edge contour image The pixel value is 1 for edge points and 0 for non-edge points;

[0146] Finally, iterate through all columns. The image is divided into multiple character regions using the local minimum value as the dividing point. Each character region corresponds to a character in the trailer identifier. The segmented character regions are stored in the form of binary images, carrying the original timestamp and node type label, and transmitted to the identity binding and association module 24.

[0147] In this embodiment, the container number character parsing module 22 receives multi-view image data of the container transmitted by the multi-source visual acquisition module 21, performs character recognition processing on the multi-view image data to generate several candidate complete container numbers, and transmits them to the container number validity verification module 23. The process by which the container number character parsing module 22 performs character recognition processing on the multi-view image data of the container to generate several candidate complete container numbers includes the following steps:

[0148] S22.1. According to the ISO 6346 standard, the container number identification target is decomposed into the owner code segment, equipment category identifier segment, serial number segment, and check code segment. The first segment in a single-view, single-frame image is recorded. The recognition result of each character segment is Following the internationally recognized ISO 6346 standard, the encoding rules and boundaries of each character segment of the container number are clearly defined. Standardized symbols are used to mark the character segment recognition results from a single viewpoint and single frame, enabling structured classification and end-to-end traceability of the recognition data. This lays a unified data source foundation for subsequent confidence calculations and time-series stitching. The specific implementation is as follows:

[0149] Specifically, the container number character parsing module 22 first retrieves the container number encoding specification in the ISO 6346 standard, and splits the 11-digit standard container number into four fixed character segments: all personnel code segment (1st-3rd digits, letter + number combination), equipment category identifier segment (4th digit, fixed letter), serial number segment (5th-10th digits, pure numbers), and check code segment (11th digit, numbers / letters).

[0150] Simultaneously, the character segment recognition results of a single-view, single-frame image are symbolically defined, and the character segment in the single-view, single-frame image is denoted as the... The recognition result of each character segment is ;in: This represents the character segment number, with values ​​of 1, 2, 3, and 4, corresponding to the owner code segment, device category identifier segment, serial number segment, and checksum segment, respectively. This indicates the number of times the same character segment is recognized, and its value is [value]. , This represents the total number of times the same character segment is recognized in multi-view, multi-frame images; Indicates the first The first character segment The character content identified this time strictly matches the ISO6346 encoding rules of the corresponding character segment;

[0151] In addition, the box number character parsing module 22 adds viewpoint number and frame sequence number tags to each frame of image transmitted by the multi-source visual acquisition module 21 to ensure... It can accurately trace back to the acquisition perspective and frame sequence of the original image.

[0152] S22.2. Locate the local character segments in each frame of the image and calculate the segment-level confidence score for each character segment. The system filters out valid character segments with confidence levels meeting a preset threshold. It then uses segment position marking to pinpoint the precise recognition region to eliminate background interference. Multi-dimensional feature weighting is used to calculate segment-level confidence levels to quantify the reliability of the recognition results. Finally, threshold filtering eliminates low-confidence results, ensuring that subsequent candidate box numbers are composed of highly reliable valid character segments. The specific implementation is as follows:

[0153] Specifically, the container number character parsing module 22 first performs character segment position calibration on each frame of the preprocessed container number image. Based on the container number printing and layout characteristics, it locates the rectangular pixel coordinate range of each character segment. ;in: Indicates the first The horizontal coordinate of the top left corner of the pixel region of each character segment; Indicates the first The top-left y-coordinate of the pixel region of each character segment; Indicates the first The horizontal coordinate of the bottom right corner of the pixel region of each character segment; Indicates the first The bottom right ordinate of the pixel region of each character segment;

[0154] Simultaneously, based on the recognition results and image features within the calibration area, the segment-level confidence level of each character segment is calculated. The calculation formula is:

[0155] ;

[0156] in:

[0157] Represents the first frame in a single-view image. The first character segment The segment-level confidence score of the identification result, with a value range of [0,1];

[0158] This represents the character matching score weight coefficient, with a value of 0.5.

[0159] Indicates the first The first character segment The feature matching value of the recognition result with the standard character library ranges from [0,1].

[0160] This represents the image sharpness weighting coefficient, with a value of 0.3.

[0161] Indicates the first The image sharpness of the character segment labeled area is obtained by normalizing the mean value of the image gradient magnitude, and the value range is [0,1].

[0162] This represents the texture feature similarity weight coefficient, with a value of 0.2.

[0163] Indicates the first The similarity between the texture features of the character segment recognition area and the texture of the container number printing is in the range of [0,1].

[0164] and Satisfy the normalization condition of the weight coefficients: ;

[0165] Furthermore, the box number character parsing module 22 sets the fragment-level confidence threshold to 0.7, and... The character segment recognition result is determined to be a valid character segment. The identification results are discarded, and only the highly reliable identification results are retained.

[0166] S22.3. Perform time-sequential splicing of valid character segments from different frames to generate several candidate complete box numbers. And calculate the overall confidence of the candidate complete box number based on the fragment-level confidence. The system concatenates valid character segments according to the character segment order of the ISO 6346 standard to generate complete candidate box numbers. It then calculates the overall confidence metric for each candidate box number using the arithmetic mean method, quantifying the overall reliability of each candidate box number. This provides a high-quality candidate dataset for subsequent box number validity verification. The specific implementation is as follows:

[0167] Specifically, the container number character parsing module 22 sequentially concatenates all filtered valid character segments in the order of owner code segment → equipment category identifier segment → serial number segment → check code segment, and combines the valid recognition results of different character segments to generate several 11-digit complete container numbers, which are recorded as candidate complete container numbers. ;in: This represents the candidate box number, with a value of [value]. , This represents the total number of candidate box numbers generated after splicing. Indicates the first Each candidate complete box number is formed by sequentially concatenating four valid character segments;

[0168] Simultaneously, based on the fragment-level confidence of each character segment, the overall confidence of each candidate complete box number is calculated. The calculation formula is:

[0169] ;

[0170] in:

[0171] Indicates the first The overall confidence level of each candidate complete box number, with a value range of [0,1];

[0172] This indicates the total number of character segments that make up the container number, consistent with the breakdown result of the ISO 6346 standard;

[0173] This indicates the segment number, with values ​​of 1, 2, 3, and 4.

[0174] Indicates the first The candidate box number is the first The fragment-level confidence of the valid character segments corresponding to each character segment, and satisfying the following conditions: ;

[0175] Finally, all candidate complete box numbers With the corresponding overall confidence level The associated storage is a structured dataset, carrying the source information of each character segment, and is transmitted to the box number validity verification module 23.

[0176] In this embodiment, the container number validity verification module 23 receives the candidate complete container numbers transmitted by the container number character parsing module 22, introduces the standard encoding rules for container numbers to perform a preliminary screening of the candidate complete container numbers for validity, and transmits the preliminary screened valid candidate container numbers to the container number identity verification module 25. The process of the container number validity verification module 23 performing the preliminary screening of the candidate complete container numbers for validity includes the following steps:

[0177] S23.1, For candidate complete container numbers Perform format compliance verification, eliminate candidate container numbers that do not conform to the ISO 6346 standard format, and, based on the container number coding format specification of the ISO 6346 standard, verify the complete container numbers of the candidates. Multi-dimensional format verification is conducted to eliminate invalid candidate box numbers with incorrect bit length, character type violations, or structural mismatches. This ensures that all candidate box numbers entering the checksum calculation stage conform to the standard encoding structure, avoiding unnecessary computational waste and improving overall verification efficiency. The specific implementation is as follows:

[0178] Specifically, the container number validity verification module 23 preloads the container number format rules of the ISO 6346 standard, and verifies each candidate complete container number. Execute the following verification logic:

[0179] Bit verification: Verification If the total length of the characters is 11, and the length is not equal to 11, then it is directly marked as an invalid candidate box number and removed.

[0180] Character field type validation:

[0181] The character type is verified segment by segment according to the four-segment structure of ISO 6346, as follows:

[0182] The first three characters (all code segments): check if the character is AZ (uppercase letter) or 0-9 (number); if other characters are present, they are discarded.

[0183] The 4th character (equipment category identifier): Check if the character is U / J / Z (the standard equipment category letters for containers). If it is any other character, it will be rejected.

[0184] The 5th to 10th digits (serial number segment): check if the character is a pure number 0-9. If there are letters or other symbols, they will be removed.

[0185] The 11th bit (check code segment): checks whether the character is a number 0-9. If it is a letter or other symbol, it is discarded.

[0186] Structural verification: Confirmed The character field order strictly follows the ISO 6346 standard structure of "3-digit owner code + 1-digit device category + 6-digit serial number + 1-digit checksum". Candidate complete box numbers that pass verification are discarded. Enter S23.2 to perform check code calculation. Candidate box numbers that fail the check are marked as invalid and logged (including box number content and reason for failure), and are not included in subsequent processing.

[0187] S23.2 Perform a checksum operation on the candidate complete container numbers that meet the format requirements, and convert the first 10 characters of the container number into a numeric value. Based on numerical values Calculate the check code Strictly adhering to the ISO 6346 standard check code algorithm logic, for candidate complete box numbers with compliant format. Perform a checksum calculation, mapping the first 10 characters to standard numeric values. The theoretical check code is obtained by calculating the weighted sum and modulo. This provides an accurate benchmark value for subsequent checksum comparison, ensuring that the verification logic is completely consistent with international standards. The specific implementation is as follows:

[0188] Specifically, the container number validity verification module 23 first loads the character-numeric mapping rules of the ISO 6346 standard, and then verifies the candidate complete container numbers. The first 10 characters (denoted as Perform the following operations:

[0189] Character-to-number mapping: mapping the first 10 characters ( Convert to the corresponding numeric value The mapping rules strictly follow the ISO 6346 standard:

[0190] Numeric characters 0-9: directly mapped to their own numerical values, i.e. (for example but 5);

[0191] Alphanumeric characters AZ: follow the letter-number mapping rules specified in ISO 6346 (skip 11 to avoid ambiguity with the check code modulus value 11).

[0192] Weighted summation operation:

[0193] Based on the mapped digital value By weight ( (i.e., weights are 1, 2, 4, 8, 16, 32, 64, 128, 256, 512 respectively) A weighted summation is performed, calculated using the following formula:

[0194] ;

[0195] in:

[0196] This represents the weighted sum, and is a positive integer.

[0197] Indicates the first The weight value corresponding to each bit character;

[0198] Check code calculation:

[0199] Weighted sum Perform modulo-11 operations to obtain the theoretical check code. The calculation formula is:

[0200] ;

[0201] in:

[0202] This represents the theoretical check digit, with a value range of 0-10; if Then, according to ISO 6346 standard, Corrected to 0 (because the check code segment only contains numbers from 0 to 9).

[0203] S23.3. Compare the calculated check code with the check code segment of the candidate complete box number. If they match, the candidate box number is determined to be valid. The theoretical check code calculated in S23.2 is then used. With candidate complete box number The original check code segment is accurately compared to verify the encoding legality of the candidate box number. Only the candidate box numbers with matching check codes are retained, ensuring that all output legal candidate box numbers comply with the encoding verification rules of ISO6346. This provides a real and reliable candidate dataset for the subsequent box number identity verification module 25. The specific implementation is as follows:

[0204] Specifically, the container number validity verification module 23 executes the following comparison logic:

[0205] Extracting the original checksum: from the candidate complete box number Extract the 11th character and record it as the original checksum. Convert it to a numeric type (because the check digit range in the ISO 6346 standard is 0-9). );

[0206] Checksum comparison: The theoretical checksum... With the original check code Compare:

[0207] like and Then determine the complete box number of the candidate. This is a valid candidate container number;

[0208] like and Then determine the complete box number of the candidate. This is a valid candidate container number;

[0209] Except for the two situations mentioned above, all other candidate box numbers are deemed invalid and are removed.

[0210] Result processing:

[0211] Valid candidate container number Associate its overall confidence level The source information of each character segment is transmitted to the box number identity verification module 25 through the local high-speed PCIe interface;

[0212] Invalid candidate container numbers are logged in detail, including the complete container number and its theoretical check code. Original check code The reasons for the failure are compared and used for subsequent algorithm optimization and problem tracing.

[0213] In this embodiment, the identity binding and association module 24 synchronously collects corresponding container number and vehicle number information at key nodes of container operations and key nodes of trailer operations, establishes a dynamic binding relationship between operation cycle, trailer identification, and container number, and transmits it to the container number identity verification module 25. Through precise operation node timing alignment, data association, and logical verification, a unique binding between the container and the trailer is achieved within a single operation cycle, providing the container number identity verification module 25 with the basis for identity association in the operation scenario, avoiding identity confusion in mixed operation scenarios with multiple containers and multiple vehicles, and improving the accuracy of cargo handling operations and the traceability of the entire chain. The specific implementation is as follows:

[0214] Specifically, the identity binding and association module 24 receives the operation node trigger signal issued by the full-process cargo handling operation control unit 3, and identifies two types of core operation nodes:

[0215] Key milestones in container operations: Spreader placement, spreader unlocking, and container unloading.

[0216] Key milestones in towing operations: tow truck arrival and departure.

[0217] Meanwhile, the identity binding and association module 24 assigns a unified millisecond-level timestamp to all nodes and executes time sequence alignment logic: it determines that the nodes in the same operation cycle belong to the same operation cycle when the time difference between the container operation node and the trailer operation node is ≤5s, thus providing a time sequence basis for subsequent data collection and binding; the operation cycle ID is generated by timestamp + spreader number and establishes a one-to-one mapping relationship with the global operation cycle ID synchronized from the terminal operating system by the operation plan synchronization module 31 to ensure the uniqueness of each operation cycle.

[0218] Furthermore, under the triggering of the aligned job node, the identity binding and association module 24 synchronously collects two types of core data:

[0219] Container number information: Obtain all valid candidate complete container numbers within the current job loop from the container number character parsing module 22. ( ), including container number characters and overall confidence level Source information for each character segment;

[0220] Vehicle number information: Obtain trailer identification character data for the current work cycle from the trailer identification preprocessing submodule. ( The module identifies the trailer identification serial number, which includes the trailer identification character content, node type label (box in / unlocked), and collection timestamp. At the same time, the module initially associates the box number information and vehicle number information under the same job cycle ID to generate an intermediate dataset in the format of job cycle ID → [box number candidate set] → [vehicle number identification set], providing structured input for subsequent dynamic binding.

[0221] Meanwhile, the identity binding and association module 24 performs the following dynamic binding logic on the intermediate dataset:

[0222] Single data binding: If only one set of valid container numbers exists within the same work cycle. With 1 set of valid vehicle number identifiers Then directly establish the binding relationship: job cycle ;

[0223] Multiple data filtering: If multiple valid box numbers or multiple valid car numbers exist within the same work cycle, they will be filtered according to the following priority:

[0224] Priority 1: Select the container / car number with the smallest difference from the work node timestamp;

[0225] Priority 2: If the timestamp differences are the same, select the box number with the highest confidence level (i.e., (Largest) or the vehicle number with the most complete recognition result;

[0226] Uniqueness check: Verify the uniqueness of the binding relationship.

[0227] Only one binding relationship corresponds to the same job cycle ID;

[0228] Same trailer markings Only one box number is bound within the same work cycle. ;

[0229] Same box number Only one trailer tag is bound within the same work cycle. If the verification fails, the binding relationship is marked as abnormal, the log is recorded (including job cycle ID, abnormal data, and failure reason) and removed, and only the binding relationships that pass the verification are retained.

[0230] In addition, the identity binding association module 24 encapsulates the verified dynamic binding relationships into a structured dataset, with each binding data entry containing the following fields:

[0231] Job Cycle ID: Uniquely identifies a single job cycle;

[0232] Trailer sign : The trailer identifier character currently bound to the job cycle;

[0233] Target container number : The valid container number currently bound to the current job cycle;

[0234] Overall confidence level of container number Reliability score of the target container number;

[0235] Job node timestamp: Binds to the collection time of the corresponding job node;

[0236] Spreading Gear Number: The spreading gear number of the gantry crane used for this operation;

[0237] Node type label: Container operation node type (loading / unloading / dropping);

[0238] Finally, the identity binding and association module 24 transmits the dataset to the container number identity verification module 25 through the local high-speed PCIe interface, and at the same time, it stores and backs up the data locally on the edge computing industrial control computer. The backup data is retained for 30 days to meet the full-chain traceability requirements of port cargo handling operations.

[0239] In this embodiment, the container number identity verification module 25 receives the legitimate candidate container numbers transmitted by the container number validity verification module 23 and the dynamic binding relationship transmitted by the identity binding association module 24. It then sorts and confirms the initially screened legitimate candidate container numbers in conjunction with the terminal operation plan data and vehicle number binding results. Finally, it performs cross-verification of the container number identification results with the bound vehicle number information and the terminal operation plan data. The process by which the container number identity verification module 25 sorts and confirms the legitimate candidate container numbers in conjunction with the terminal operation plan data and vehicle number binding results and completes the cross-verification includes the following steps:

[0240] S25.1. Set three weighted dimensions: overall confidence level, planned character matching degree, and vehicle number association matching degree, and assign weights to each weighted dimension. Quantify the legitimate candidate container numbers through a multi-dimensional weighted scoring system. The overall reliability is assessed by incorporating the confidence level of the container number itself, its matching degree with the work plan, and its association matching degree with the bound vehicle number into the evaluation system. This provides an objective and quantifiable scoring basis for subsequent sorting and confirmation, avoiding the one-sidedness of a single-dimensional evaluation. The specific implementation is as follows:

[0241] Specifically, the container number identification module 25 predefines three weighted dimensions and their corresponding weights. All weights satisfy the normalization condition to ensure that the comprehensive score can be directly used for ranking.

[0242] Overall confidence level dimension: corresponding to the valid candidate container number Overall confidence level This reflects the reliability of the container number identification result, and the weight is denoted as [weight value]. The possible value is 0.4;

[0243] Planned character matching degree dimension: corresponding valid candidate box number The degree of character matching between the container number and the target container number in the terminal operation plan reflects the fit between the container number and the operation plan, and the weight is denoted as [weight not specified]. The possible value is 0.3;

[0244] Vehicle number association matching degree dimension: corresponding legitimate candidate container numbers The degree of association and matching between the container number and the bound trailer number in the work plan reflects the operational relevance between the container number and the bound trailer number, and the weight is denoted as [weight not specified]. The possible value is 0.3;

[0245] At the same time, ensure that the weights are normalized: The weight values ​​can be fine-tuned according to the port operation scenario, but the normalization constraint is always maintained.

[0246] S25.2 Overall confidence level based on candidate complete container numbers Calculate the complete box number of the legal candidate based on the matching degree of each dimension. Overall score Valid candidate container numbers are calculated using a weighted summation formula. The comprehensive score integrates the evaluation results from the three dimensions into a single quantifiable score, providing a unified standard for subsequent ranking. The specific implementation is as follows:

[0247] Specifically, the container number identity verification module 25 first calculates the matching degree values ​​for each dimension:

[0248] Overall confidence level: Directly reuses the output of box number character parsing module 22. The value range is [0,1];

[0249] Planned character matching degree:

[0250] Record the target container number corresponding to the current operation cycle in the terminal operation plan as follows: ,Will and The planned character matching degree is the proportion of matched characters to the total number of characters compared character by character. The calculation formula is:

[0251] ;

[0252] in express and The number of consecutive characters in the middle section, 11 is the total number of digits in the box number;

[0253] Vehicle number association matching degree:

[0254] The vehicle number transmitted by the identity binding association module 24 is recorded as follows: Inquiry about the dock operation plan The corresponding planned container number is ,Will and The vehicle number association matching degree is the ratio of the number of matched characters to the total number of characters compared character by character. The calculation formula is:

[0255] ;

[0256] in express and The number of characters in the same length;

[0257] At the same time, a comprehensive score is calculated based on the values ​​of the above three dimensions. The calculation formula is:

[0258] ;

[0259] in:

[0260] Indicates the valid candidate container number The overall score ranges from [0,1]. The closer the value is to 1, the higher the overall reliability of the candidate box number.

[0261] , , These are the matching scores for the corresponding dimensions, with values ​​ranging from [0,1].

[0262] S25.3, Based on the overall score The valid candidate container numbers are sorted in descending order, and the candidate container number with the highest score is selected as the target container number. The final selection is based on the overall score. The corresponding judgment operation is performed, and the most reliable target box number is selected by sorting the comprehensive scores. At the same time, a differentiated judgment operation is performed based on the score range to balance the recognition accuracy and operation efficiency. The specific implementation is as follows:

[0263] Specifically, the container number identity verification module 25 first verifies all valid candidate container numbers. according to Sort the candidate box numbers in descending order from highest to lowest to generate a sorted list of candidate box numbers;

[0264] Further, select from the list The highest candidate container number is used as the target container number. If multiple candidate box numbers have the same score and all are the highest value, then one of them will be selected. The highest candidate container number is used as the target container number;

[0265] Finally, based on the target container number Overall score Perform the corresponding judgment operation:

[0266] like If a container number is determined to be a highly reliable target container number, it is directly confirmed as a valid container number for the current work cycle.

[0267] like If the container number is determined to be of medium reliability, a manual review process will be triggered, and the tally personnel will verify the container number information for confirmation.

[0268] like If the target container number is determined to be of low reliability, it is marked as an identification anomaly, triggering the container number re-identification process, and returning to the container number character parsing module 22 to reprocess the image data of the current job cycle.

[0269] S25.4. Perform a consistency comparison between the target container number and the terminal operation plan data, as well as the planned truck information corresponding to the bound vehicle number, to complete the cross-verification of the container number identity. A two-way comparison of the planned trailer information verifies the consistency between the container number and the operational scenario, avoiding errors from single-dimensional verification and ensuring the accuracy and traceability of cargo handling data. The specific implementation is as follows:

[0270] Specifically, the container number identification module 25 first obtains the terminal operation plan data for the current operation cycle from the full-process cargo handling operation control unit 3, including the planned container number. Planned trailer markings Simultaneously, the bound vehicle number is obtained from the identity binding association module 24. The corresponding planned trailer information; simultaneously, the module performs a two-layer consistency comparison:

[0271] Container number - work plan comparison: Target container number With the planned container number The system compares each character. If the characters are completely identical, the box number is determined to be consistent with the work plan. If there are character differences, it is marked as a plan inconsistency, triggering an exception handling process.

[0272] Container number - vehicle number comparison: Query the vehicle numbers bound to the terminal operation plan. Corresponding planned container number Target container number and The system compares each character. If the characters are completely identical, the box number and the bound vehicle number are considered to be consistent. If there are character differences, the system marks the vehicle number as inconsistent and triggers exception handling.

[0273] Furthermore, cross-validation is performed based on the comparison results of the two layers:

[0274] If both comparisons match: the cross-verification of the container number is deemed successful, and the target container number is... Bind vehicle number The operation cycle ID is encapsulated into the final sorting data and transmitted to the edge computing and cloud collaborative control unit 4;

[0275] If any layer of comparison is inconsistent: cross-validation is deemed to have failed, a detailed exception log is recorded (including target container number, planned container number, bound vehicle number, and difference location), triggering a manual intervention process, where tally personnel verify and correct the data on-site;

[0276] Finally, the cross-validation result (pass / fail), target container number, and bound vehicle number will be included. The operation cycle ID is stored in the local database and simultaneously synchronized to the cloud-based cargo handling system to meet the full-chain traceability requirements of port cargo handling operations.

[0277] The full-process cargo handling operation control unit 3 automatically synchronizes the shift operation plan data from the terminal operating system and provides it to the container identification and container number parsing unit 2 as terminal operation plan data. Based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit 1, it triggers the container number visual acquisition component of the container identification and container number parsing unit 2 to perform multi-view image data acquisition of the container. At the container landing and gantry crane spreader unlocking nodes, it triggers the vehicle identification acquisition component of the container identification and container number parsing unit 2 to collect trailer identification data, executes the cross-verification logic of the container identification and container number parsing unit 2, and feeds back the manual verification results to the system as training data for the self-learning model. The full-process cargo handling operation control unit 3 includes an operation plan synchronization module 31, a container number acquisition triggering module 32, a vehicle number acquisition triggering module 33, a verification logic execution module 34, and a verification data feedback module 35, wherein:

[0278] In this embodiment, the operation plan synchronization module 31 automatically synchronizes the shift operation plan data from the terminal operating system and provides the shift operation plan data to the container identification and container number parsing unit 2 to support subsequent business logic such as container number matching and cross-validation. The specific implementation is as follows:

[0279] Specifically, the operation plan synchronization module 31 establishes data interaction with the terminal operating system through a standardized API interface, and automatically synchronizes the operation plan data for the current shift according to the shift number. The core fields include operation cycle ID, planned container number, planned trailer identifier, spreader number, and operation node sequence. The synchronized data is cached locally on the edge computing industrial control computer and stored according to the operation cycle ID index. It responds to the data query requests of the container identification and container number parsing unit 2 in real time to ensure the timeliness and availability of the plan data.

[0280] In this embodiment, the container number acquisition triggering module 32 triggers the container number visual acquisition component of the container identification and container number parsing unit 2 to perform multi-view image data acquisition of the container based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit 1. This avoids invalid acquisition and adapts to the dynamic operation scenario of the gantry crane. The specific implementation is as follows:

[0281] Specifically, the container number acquisition trigger module 32 subscribes in real time to the three-dimensional position and rotation angle data of the spreader output by the multi-source fusion positioning perception and processing unit 1. When the triggering condition of "spreader lifting height ≥ 2m and rotation angle ≥ 5°" is met, it sends an acquisition command to the container number visual acquisition component of the container identification and container number parsing unit 2 to start the multi-view image acquisition of the container; it simultaneously records the position and angle of the spreader at the time of triggering and associates it with the image data acquired subsequently, providing a basis for scene traceability for container number recognition.

[0282] In this embodiment, the vehicle number acquisition triggering module 33 triggers the vehicle number identification acquisition component of the container identification and container number parsing unit 2 to collect trailer identification data at the container loading and gantry crane spreader unlocking nodes, ensuring that the collected data is strongly correlated with the operation scenario and reducing redundant data. The specific implementation is as follows:

[0283] Specifically, the vehicle number acquisition trigger module 33 receives node signals from the limit switch of the gantry crane and the spreader unlocking sensor. When it detects the signal of "container in place" or "spreader unlocking action completed", it immediately sends an acquisition command to the vehicle number identification acquisition component of the container identification and container number parsing unit 2 to acquire the trailer identification data corresponding to the current operation cycle. It adds a node type label (container in / unlocked) and operation cycle ID to each acquisition to ensure that the vehicle number data is accurately bound to the operation scenario.

[0284] In this embodiment, the verification logic execution module 34 executes the cross-verification logic of the container identification and container number parsing unit 2, linking the operation plan data and the bound vehicle number information to verify the accuracy of the container number identity. The specific implementation is as follows:

[0285] Specifically, the verification logic execution module 34 calls the verification interface of the container identification and container number parsing unit 2, and inputs the current operation cycle ID, the synchronized shift operation plan data, and the dynamic binding relationship output by the identity binding association module 24 to trigger the container number identity cross-verification process; at the same time, it receives the verification result (pass / fail) and triggers the subsequent process: if the verification passes, standardized cargo handling data is generated; if the verification fails, a manual review process is initiated to ensure the accuracy of the cargo handling data.

[0286] In this embodiment, the verification data feedback module 35 feeds back the manual verification results to the edge computing and cloud collaborative control unit 4 as training data for the self-learning model, continuously optimizing the accuracy of box number recognition and verification. The specific implementation is as follows:

[0287] Specifically, the verification data feedback module 35 receives the corrected data after manual verification (including the original identification box number, the corrected box number, the bound vehicle number, the operation cycle ID, etc.), encapsulates it into standardized training samples, and transmits it to the edge computing and cloud collaborative control unit 4 via industrial Ethernet; and simultaneously records the verification time and verification personnel information to ensure the traceability of training data and provide high-quality samples for the system's self-learning iteration.

[0288] The edge computing and cloud-based collaborative control unit 4 deploys an edge computing industrial control computer at the gantry crane end. It runs the data fusion algorithm of the multi-source fusion positioning perception and processing unit 1, the character recognition processing, container number parsing, and visual recognition algorithm of the container identification and container number parsing unit 2, and deploys a system server in the cloud to perform data interaction with the terminal operating system. It receives cargo handling business scheduling instructions from the full-process cargo handling operation control unit 3, pushes abnormal alarm information from the container identification and container number parsing unit 2, and stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit 1, the container identification and container number parsing unit 2, and the full-process cargo handling operation control unit 3. The edge computing and cloud-based collaborative control unit 4 includes an edge computing deployment module 41, a cloud interaction module 42, a scheduling instruction receiving module 43, an abnormal alarm push module 44, and a full-link data storage module 45, wherein:

[0289] The edge computing deployment module 41 deploys an edge computing industrial control computer at the end of the gantry crane, running the data fusion algorithm of the multi-source fusion positioning perception and processing unit 1, and the character recognition processing, container number parsing and visual recognition algorithm of the container identification and container number parsing unit 2, reducing data transmission latency and ensuring the real-time performance of operation control. The specific implementation is as follows:

[0290] Specifically, the edge computing deployment module 41 is equipped with an edge computing industrial control computer at the gantry crane end. It is directly connected to the multi-source fusion positioning perception and processing unit 1, the container identification and container number parsing unit 2, and the full-process cargo handling operation control unit 3 via industrial Ethernet. It runs local data fusion algorithm, character recognition processing algorithm, container number parsing algorithm and visual recognition algorithm. All real-time calculations are completed on the edge side, and the calculation results are directly sent to the corresponding execution unit without relying on cloud network transmission.

[0291] The cloud-based interaction module 42 deploys a system server in the cloud to perform data interaction with the terminal's operating system, achieving global communication of work plans, equipment status, and cargo handling results. The specific implementation is as follows:

[0292] Specifically, the cloud interaction module 42 deploys a system server in the cloud and establishes an encrypted data channel with the terminal operating system through the TCP / IP protocol. It synchronizes the shift operation plan, cargo handling results, and equipment operation status data in both directions, providing cloud-level data support for the full-process cargo handling operation control unit 3. At the same time, it synchronizes the edge processing results to the terminal operating system to complete the business closed loop.

[0293] The scheduling instruction receiving module 43 receives the cargo handling business scheduling instructions from the full-process cargo handling operation control unit 3, realizing the unified flow of business instructions between the edge side and the cloud side. The specific implementation is as follows:

[0294] Specifically, the scheduling instruction receiving module 43 monitors in real time the scheduling instructions issued by the full-process sorting operation control unit 3, including instructions such as collection start, verification execution, manual review, and data retransmission. After parsing the instructions, it forwards them to the edge computing deployment module 41 or the corresponding functional unit to ensure the orderly execution of scheduling instructions and status feedback.

[0295] The anomaly alarm push module 44 pushes anomaly alarm information from the container identification and container number parsing unit 2 for quick handling. The specific implementation is as follows:

[0296] Specifically, the abnormal alarm push module 44 collects the abnormal alarm information output by the container identification and container number parsing unit 2 in real time, classifies and encapsulates it according to the abnormality level, and pushes it to the full-process cargo handling operation control unit 3 and the remote interaction and operation data traceability unit 5 through the industrial Ethernet, and records the abnormality occurrence time, operation cycle ID and abnormality type in a traceable abnormality log.

[0297] The full-link data storage module 45 stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit 1, the container identification and container number parsing unit 2, and the full-process cargo handling operation control unit 3.

[0298] The remote interaction and operation data traceability unit 5 displays the cargo handling operation status of the full-process cargo handling operation control unit 3 and the abnormal alarm information of the container identification and container number parsing unit 2 through the remote cargo handling center monitoring screen and mobile application. It receives remote manual review instructions and feeds them back to the full-process cargo handling operation control unit 3. It encrypts and stores the full-link operation data stored in the edge computing and cloud collaborative control unit 4, and supports the retrieval of operation chain data records for any container.

[0299] Specifically, the remote interaction and operation data traceability unit 5 displays the cargo handling operation status of the full-process cargo handling operation control unit 3 and the abnormal alarm information of the container identification and container number parsing unit 2 through the remote cargo handling center monitoring screen and mobile application. It receives remote manual review instructions and feeds them back to the full-process cargo handling operation control unit 3. It encrypts and stores the full-link operation data stored in the edge computing and cloud collaborative control unit 4, and supports the retrieval of the corresponding complete operation chain data record based on any container information, so as to realize remote operation monitoring, manual intervention and historical data traceability.

[0300] This embodiment also provides an unmanned intelligent cargo handling method based on a gantry crane. Based on the aforementioned unmanned intelligent cargo handling system based on a gantry crane, the method includes the following steps:

[0301] S1. The absolute azimuth angle data of the lifting device is collected by the dual-axis digital compass sensor installed on the lifting device of the gantry crane. The relative displacement data of each mechanism is collected by the encoder of the gantry crane's own mechanism. After the absolute direction reference and relative displacement data are fused and corrected by the data fusion algorithm, the real-time position and angle data of the lifting device of the gantry crane are output.

[0302] S2. Automatically synchronize the shift operation plan data from the terminal operating system, and use the shift operation plan data as the terminal operation plan data for subsequent container number identification and verification.

[0303] S3. Based on the real-time position and angle of the gantry crane spreader output in step S1, trigger the acquisition of multi-view image data of the container. Perform character recognition processing on the acquired multi-view image data of the container to generate several candidate complete container numbers. Introduce the standard coding rules for container numbers to perform a preliminary screening of the legality of the candidate complete container numbers and obtain legal candidate container numbers.

[0304] S4. At the container loading and gantry crane spreader unlocking nodes, trigger the collection of trailer identification data, and simultaneously collect the corresponding container number information and vehicle number information at the key nodes of container operation and trailer operation, and establish a dynamic binding relationship between operation cycle, trailer identification and container number.

[0305] S5. The legal candidate container numbers obtained in step S3 are combined with the terminal operation plan data and vehicle number binding results to perform a weighted comprehensive score sorting to confirm the target container number. The target container number is then compared with the terminal operation plan data and the planned trailer information corresponding to the bound vehicle number to complete the cross-verification of the container number identity.

[0306] S6. Runs data fusion algorithms, character recognition processing, container number parsing and visual recognition algorithms at the gantry crane end, performs data interaction with the terminal operating system in the cloud, receives tallying business scheduling instructions, pushes abnormal alarm information in the container number recognition and verification process, and stores the full-link operation data generated in the entire tallying operation process.

[0307] S7. Displays the status of cargo handling operations and abnormal alarm information for container number identification and verification through a remote cargo handling center monitoring screen and mobile application. Receives remote manual review instructions and feeds them back into the cargo handling process. Encrypts and stores data across the entire operation chain and supports retrieving operation chain data records for any container.

[0308] S8. Feedback the results of remote manual review as training data for the self-learning model of container number character recognition, container number parsing, and identity cross-verification.

[0309] Those skilled in the art will understand that the process of implementing all or part of the steps of the above embodiments can be carried out by hardware or by a program instructing the relevant hardware.

[0310] 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 claimed invention.

Claims

1. An unmanned intelligent cargo handling system based on a gantry crane, characterized in that, include: Multi-source fusion positioning perception and processing unit (1) uses a dual-axis digital compass sensor installed on the gantry crane spreader to collect the absolute azimuth angle data of the gantry crane spreader, reuses the encoder of the gantry crane's own mechanism to collect the relative displacement data of each mechanism, and fuses and corrects the absolute direction reference and relative displacement data through a data fusion algorithm to output the real-time position and angle data of the gantry crane spreader. The container identification and container number parsing unit (2) uses a container number visual acquisition component and a vehicle number identification acquisition component installed on the gantry crane spreader to collect container multi-view image data and trailer identification data respectively. The collected container multi-view image data is processed by character recognition to generate several candidate complete container numbers. The container number standard coding rules are introduced to screen the candidate complete container numbers for legality. At the same time, the corresponding container number information and vehicle number information are collected at the key nodes of container operation and key nodes of trailer operation. The dynamic binding relationship between operation cycle, trailer identification and container number is established. The candidate complete container numbers after initial screening are sorted and confirmed in combination with the terminal operation plan data and vehicle number binding results. The container number recognition results are cross-validated in combination with the bound vehicle number information and the terminal operation plan data. The full-process cargo handling operation control unit (3) automatically synchronizes the on-duty operation plan data from the terminal operating system and provides it to the container identification and container number parsing unit (2) as terminal operation plan data. Based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit (1), the container number visual acquisition component of the container identification and container number parsing unit (2) is triggered to perform container multi-view image data acquisition. At the container loading and gantry crane spreader unlocking nodes, the vehicle number identification acquisition component of the container identification and container number parsing unit (2) is triggered to collect trailer identification data. The cross-verification logic of the container identification and container number parsing unit (2) is executed, and the manual verification result is fed back to the system as training data for the self-learning model. Edge computing and cloud collaborative control unit (4), the edge computing and cloud collaborative control unit (4) deploys an edge computing industrial control computer at the gantry crane end, runs the data fusion algorithm of the multi-source fusion positioning perception and processing unit (1), the character recognition processing, container number parsing and visual recognition algorithm of the container identification and container number parsing unit (2), deploys a system server in the cloud to perform data interaction with the terminal operating system, undertakes the cargo handling business scheduling instructions of the full-process cargo handling operation control unit (3), pushes abnormal alarm information of the container identification and container number parsing unit (2), and stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit (1), the container identification and container number parsing unit (2), and the full-process cargo handling operation control unit (3); The remote interaction and operation data traceability unit (5) displays the cargo handling operation status of the full-process cargo handling operation control unit (3) and the abnormal alarm information of the container identification and container number parsing unit (2) through the remote cargo handling center monitoring screen and mobile application. It receives remote manual review instructions and feeds them back to the full-process cargo handling operation control unit (3). It encrypts and stores the full-link operation data stored in the edge computing and cloud collaborative control unit (4) and supports the retrieval of operation chain data records of any container.

2. The unmanned intelligent cargo handling system based on a gantry crane according to claim 1, characterized in that, The multi-source fusion positioning sensing and processing unit (1) includes an azimuth data acquisition module (11), a relative displacement data acquisition module (12), a data fusion correction module (13), and a positioning data output module (14), wherein: The azimuth data acquisition module (11) uses a dual-axis digital compass sensor installed on the gantry crane spreader to acquire the absolute azimuth data of the gantry crane spreader and transmit it to the data fusion correction module (13). The relative displacement data acquisition module (12) reuses the encoders of the hoisting mechanism, the luffing mechanism, and the slewing mechanism of the gantry crane to collect the relative displacement data of the hoisting height of the gantry crane spreader, the relative displacement data of the boom pitch angle of the gantry crane, and the relative displacement data of the slewing angle of the gantry crane, and transmits them to the data fusion and correction module (13). The data fusion correction module (13) receives the absolute azimuth data transmitted by the azimuth data acquisition module (11), the lifting height relative displacement data, the boom pitch angle relative displacement data, and the slewing angle relative displacement data transmitted by the relative displacement data acquisition module (12). It then uses a data fusion algorithm to fuse and correct the absolute direction reference corresponding to the absolute azimuth data with the three types of relative displacement data, and transmits the corrected positioning data to the positioning data output module (14). The positioning data output module (14) receives the corrected positioning data transmitted by the data fusion correction module (13) and outputs the real-time position and angle data of the gantry crane lifting device.

3. The unmanned intelligent cargo handling system based on a gantry crane according to claim 1, characterized in that, The container identification and container number parsing unit (2) includes a multi-source visual acquisition module (21), a container number character parsing module (22), a container number validity verification module (23), an identity binding and association module (24), and a container number identity verification module (25), wherein: The multi-source visual acquisition module (21) uses the container number visual acquisition component and the vehicle number identification acquisition component installed on the gantry crane spreader to acquire container multi-view image data and trailer identification data respectively, and transmits the container multi-view image data to the container number character parsing module (22) and the trailer identification data to the identity binding association module (24). The container number character parsing module (22) receives container multi-view image data transmitted by the multi-source visual acquisition module (21), performs character recognition processing on the container multi-view image data to generate several candidate complete container numbers and transmits them to the container number legality verification module (23). The container number legality verification module (23) receives the candidate complete container number transmitted by the container number character parsing module (22), introduces the container number standard encoding rules to perform a preliminary screening of the candidate complete container number, and transmits the preliminary screened legal candidate container number to the container number identity verification module (25). The identity binding association module (24) synchronously collects the corresponding container number information and vehicle number information at the key nodes of container operation and trailer operation, establishes a dynamic binding relationship between operation cycle, trailer identification and container number, and transmits it to the container number identity verification module (25). The container number identity verification module (25) receives the legal candidate container number transmitted by the container number legality verification module (23) and the dynamic binding relationship transmitted by the identity binding association module (24). It sorts and confirms the legal candidate container numbers after initial screening in combination with the terminal operation plan data and vehicle number binding results. It also performs cross-verification of the container number identification results in combination with the bound vehicle number information and the terminal operation plan data.

4. The unmanned intelligent cargo handling system based on a gantry crane according to claim 3, characterized in that, The multi-source visual acquisition module (21) includes a container number image acquisition submodule, a container number image preprocessing submodule, a trailer identification acquisition submodule, and a trailer identification preprocessing submodule, wherein: The container number image acquisition submodule is used to continuously acquire container number image data from multiple perspectives of the container at a preset acquisition frequency during the operation of lifting and rotating containers by the gantry crane spreader. The container number image preprocessing submodule is used to sequentially perform Gaussian filtering noise reduction and contrast-limited adaptive histogram equalization contrast enhancement operations on the single frame container number image data acquired by the container number image acquisition submodule, and transmit the processed container number image data to the container number character parsing module (22). The trailer identification acquisition submodule is used to acquire trailer identification image data at the container loading node and the gantry crane spreader unlocking node. The trailer identification preprocessing submodule is used to sequentially perform Canny edge detection contour extraction and projection character segmentation operations on the trailer identification image data collected by the trailer identification acquisition submodule, and transmit the processed trailer identification data to the identity binding association module (24).

5. The unmanned intelligent cargo handling system based on a gantry crane according to claim 4, characterized in that, The process by which the container number character parsing module (22) performs character recognition processing on multi-view image data of the container to generate several candidate complete container numbers includes the following steps: S22.

1. According to the ISO 6346 standard, the container number identification target is decomposed into the owner code segment, equipment category identifier segment, serial number segment, and check code segment. The first segment in a single-view, single-frame image is recorded. The recognition result of each character segment is ; S22.

2. Locate the local character segments in each frame of the image and calculate the segment-level confidence score for each character segment. Filter out valid character segments whose confidence level meets the preset threshold; S22.

3. Perform time-sequential splicing of valid character segments from different frames to generate several candidate complete box numbers. And calculate the overall confidence of the candidate complete box number based on the fragment-level confidence. .

6. The unmanned intelligent cargo handling system based on a gantry crane according to claim 5, characterized in that, The process of the container number validity verification module (23) performing preliminary screening of the candidate complete container numbers includes the following steps: S23.1, For candidate complete container numbers Perform format compliance verification and eliminate candidate box numbers that do not conform to the ISO6346 standard format; S23.2 Perform a checksum operation on the candidate complete container numbers that meet the format requirements, and convert the first 10 characters of the container number into a numeric value. Based on numerical values Calculate the check code ; S23.

3. Compare the calculated check code with the check code segment of the candidate complete box number. If the comparison is consistent, it is determined to be a valid candidate box number.

7. The unmanned intelligent cargo handling system based on a gantry crane according to claim 6, characterized in that, The container number identity verification module (25) sorts and confirms legitimate candidate container numbers in conjunction with terminal operation plan data and vehicle number binding results, and completes the cross-verification process, which includes the following steps: S25.1 Set three weighted dimensions: overall confidence level, planned character matching degree, and vehicle number association matching degree, and assign weights to each weighted dimension; S25.2 Overall confidence level based on candidate complete container numbers Calculate the complete box number of the legal candidate based on the matching degree of each dimension. Overall score ; S25.3, Based on the overall score The valid candidate container numbers are sorted in descending order, and the candidate container number with the highest score is selected as the target container number. The final selection is based on the overall score. Execute the corresponding judgment operation; S25.

4. Perform a consistency comparison between the target container number and the terminal operation plan data and the planned trailer information corresponding to the bound vehicle number to complete the cross-verification of the container number identity.

8. The unmanned intelligent cargo handling system based on a gantry crane according to claim 1, characterized in that, The full-process cargo handling operation control unit (3) includes an operation plan synchronization module (31), a container number acquisition trigger module (32), a vehicle number acquisition trigger module (33), a verification logic execution module (34), and a verification data feedback module (35), wherein: The operation plan synchronization module (31) automatically synchronizes the shift operation plan data from the terminal operating system and provides the shift operation plan data to the container identification and container number parsing unit (2). The container number acquisition trigger module (32) triggers the container number visual acquisition component of the container identification and container number parsing unit (2) to perform container multi-view image data acquisition based on the real-time position and angle of the gantry crane spreader output by the multi-source fusion positioning perception and processing unit (1). The vehicle number acquisition triggering module (33) triggers the vehicle number identification acquisition component of the container identification and container number parsing unit (2) to collect trailer identification data at the container loading and gantry crane spreader unlocking nodes; The verification logic execution module (34) executes the cross-verification logic of the container identification and container number parsing unit (2); The verification data feedback module (35) feeds back the manual verification results to the edge computing and cloud collaborative control unit (4) as training data for the self-learning model.

9. The unmanned intelligent cargo handling system based on a gantry crane according to claim 1, characterized in that, The edge computing and cloud collaborative control unit (4) includes an edge computing deployment module (41), a cloud interaction module (42), a scheduling instruction receiving module (43), an anomaly alarm push module (44), and a full-link data storage module (45), wherein: The edge computing deployment module (41) deploys an edge computing industrial control computer at the end of the gantry crane, and runs the data fusion algorithm of the multi-source fusion positioning perception and processing unit (1), the character recognition processing of the container identity recognition and container number parsing unit (2), the container number parsing and visual recognition algorithm; The cloud interaction module (42) deploys a system server in the cloud and performs data interaction with the dock operating system; The scheduling instruction receiving module (43) receives the cargo handling business scheduling instructions from the full-process cargo handling operation control unit (3); The abnormal alarm push module (44) pushes abnormal alarm information from the container identification and container number parsing unit (2); The full-link data storage module (45) stores the full-link operation data generated by the multi-source fusion positioning perception and processing unit (1), the container identification and container number parsing unit (2), and the full-process cargo handling operation control unit (3).

10. An unmanned intelligent cargo handling method based on a gantry crane, based on the unmanned intelligent cargo handling system based on a gantry crane as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. The absolute azimuth angle data of the lifting device is collected by the dual-axis digital compass sensor installed on the lifting device of the gantry crane. The relative displacement data of each mechanism is collected by the encoder of the gantry crane's own mechanism. After the absolute direction reference and relative displacement data are fused and corrected by the data fusion algorithm, the real-time position and angle data of the lifting device of the gantry crane are output. S2. Automatically synchronize the shift operation plan data from the terminal operating system, and use the shift operation plan data as the terminal operation plan data for subsequent container number identification and verification. S3. Based on the real-time position and angle of the gantry crane spreader output in step S1, trigger the acquisition of multi-view image data of the container. Perform character recognition processing on the acquired multi-view image data of the container to generate several candidate complete container numbers. Introduce the standard coding rules for container numbers to perform a preliminary screening of the legality of the candidate complete container numbers and obtain legal candidate container numbers. S4. At the container loading and gantry crane spreader unlocking nodes, trigger the collection of trailer identification data, and simultaneously collect the corresponding container number information and vehicle number information at the key nodes of container operation and trailer operation, and establish a dynamic binding relationship between operation cycle, trailer identification and container number. S5. The legal candidate container numbers obtained in step S3 are combined with the terminal operation plan data and vehicle number binding results to perform a weighted comprehensive score sorting to confirm the target container number. The target container number is then compared with the terminal operation plan data and the planned trailer information corresponding to the bound vehicle number to complete the cross-verification of the container number identity. S6. Runs data fusion algorithms, character recognition processing, container number parsing and visual recognition algorithms at the gantry crane end, performs data interaction with the terminal operating system in the cloud, receives tallying business scheduling instructions, pushes abnormal alarm information in the container number recognition and verification process, and stores the full-link operation data generated in the entire tallying operation process. S7. Displays the status of cargo handling operations and abnormal alarm information for container number identification and verification through a remote cargo handling center monitoring screen and mobile application. Receives remote manual review instructions and feeds them back into the cargo handling process. Encrypts and stores data across the entire operation chain and supports retrieving operation chain data records for any container. S8. Feed back the results of remote manual review as training data for the self-learning model of container number character recognition, container number parsing and identity cross-verification.

Images