A smart container sorting system for the international logistics lubricant industry

By employing WebGL technology and intelligent bin-splitting algorithms on the browser side, the problem of mixed loading of multi-form goods in the lubricating oil industry has been solved, enabling real-time 3D rendering and fine-grained interaction, improving bin-splitting efficiency and space utilization, and reducing deployment costs.

CN123070828APending Publication Date: 2026-11-10SHANGHAI TOGETHER DELIVERY NETWORK TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202611425401.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-15
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies cannot effectively handle the mixed loading needs of various forms of goods such as drums, barrels, cartons, and pallets in the lubricating oil industry. Furthermore, the visualization of packing solutions relies on offline rendering or dedicated hardware equipment, and real-time 3D rendering and fine-grained interaction cannot be achieved on the browser side. The independent operation process results in the separation of the packing calculation and visualization modules.

Method used

Using WebGL technology, real-time 3D rendering of containers and cargo is achieved on the browser side. Combined with intelligent container sorting algorithm, through cargo attribute recognition module, loading constraint engine, automatic container sorting module, Web-side 3D rendering module, layered expansion submodule, layer-by-layer filtering submodule and floating interaction submodule, it realizes automated container sorting and real-time 3D visualization of multi-form cargo. It supports layered expansion, layer-by-layer filtering and floating interaction, and achieves seamless integration of container sorting calculation and 3D rendering.

Benefits of technology

It enables automated sorting of lubricating oil in various forms, improving sorting efficiency and space utilization, reducing deployment costs, and supporting real-time 3D visualization and fine-grained interaction on the browser side to ensure smooth and consistent operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123070828A_ABST
    Figure CN123070828A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent box sorting system for the international logistics lubricant industry, belonging to the field of international logistics packing planning technology. The system includes: a cargo attribute recognition module, which classifies cargo into three types—cartons, medium-sized drums, and large drums—based on shape type fields, and performs packing calculations in conjunction with cargo weight and packaging size, configuring geometric modeling parameters and rendering attributes; a loading constraint engine, with built-in heavy-bottom and light-top, large-bottom-small-top constraints, prioritizing the allocation of heavy drums and entire pallets to the bottom layer, and real-time verification of the total cargo weight and container load capacity limit; an automatic box sorting module, where the server automatically calculates the box sorting scheme based on cargo specifications, weight, batch, quantity, and container size constraints; a web-based 3D rendering module, which renders multi-form cargo models on the browser side based on a WebGL engine; and a layered expansion, layer-by-layer filtering, and ray projection floating interaction sub-module, enabling layered viewing and item-by-item identification of cargo within the box. This invention achieves automated box sorting and 3D visualization without requiring dedicated hardware.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of international logistics container planning technology, and in particular to an intelligent container sorting system for the international logistics lubricant industry. This invention comprehensively utilizes WebGL 3D graphics rendering engines (such as Three.js), perspective projection and track interaction control, and ray projection collision detection algorithms to achieve real-time 3D modeling and interactive display of the stacking shape of lubricant goods inside containers. Combined with an intelligent container sorting algorithm, it automatically calculates the packing plan based on parameters such as cargo specifications, weight, and shape, following the loading constraint principle of heavy at the bottom and light at the top, and large at the bottom and small at the top. The calculation results are presented in 3D using multi-morphological geometries (cartons, medium drums, large drums), layered unfolded views, layer-by-layer filtering, and floating information prompts. This belongs to the interdisciplinary application field of logistics information technology and computer graphics. Background Technology

[0002] The Three-Dimensional Container Loading Problem (3D-CLP) is a classic NP-hard problem in operations research. Its core is to load goods of different sizes, weights, and shapes into a given three-dimensional container (such as a shipping container) in a rational arrangement to maximize space utilization and satisfy loading constraints. It has significant engineering application value in international logistics. Currently, container loading solutions in the lubricant industry's international logistics mainly rely on manual experience. Loading personnel need to determine the placement of each barrel of lubricant and each carton based on the cargo details on the execution order, resulting in low efficiency, unstable space utilization, and a lack of intuitive solutions.

[0003] In terms of algorithm research, existing technologies have accumulated a certain amount of expertise. The paper "Multi-level Tree Search Algorithm for Solving the 3D Packing Problem" (Acta Automatica Sinica, DOI:10.16383 / j.aas.c180795) proposes a multi-level tree search algorithm for space allocation, recursively partitioning and searching the container space by constructing a multi-level tree structure. The paper "Research on Solving Container Loading Problem Based on Packing Tree Algorithm" (Information and Control) combines heuristic algorithms to establish a packing tree model, transforming the loading problem into the construction and traversal of the packing tree. The above algorithm research mainly targets the homogeneous or weakly heterogeneous loading scenarios of general standard cardboard box goods. However, for the special needs of the lubricating oil industry's mixed loading of various forms of goods such as drums, boxes, and pallet combinations, the adaptability of existing algorithms is insufficient, and they cannot directly handle the mixed stacking of medium and large drums with cardboard boxes and their center of gravity and load-bearing constraints.

[0004] In terms of 3D visualization and interaction, existing technologies also have significant shortcomings. Chinese patent CN112607234B discloses a 3D packing method for modular containers. This method involves deploying depth cameras at a dedicated workstation to acquire depth images of the sides and top of the cargo, obtaining cargo dimensions through image recognition, arranging the cargo using a 3D packing optimization algorithm, and finally generating a 3D spatial model of the storage unit adjustment and cargo stacking within the modular container using parametric families in Autodesk Revit software. The shortcomings of this solution are as follows: First, its cargo size acquisition relies on dedicated hardware equipment such as depth cameras, which need to be deployed at dedicated workstations, resulting in high deployment costs and strong environmental adaptability. Second, its 3D spatial model relies on desktop offline modeling software such as Revit for generation, which cannot be rendered and interacted with in real time in a regular web browser. On-site personnel can only view the offline-generated static model and cannot perform fine-grained verification operations such as layered viewing and item-by-item identification of the cargo inside the container. Third, its packing algorithm is designed for the dynamic adaptation scenario of storage units in modular containers and is not specifically designed for the mixed loading constraints of various cargo forms such as barrels, large barrels, cartons, and pallet combinations in the lubricating oil industry.

[0005] US Patent 11820611B2 discloses a load planning optimization method using automated 3D packing, which can generate a final packing plan and provide a 3D rendering display. However, this method is not adapted to the specific loading requirements of the lubricating oil industry. Its cargo model is mainly based on general regular geometry, lacking differentiated modeling and rendering for lubricating oil industry-specific loading materials such as oil drums, pallets, pads, partitions, and strapping. At the same time, its 3D display has limited interactive capabilities, failing to provide interactive means for fine-grained verification of the loading plan, such as layered expansion, layer-by-layer filtering, and item-by-item floating recognition.

[0006] US Patent 20240338911A1 discloses an augmented reality (AR) visualization method and system for ship loading, which uses AR devices to overlay virtual cargo models onto a real-world scene to assist in loading operations. This solution requires AR hardware (such as AR glasses or AR terminals), resulting in high deployment costs. Furthermore, its application is limited to on-site assistance during port loading operations and cannot meet the needs of international logistics personnel for remote viewing, verification, and adjustment of loading plans via a web browser.

[0007] Furthermore, Chinese patent CN114067048A discloses a visual intelligent system and construction method for material transfer, which builds a 3D scene and visualizes the material transfer plan; Chinese patent CN109840726A discloses a method, device, and computer-readable storage medium for item sorting, which automatically generates the sorting configuration of pending orders using the sorting configuration of historical orders. The above solutions involve 3D visualization and sorting calculation, but none achieve integrated implementation where the sorting algorithm calculation results directly drive real-time rendering of the 3D scene. The sorting results and 3D display require intermediate steps of data export, conversion, and re-import, resulting in fragmented operation processes. Moreover, none of these solutions are specifically designed for the loading constraints and fine-grained 3D interaction of multi-form goods in the lubricant industry.

[0008] In summary, existing technologies have the following technical problems: First, general packing algorithms are designed for standard cardboard box scenarios and cannot handle the mixed loading needs of various cargo types, such as drums, large drums, cardboard boxes, and pallet combinations, in the lubricant industry; Second, the visualization of packing solutions largely relies on offline rendering or dedicated hardware devices (depth cameras, AR devices), resulting in high deployment costs and the inability to run in real time in regular browsers; Third, 3D display solutions only provide static or simple rotational displays from a single perspective, and cannot perform layered viewing, layer-by-layer filtering, and item-by-item identification of the cargo inside the box; Fourth, the boxing calculation module and the visualization module are independent of each other, and the boxing results need to be exported and then imported into the visualization tool for viewing, resulting in a fragmented operation process and an inability to achieve seamless integration from algorithm calculation to 3D presentation. Summary of the Invention

[0009] To address the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent container sorting system for the international logistics lubricant industry, aiming to solve the following technical problems: First, to solve the problem of automated container sorting for mixed loading of lubricant products in various forms, i.e., how to automatically calculate the optimal sorting scheme based on parameters such as specifications, weight, and batch of different forms of goods such as medium drums, large drums, cartons, and pallet combinations, while meeting loading constraints such as heavy at the bottom and light at the top, large at the bottom and small at the top, and upper load limits, replacing manual planning item by item, and improving sorting efficiency and space utilization; Second, to solve the problem of real-time 3D visualization of the packing scheme on the Web, i.e., how to achieve real-time 3D rendering of containers and goods on the browser based on WebGL technology (Three.js engine), supporting multiple geometric shapes (paper... The system features differentiated modeling and coloring of various components (boxes, drums, large drums, pallet models, mats, partitions, strapping, etc.) to allow for intuitive presentation of packing solutions without the need for any plugins. Thirdly, it addresses the issue of fine-grained interactive operation in 3D scenes, namely, how to achieve layered view expansion, layer-by-layer filtering, and cargo floating detection and highlighting based on ray casting algorithms, as well as information floating window display, to meet users' needs for refined verification of packing solutions. Fourthly, it solves the problem of integrated boxing calculation and 3D visualization, namely, how to directly drive the real-time rendering of 3D scenes with the calculation results of intelligent boxing algorithms (cargo coordinates, rotation angle, layer information, shape type, batch identifier, etc.), achieving seamless connection from algorithm calculation to 3D presentation, and supporting automatic scene refresh and camera adaptive adjustment when switching between multiple containers.

[0010] The above-mentioned objective of this invention is achieved through the following technical solutions: An intelligent box sorting system for the international logistics lubricant industry includes a cargo attribute recognition module, a loading constraint engine, an automatic box sorting module, a web-based 3D rendering module, a layered expansion submodule, a layered filtering submodule, and a floating interactive submodule.

[0011] The cargo attribute recognition module is used to acquire cargo attribute data of the goods to be sorted and identify the cargo form. The cargo attribute data includes at least a cargo shape type field, specifications and dimensions, weight, batch number, and quantity. The cargo attribute recognition module classifies the goods into three forms—carton, medium drum, and large drum—based on the cargo shape type field, and performs packing calculations in conjunction with the weight and packaging size of the goods. It also configures corresponding geometric modeling parameters and rendering attributes for each form. Through this module, the system can identify multi-form goods unique to the lubricant industry, mapping the cargo shape type field (shape_type) returned from the backend to the corresponding three-dimensional geometric model, providing a unified cargo data foundation for subsequent automatic packing calculations and 3D rendering.

[0012] The loading constraint engine incorporates built-in rules for heavier items at the bottom and larger items at the top, prioritizing heavy drums and palletized goods exceeding preset weight thresholds at the bottom of the container. Lightweight, small items like cardboard boxes are placed on top. During the sorting process, the engine continuously monitors the total weight of the goods to ensure it does not exceed the container's weight limit, triggering an overload warning if it does. This loading constraint engine is a core safety feature in the lubricant industry's packing solutions. Its built-in principle of heavier items at the bottom and larger items at the top meets the safety requirements for transporting lubricant goods: placing heavier drums or palletized goods at the bottom lowers the container's center of gravity, preventing tipping during transport; placing lightweight cardboard boxes on top prevents damage from pressure. Simultaneously, real-time monitoring of the container's weight limit prevents overloading.

[0013] The automatic container sorting module sends the execution order identifier of the goods to be sorted to the server. The server then automatically calculates the sorting scheme based on the goods' specifications, weight, batch, quantity, and container size constraints, returning the container allocation result, internal coordinate position, rotation angle, layer information, and shape type for each item. The automatic sorting module server-sides the sorting calculation, delegating the complex NP-hard 3D container optimization calculations to the server, while the front-end only handles data presentation and interaction, balancing computational efficiency and user experience.

[0014] The web-based 3D rendering module is used to construct a 3D scene of containers and cargo on the browser side based on the WebGL graphics rendering engine. It directly drives the 3D scene rendering based on the container sorting results returned by the automatic container sorting module, generating 3D models of various cargo types. WebGL technology allows the browser to perform hardware-accelerated 3D rendering without installing any plugins. The Three.js engine provides comprehensive abstract interfaces for scenes, cameras, renderers, geometry, materials, and lighting, enabling the efficient construction and rendering of various models such as container wireframes, medium drums, large drums, cartons, pallets, dunnage, partitions, and strapping. The web-based 3D rendering module includes a layered expansion submodule, a layer-by-layer filtering submodule, and a floating interaction submodule. These three submodules work together to enable fine-grained viewing and verification of the packing plan in the 3D scene.

[0015] The layered unfolding submodule (belonging to the web-based 3D rendering module) is used to calculate the inter-layer offset based on the height of each layer of goods inside the container, and to separate and display each layer of goods vertically, distinguishing the boundaries of each layer with different colored wireframes. Through the layered unfolding view, the arrangement of the bottom layer of goods, which was originally obscured by the upper layer of goods, can be clearly presented, allowing users to intuitively view the arrangement of each layer of goods and meet the needs of detailed verification.

[0016] The layer-by-layer filtering submodule is used to maintain layer-level filtering variables. During the rendering loop, it determines whether the layer number of each item matches the layer-level filtering variables. Items that do not match are skipped from rendering, allowing items at the specified layer to be displayed separately. Layer-by-layer filtering and layer-by-layer expansion complement each other: layer-by-layer expansion separates each layer vertically for overall observation, while layer-by-layer filtering renders only the layer specified by the user, eliminating visual interference from other layers and facilitating layer-by-layer verification.

[0017] The floating interaction submodule (part of the web-based 3D rendering module) listens for mouse movement events and detects goods below the mouse position using a raycasting algorithm. Upon a hit, the hit goods are highlighted and a floating information window pops up. This window displays at least the goods' SKU, stack number, layer, size, and shape type. Through raycasting collision detection, the system can accurately identify the individual goods the mouse is pointing at and present their detailed information using highlighted materials and a floating information window, enabling item-by-item identification and information traceability within the container.

[0018] Furthermore, in the cargo attribute recognition module, the geometric modeling parameters for the three types of shapes—carton, medium drum, and large drum—include length, width, and height dimensions or bottom diameter and height, and their respective rendering attributes include color, texture, and transparency. The system also includes pallet modeling subunits, pallet modeling subunits, partition modeling subunits, and strapping rendering subunits, used to render the models of pallets, pallets, partitions, and strapping in the 3D scene respectively, to recreate the actual loading form of the lubricating oil cargo. Through these extensions, the 3D scene not only presents the cargo itself but also the loading accessories such as pallets, pallets, partitions, and strapping, ensuring a high degree of consistency between the packing plan and the actual loading form on site.

[0019] Furthermore, the loading constraint engine is also used to compare the total weight of the goods in the current compartmentation scheme with the container's maximum load capacity in real time during the compartmentation process. If the total weight exceeds the container's maximum load capacity, a secondary confirmation prompt will appear. The total weight is obtained in real time through a high-precision calculation library, where the weight of a single item is the product of the item's unit weight and the quantity of items. The use of a high-precision calculation library (such as BigNumber) avoids the loss of precision in floating-point operations, ensuring the accuracy of weight accumulation. The secondary confirmation prompt gives the user the final decision-making power, balancing security and operational flexibility.

[0020] Furthermore, the automatic container splitting module completes the container splitting calculation by calling the automatic container splitting interface. The automatic container splitting interface receives the execution order identifier. The server calculates the container splitting scheme based on the cargo specifications, weight, batch, quantity, and container size and loading constraint rules, and returns the container allocation result and internal coordinate position of each cargo through the automatic container splitting interface.

[0021] Furthermore, the system also includes a manual sorting assistance module, which provides a two-way interface for listing goods to be sorted and goods already sorted. After the user selects the goods to be sorted and enters the quantity, the system uses a high-precision calculation library to accumulate the total weight of the selected goods in real time and compares the accumulated result with the container's weight limit. If the weight limit is exceeded, a secondary confirmation pop-up appears. The manual sorting assistance module also supports batch sorting via template import and one-click reset of sorting results. The manual sorting assistance module provides a convenient channel for manual intervention and is suitable for scenarios where automatic sorting results need fine-tuning or where users want to adjust the loading plan according to their own experience.

[0022] Furthermore, in the layered unfolding submodule, the inter-layer offset is calculated based on the maximum height of each layer of goods. After each layer of goods is separated vertically, the boundaries between layers are distinguished by blue, green, and orange colored frames to visually represent the placement range of each layer of goods. The use of three-color frames to distinguish layer boundaries provides a clear color contrast, making it easy for users to quickly identify different layers.

[0023] Furthermore, the system also includes a compact stacking submodule, used to sort the goods in the 3D scene according to a three-level sorting rule of hierarchy, shape weight, and index in a merged view. It calculates the compact reference plane of each item through footprint overlap detection, ensuring that adjacent layers of goods fit tightly together vertically, eliminating gaps between layers. Compact stacking can restore the true loading state of the goods after compaction in a merged view, avoiding visual misleading effects caused by gaps between layers in the unfolded state.

[0024] Furthermore, in the floating interaction submodule, highlighting the hit goods is achieved by enhancing the self-illuminating property of the hit goods' material; the floating information window is a tooltip that follows the mouse movement and automatically hides when the mouse moves away from the hit goods. Highlighting is achieved by enhancing the material's self-illuminating property, significantly improving the visual recognition of the hit goods without changing the model's geometry; the floating information window follows the mouse movement, avoiding obscuring other goods and providing a smooth operating experience.

[0025] Furthermore, the system also includes a camera adaptation submodule and a multi-container switching submodule. The camera adaptation submodule automatically calculates the optimal viewing distance and target point based on the container size and the current view mode, which includes a merged view mode and a layered view mode. The multi-container switching submodule matches the corresponding stacking data in the 3D data according to the container identifier and triggers re-rendering when switching containers in the container list, while simultaneously driving the camera adaptation submodule to adjust the camera to the optimal viewing angle. Camera adaptation ensures that the user has a complete field of view regardless of the container size or whether it is in a merged or layered view; multi-container switching enables seamless browsing between multiple containers.

[0026] Furthermore, the system also includes a direct-drive integration module, which acquires container dimensions, stack attributes, and cargo coordinates, rotation angles, shape types, and hierarchical data through a 3D data interface. This data directly drives the web-based 3D rendering module to perform scene rendering without intermediate data conversion. The direct-drive integration module also monitors data changes through a responsive listening mechanism, automatically performing data loading and scene reconstruction when data changes, and releasing the geometry and material resources of the old model before scene reconstruction to avoid memory leaks. The direct-drive integration module is key to achieving the "integration of container segmentation calculation and 3D visualization" in this invention: the backend container segmentation calculation results directly drive the frontend rendering in the form of 3D data, eliminating the intermediate steps of data export, format conversion, and then importing into visualization in traditional solutions; the responsive listening mechanism and resource release mechanism ensure system stability and smoothness in frequent operation scenarios such as multi-container switching and data updates.

[0027] In summary, compared with the prior art, the present invention includes at least one of the following beneficial technical effects: This invention achieves automated packing of lubricating oil in various forms. Through a cargo shape classification and loading constraint engine, the system can automatically calculate packing schemes based on parameters such as cargo specifications, weight, and batch number, replacing manual planning for each item. This significantly improves packing efficiency and space utilization, and reduces transportation costs. Compared to the general 3D cargo packing method based on a depth camera disclosed in CN112607234B, this invention specifically designs three types of shape classification and differentiated geometric modeling for mixed loading scenarios of lubricating oil cargo in various forms, including drums, large drums, cartons, and pallet combinations. It also incorporates industry-specific loading constraints such as heavy-on-bottom, large-on-bottom, small-on-bottom, and load-bearing limits, overcoming the shortcomings of general packing algorithms in handling mixed loading of multiple forms and industry-specific constraints.

[0028] This invention enables real-time 3D visualization of packing solutions on a web-based platform. Based on the Three.js WebGL engine, 3D rendering is completed in the browser without the need for plugins or dedicated hardware. It supports differentiated modeling of various shapes, including cartons, drums, large drums, pallets, pallets, and strapping, making packing solutions intuitively identifiable. Unlike CN112607234B, which relies on depth camera acquisition and offline modeling on the Revit desktop, and US20240338911A1, which relies on AR hardware, this invention achieves 3D rendering entirely within a regular browser using WebGL hardware acceleration. This significantly reduces deployment costs and improves the accessibility and versatility of packing solutions, allowing direct viewing on any terminal device with a browser.

[0029] This invention enables fine-grained interaction within a 3D scene. Through layered expansion and layer-by-layer filtering, users can view the arrangement of goods at the bottom layer that are obscured by the upper layers. By using light projection and hover detection, each item can be identified, and detailed information such as SKU, layer, and size can be viewed, meeting the needs of refined verification. Existing technologies (such as US11820611B2) only provide basic interactions such as overall rotation and scaling in 3D display, and cannot perform layered viewing and item-by-item identification of goods within the box. The fine-grained interaction method of this invention fills this gap.

[0030] This invention achieves integrated processing of container segmentation calculations and 3D visualization. Backend calculation results directly drive frontend 3D scene rendering, eliminating the need for intermediate data export and import. Scenes automatically refresh and cameras adaptively adjust during multi-container switching, ensuring smooth and seamless operation. Compared to solutions such as CN114067048A and CN109840726A that separate container segmentation calculations from visualization, this invention's direct data-driven mechanism achieves a seamless connection between algorithm calculation results and 3D presentation.

[0031] By implementing an automatic scene resource release mechanism, the geometry and material resources of the old model are released in a timely manner when switching containers or updating data, avoiding memory leaks caused by prolonged use and ensuring the stability of system operation. Attached Figure Description

[0032] Figure 1 This is a system structure block diagram of the intelligent box sorting system for the international logistics lubricating oil industry of the present invention.

[0033] Figure 2 This is an operation flowchart of the intelligent box sorting system for the international logistics lubricant industry of the present invention. Detailed Implementation

[0034] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0035] In the description of this application, it should be noted that the terms "upper," "lower," "inner," "outer," "top / bottom," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0036] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installed," "equipped with," "sleeved / connected," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0037] Example 1: System Overall Architecture like Figure 1 As shown in the figure, this embodiment provides an overall architecture of an intelligent box sorting system for the international logistics lubricant industry. The architecture adopts a three-layer structure, which consists of a data / interface layer, an application layer, and a user layer from bottom to top. The layers interact with each other through standard interfaces.

[0038] The user layer is a browser (web-based). Users access the binning system through a browser without needing to install any plugins or dedicated clients. The browser runs a WebGL-based Three.js rendering engine and communicates with the application layer via standard network protocols (such as HTTP / HTTPS). As the system's entry point, the user layer provides users with a binning operation interface and a 3D visualization interface. Its core feature is that all 3D rendering and interactive functions are completed on the browser side, without relying on dedicated hardware such as depth cameras or AR devices, nor on desktop offline modeling software such as Revit. This significantly reduces deployment costs and improves the accessibility and versatility of the solution.

[0039] The application layer is the core business logic layer of the system, which includes five major functional modules: intelligent container sorting module, 3D visualization module (core module), container management module, WMS collaboration module, and sorting and shipping inspection module.

[0040] The intelligent container sorting module provides functions such as automatic sorting algorithm, manual sorting operation, template import sorting, load-bearing verification, resetting sorting, and completing packing. The automatic sorting algorithm calls the autoPack interface in the data / interface layer, passing the execution single ID to the server, which then performs the sorting calculation and returns the result. The manual sorting operation provides bidirectional operation between the list of goods to be sorted and the list of already sorted goods. Users can select goods and enter the quantity to be moved. The system uses the BigNumber high-precision calculation library to accumulate the total weight of the selected goods in real time and compares it with the container's load-bearing limit. Template import sorting allows users to batch import sorting configurations using templates such as Excel. The load-bearing verification function verifies the relationship between the total weight of the goods and the container's load-bearing limit (containerLoadGw) in real time at each stage of the sorting process, triggering a warning prompt and requiring user confirmation when the load exceeds the limit. Resetting sorting clears the current sorting result with a single click. Completing packing confirms the final packing plan and submits it.

[0041] The 3D visualization module is the core module of this invention. Its internal functions include: a Three.js rendering engine, responsible for the creation, rendering loop, and scene management of 3D scenes; container wireframe modeling, used to present the boundaries and internal space of containers in wireframe mode; multi-form cargo rendering, used to render models of various forms of cargo such as cartons, medium drums, and large drums; pallet / dustboard modeling, used to render pallet and dustboard models; strapping rendering, used to render strapping models; layered expansion / layered filtering, used to achieve fine-grained hierarchical viewing of 3D scenes; light projection and hover interaction, used to achieve cargo detection and information display when the mouse hovers; and multi-container switching / camera adaptation, used to achieve seamless switching and browsing between multiple containers and automatic adjustment of the best viewing angle.

[0042] The container management module provides functions such as container list display, container type editing and modification, loading method setting, fumigation requirement arrangement, and stamping / certification marking. The container list display shows all containers associated with the current execution order and their status; container type editing and modification allows adjustment of container dimensions and type parameters; loading method setting specifies the loading method for the container (e.g., full container load, less-than-container load, etc.); fumigation requirement arrangement marks containers requiring fumigation and records their fumigation status; and stamping / certification marking records the status of relevant certification documents for the container.

[0043] The WMS collaboration module provides functions such as synchronizing WMS, canceling WMS synchronization, and tracking synchronization status. It is used to achieve data collaboration between the bin sorting system and the warehouse management system (WMS), synchronize bin sorting results to the WMS system, cancel executed synchronization operations, and track synchronization status to ensure data consistency between systems.

[0044] The Tally Sheet Delivery Order (TSDO) module provides functions such as TSDO PDF preview, TSDO Excel export, delivery inspection confirmation, special requirements viewing, and approval record management. TSDOs are Tally Sheet Delivery Orders. The module supports online preview of TSDO documents in PDF format and export in Excel format for easy download and archiving. The delivery inspection confirmation is used to confirm various pre-delivery checks (such as contracts, payments, business transactions, EUC, etc.). The special requirements viewing module allows users to view the special requirements of the execution order and its approval records.

[0045] The data / interface layer provides various backend service interfaces required for system operation, including: the `packingFind` interface for querying container allocation plans; the `autoPack` interface for performing automatic container allocation calculations; the `getExec3D` interface for obtaining 3D data of the execution order (container dimensions, stack attributes, cargo coordinates / rotation / shape / layer, etc.); the `getContainers` interface for obtaining a list of containers and container information; the `snyWms` interface for synchronizing WMS data; the `reqfumigated` interface for submitting fumigation requests; and the `finishPack` interface for completing packing and submitting the final plan. The data / interface layer exposes the server-side computing capabilities to the application layer in the form of interfaces, achieving a front-end / back-end separation architecture, decoupling yet linking the front-end 3D rendering with the back-end container allocation calculations.

[0046] In this embodiment, the functional modules of the application layer interact with the backend service through the data / interface layer: the intelligent container sorting module calls the autoPack, packingFind, and finishPack interfaces to complete container sorting calculations and submissions; the 3D visualization module calls the getExec3D and getContainers interfaces to obtain 3D rendering data; the container management module calls the getContainers and reqfumigated interfaces to manage container and fumigation requirements; the WMS collaboration module calls the snyWms interface to synchronize data; and the sorting list and shipping inspection module obtains TSDO document data and approval record data through the backend service.

[0047] This embodiment also includes a cargo attribute recognition module, used to acquire cargo attribute data of the goods to be sorted and identify the cargo form. The cargo attribute data includes at least a cargo shape type field, specifications and dimensions, weight, batch number, and quantity. The cargo attribute recognition module classifies the goods into three forms—carton, medium drum, and large drum—based on the cargo shape type field, and performs packing calculations in conjunction with the weight and packaging size of the goods. It also configures corresponding geometric modeling parameters and rendering attributes for each form. Through this module, the system can identify the multi-form cargo unique to the lubricant industry, mapping the cargo shape type field (shape_type) returned from the backend to the corresponding three-dimensional geometric model, providing a unified cargo data foundation for subsequent automatic packing calculations and 3D rendering. In the cargo attribute recognition module, the geometric modeling parameters for the three types of containers—cartons, medium drums, and large drums—include length, width, and height dimensions or bottom diameter and height. Their rendering attributes include color, texture, and transparency. The system also includes pallet modeling subunits, pallet modeling subunits, partition modeling subunits, and strapping rendering subunits, used to render models of pallets, pallets, partitions, and strapping in the 3D scene respectively, to recreate the actual loading form of the lubricating oil cargo. Through these extensions, the 3D scene not only presents the cargo itself but also the loading accessories such as pallets, pallets, partitions, and strapping, ensuring a high degree of consistency between the packing plan and the actual loading form on site.

[0048] The international logistics of the lubricant industry involves diverse cargo forms, primarily including: cartons (common sizes such as 6×1L gift boxes, 4×4L full cartons, and 4L small drum cartons), medium drums (common sizes such as 18L medium drums), and large drums (common sizes such as 200L large drums). The cargo attribute recognition module categorizes cargo into three types based on the cargo shape type field (shape_type) returned from the backend: cartons, medium drums, and large drums. Each type is assigned different geometric modeling parameters and rendering attributes, and packing calculations are performed in conjunction with the weight and packaging size of the goods.

[0049] Specifically, for the carton shape, its geometric modeling parameters include three dimensions: length (L), width (W), and height (H). The model is constructed using BoxGeometry in Three.js, and the rendering attributes can be configured to the common colors of the carton (such as yellow and brown) and textures. For the middle drum shape, its geometric modeling parameters include the base diameter (D) and height (H). The model is constructed using CylinderGeometry in Three.js, and the rendering attributes can be configured to the common colors of the middle drum (such as blue and silver) and the body texture. A lid model can be added if necessary. For the large drum shape, its geometric modeling parameters include the base diameter (D) and height (H). The model is constructed using CylinderGeometry in Three.js, and the rendering attributes can be configured to the common colors of the large drum (such as dark blue and green) and the body texture. A lid model can be added if necessary.

[0050] In terms of rendering attributes, all types of forms support the configuration of color, texture, and opacity. Color attributes are used to distinguish different goods or different batches of goods; texture attributes are used to present surface details such as the body of oil drums and the printing on cardboard boxes; and opacity attributes are used to present the semi-transparent features of auxiliary materials such as partitions, making the primary and secondary information in the 3D scene clearly distinguishable.

[0051] In this embodiment, the system also includes a pallet modeling subunit, a mat modeling subunit, a partition modeling subunit, and a strapping rendering subunit. The pallet modeling subunit renders the pallet model in the 3D scene. The pallet model is a combination of a cuboid and legs, representing the bottom layer of a full pallet of lubricating oil. The mat modeling subunit renders the mat model. The mat is a flat plate laid at the bottom of the pallet or container to distribute the pressure of the goods. The partition modeling subunit renders the partition model. The partitions stand vertically or lie flat between the goods to separate different batches or different sizes of goods. The strapping rendering subunit renders the strapping model. The strapping is presented as a strip-shaped geometry surrounding the goods, used to secure the goods on the pallet. Through the differentiated modeling of the above loading auxiliary materials, the 3D scene can restore the realistic form of lubricating oil loading, making the presentation of the packing scheme highly consistent with the actual site conditions.

[0052] This embodiment also includes a loading constraint engine, which incorporates loading constraint rules such as "heavy weight at the bottom, light weight at the top" and "large weight at the bottom, small weight at the top." This prioritizes heavy drums and palletized goods exceeding a preset weight threshold to the bottom of the container, while lightweight, small items like cardboard boxes are placed on the top. During the sorting process, the total weight of the goods is checked in real-time to ensure it does not exceed the container's weight limit; if it does, an overweight warning is triggered. The loading constraint engine is a core guarantee of safety in lubricant industry packing solutions. Its built-in principle of "heavy weight at the bottom, light weight at the top" and "large weight at the bottom, small weight at the top" meets the safety requirements for transporting lubricant goods: placing heavier drums or palletized goods at the bottom lowers the overall center of gravity, avoiding the risk of tipping over during transport; placing lightweight cardboard boxes on the top prevents damage from pressure. Simultaneously, real-time verification of the container's weight limit prevents overloading. The loading constraint engine also compares the total weight of the goods in the current packing scheme with the container's maximum load capacity in real time during the packing process. If the total weight exceeds the container's maximum load capacity, a secondary confirmation prompt appears. The total weight is calculated in real time using a high-precision calculation library, where the weight of a single item is the product of the item's unit weight and the quantity of items. The use of a high-precision calculation library (such as BigNumber) avoids the loss of precision in floating-point operations, ensuring the accuracy of weight accumulation. The secondary confirmation prompt gives the user the final decision-making power, balancing security and operational flexibility.

[0053] This embodiment details the implementation of the loading constraint engine. The loading constraint engine is the core guarantee of the safety of the packing scheme in the lubricating oil industry. It has two built-in constraint principles: the first is the principle of heavier items at the bottom and lighter items at the top, that is, heavier items are placed on the lower layer and lighter items are placed on the upper layer; the second is the principle of larger items at the bottom and smaller items at the top, that is, larger items are placed on the lower layer and smaller items are placed on the upper layer.

[0054] During the automatic sorting process, the loading constraint engine prioritizes heavy-duty containers and palletized goods with a weight exceeding a preset weight threshold, placing them at the bottom of the container. Specifically, the system calculates the total weight of each item or pallet based on the weight field (unitGw) and quantity field (qty) in the cargo attribute data. Heavy-duty containers and palletized goods with a total weight exceeding a preset weight threshold (e.g., 100kg) are placed at the bottom, while lightweight, small cardboard boxes are stacked on top. The loading constraint engine also verifies the load-bearing capacity of the goods: for cardboard boxes with weak load-bearing capacity, the weight of goods above them must not exceed the box's maximum load-bearing capacity to prevent damage to the bottom cardboard boxes.

[0055] At each stage of the container sorting process, the loading constraint engine verifies in real time whether the total weight of the cargo exceeds the container's maximum load capacity (containerLoadGw). The maximum load capacity is the maximum permissible weight of a container, determined by the container type (e.g., 20-foot and 40-foot containers have different maximum load capacities). When the total weight of the accumulated cargo exceeds the container's maximum load capacity, the system triggers an overweight warning, displaying a message informing the user of the current overweight situation, allowing the user to confirm whether to continue or adjust the plan.

[0056] Furthermore, the loading constraint engine is also used to compare the total weight of the current packing scheme with the container's maximum load capacity in real time during the packing process. A secondary confirmation prompt will pop up if the total weight exceeds the container's maximum load capacity. The total weight is obtained in real time through a high-precision calculation library (such as BigNumber), where the weight of a single item is the product of the unit weight (unitGw) and the quantity (qty). The reason for using the BigNumber high-precision calculation library is that floating-point arithmetic can lead to precision loss when accumulating a large number of decimal weight values ​​(such as the problem of 0.1 + 0.2 ≠ 0.3). BigNumber stores values ​​as strings or fixed-point numbers, ensuring the accuracy of the weight accumulation result and avoiding misjudgments in load capacity verification due to precision errors.

[0057] This embodiment also includes an automatic container splitting module and a manual container splitting auxiliary module. The automatic container splitting module sends the execution order identifier of the goods to be split to the server. The server automatically calculates the splitting scheme based on the goods' specifications, weight, batch, quantity, and container size constraints, and returns the container allocation result, internal coordinate position, rotation angle, layer information, and shape type for each item. The automatic container splitting module realizes server-side implementation of the splitting calculation, delegating the complex NP-hard 3D container loading optimization calculation to the server. The front end is only responsible for data presentation and interaction, balancing calculation efficiency and user experience. The automatic container splitting module completes the splitting calculation by calling the automatic container splitting interface. The automatic container splitting interface receives the execution order identifier. The server calculates the splitting scheme based on the goods' specifications, weight, batch, quantity, and container size and loading constraint rules, and returns the container allocation result and internal coordinate position for each item through the automatic container splitting interface.

[0058] The manual sorting assistance module provides a two-way interface for listing both goods to be sorted and already sorted goods. After the user selects the goods to be sorted and enters the quantity, the system uses a high-precision calculation library to accumulate the total weight of the selected goods in real time and compares the accumulated result with the container's weight limit. If the weight exceeds the limit, a secondary confirmation pop-up appears. The manual sorting assistance module also supports batch sorting via template import and one-click reset of sorting results. This module provides a convenient channel for manual intervention and is suitable for scenarios where automatic sorting results require fine-tuning or where users wish to adjust the loading plan based on their experience.

[0059] The automatic container splitting module performs container splitting calculations by calling the automatic container splitting interface (autoPack). Specifically, the system passes the execution order ID to the autoPack interface. Upon receiving the request, the server retrieves the cargo data (specifications, weight, batch, quantity) and container data (dimensions, maximum load capacity) associated with the execution order. Combining loading constraints such as heavier items on the bottom and lighter items on top, larger items on the bottom and smaller items on top, and maximum load capacity, it calls a 3D container loading optimization algorithm (such as a spatial partitioning heuristic algorithm or a multi-level tree search algorithm) to calculate the container splitting scheme. The calculation results include: the container identifier (container_id) assigned to each cargo, the 3D coordinates (x, y, z) of the cargo within the container, the rotation angle of the cargo (rotation amount around each axis), the cargo's layer level (layer_no), the cargo's shape type (shape_type), and the cargo's batch identifier (batch_id). The autoPack interface returns the above calculation results to the front end as structured data (such as JSON format).

[0060] The manual container sorting assistance module provides a two-way interface for both the list of goods to be sorted and the list of goods already sorted. Users select goods in the list of goods to be sorted and enter the quantity to be moved. The system then moves the selected goods from the list of goods to be sorted to the list of goods already sorted and updates the container's loading status. During the movement process, the system uses the BigNumber high-precision calculation library to accumulate the total weight of the selected goods in real time (unitGw × qty) and compares it with the container's maximum load capacity. If the weight exceeds the limit, a secondary confirmation prompt appears, allowing the user to choose to continue or cancel. The manual container sorting assistance module also supports importing batch sorting configurations from Excel templates, suitable for scenarios with a large number of goods and low efficiency in manual operation. It also supports one-click reset of sorting results, allowing users to clear the current sorting and replan at any time. After manual sorting is completed, the system stores both the manual and automatic sorting results for rendering and display by the 3D visualization module.

[0061] This embodiment also includes a web-based 3D rendering module, used to construct a 3D scene of containers and cargo on the browser side based on the WebGL graphics rendering engine. It directly drives the 3D scene rendering based on the container sorting results returned by the automatic sorting module, generating 3D models of various cargo types. WebGL technology allows browsers to perform hardware-accelerated 3D rendering without installing any plugins. The Three.js engine provides comprehensive abstract interfaces for scenes, cameras, renderers, geometry, materials, and lighting, enabling the efficient construction and rendering of various models such as container wireframes, medium drums, large drums, cartons, pallets, dunnage, partitions, and strapping. The web-based 3D rendering module includes a layered expansion submodule, a layer-by-layer filtering submodule, and a floating interaction submodule. These three submodules work together to achieve fine-grained viewing and verification of the packing scheme in the 3D scene.

[0062] The web-based 3D rendering module uses the WebGL graphics rendering engine (Three.js) to build 3D scenes in the browser. Three.js is a mainstream WebGL wrapper library, providing comprehensive abstract interfaces for Scene, PerspectiveCamera, WebGLRenderer, BoxGeometry, CylinderGeometry, MeshBasicMaterial / MeshStandardMaterial, and AmbientLight / DirectionalLight. The browser directly calls the GPU for hardware-accelerated rendering via the WebGL API, requiring no plugins.

[0063] The process of constructing a 3D scene is as follows: First, a Scene object is created as the root node of the scene. A WebGLRenderer is created and mounted to the page DOM element. A PerspectiveCamera is created and its initial position and viewing direction are set. Second, a container wireframe model is created based on the container size. The container wireframe model is presented in the form of EdgesGeometry or LineSegments, showing the length, width, and height boundaries of the container. Third, the container segmentation data is traversed. A corresponding geometric model is created based on the shape type (shape_type) of each cargo (BoxGeometry for cardboard boxes, CylinderGeometry for medium and large buckets). The position and orientation of the model are set according to the 3D coordinates (x, y, z) and rotation angle of the cargo. The layer of the model is set according to the layer information (layer_no) of the cargo. The material and color of the model are set according to the rendering attributes of the cargo. Finally, ambient light and directional light are added, and the rendering loop (requestAnimationFrame) is started to continuously render the scene.

[0064] The system also includes a direct-drive integration module, which acquires container dimensions, stack attributes, and cargo coordinates, rotation angles, shape types, and hierarchical data through a 3D data interface. This data directly drives the web-based 3D rendering module to perform scene rendering without intermediate data conversion. The direct-drive integration module also uses a responsive monitoring mechanism to listen for data changes. When data changes, it automatically performs data loading and scene reconstruction, releasing the geometry and material resources of the old model before scene reconstruction to avoid memory leaks. The direct-drive integration module is key to achieving the "integration of container segmentation calculation and 3D visualization" in this invention: the backend container segmentation calculation results directly drive the frontend rendering in the form of 3D data, eliminating the intermediate steps of data export, format conversion, and then importing into visualization as in traditional solutions. The responsive monitoring mechanism and resource release mechanism ensure system stability and smoothness in frequent operation scenarios such as multi-container switching and data updates.

[0065] The direct-drive integration module is the key module for achieving "integrated container splitting calculation and 3D visualization" in this invention. The container splitting results generated by the automatic or manual module are used to obtain 3D data via the getExec3D interface. This data includes container dimensions, stack attributes (height and number of layers), coordinates (x, y, z), rotation angle, shape type (shape_type), layer number (layer_no), and batch identifier (batch_id) for each cargo. This data directly drives the web-based 3D rendering module to render the scene without intermediate data conversion. When a user selects a specific container from the multi-container list, the system matches the corresponding stacking data in the 3D data according to the container identifier (container_id), triggering the regenerate function to reconstruct the 3D scene of that container.

[0066] The data-driven integration module also uses a reactive listening mechanism (such as Vue's `watch`) to monitor data changes and automatically performs data loading and scene reconstruction when data changes. Specifically, the system sets reactive listeners for bin sorting result data objects. When the data changes (such as manually adjusting bins, switching containers, or refreshing data), the listener callback is triggered, and the system automatically executes the data loading function to obtain the latest 3D data and calls the scene reconstruction function to update the 3D scene. Before scene reconstruction, the system releases the geometry and material resources of the old model. Specifically, it releases GPU resources by calling the `geometry.dispose()` and `material.dispose()` methods and removes the old model objects to avoid memory leaks caused by prolonged use and ensure system stability.

[0067] It also includes a layered unfolding submodule (part of the web-based 3D rendering module), used to calculate the inter-layer offset based on the height of each layer of goods within the container, separating and displaying each layer of goods vertically, and distinguishing the boundaries of each layer with different colored wireframes. Through the layered unfolding view, the arrangement of goods at the bottom layer, which was originally obscured by goods at the top, can be clearly presented, allowing users to intuitively view the arrangement of goods at each layer, meeting the needs of detailed verification. In the layered unfolding submodule, the inter-layer offset is calculated based on the maximum height of each layer of goods. After each layer of goods is separated vertically, the boundaries between layers are distinguished by blue, green, and orange wireframes to intuitively present the placement range of goods at each layer. The use of three-color wireframes to distinguish layer boundaries provides a clear color contrast, making it easy for users to quickly identify different layers.

[0068] The layered unfold view addresses the issue of lower-level goods being obscured by upper-level goods in a 3D scene, making them invisible. In the merged view (rendered according to actual loading positions), lower-level goods are covered by upper-level goods, preventing users from observing their specific arrangement. The layered unfold submodule separates and displays each layer of goods sequentially along the vertical direction (Y-axis), ensuring that each layer is clearly visible.

[0069] The specific implementation of layered unfolding is as follows: The system maintains the layer data of goods in each layer, calculates the maximum height of goods in each layer, and calculates the inter-layer offset based on the maximum height of goods in each layer. The inter-layer offset is calculated as follows: For the i-th layer (i=1,2,...,n), its inter-layer offset offset_i = Σ(h_j), where j=1,2,...,i-1, that is, the sum of the maximum heights of the goods in the first i-1 layers plus the layer spacing constant d (e.g., 0.5 meters), ensuring that there are clearly visible gaps between the unfolded layers. During rendering, all goods in the i-th layer are offset along the Y-axis based on their original coordinates to achieve vertical separation of the layers.

[0070] After the goods are separated vertically, the boundaries between layers are distinguished by blue, green, and orange colored outlines. Specifically, the system creates a layer boundary outline for each layer: the boundary outline of layer 1 (bottom layer) is set to blue, the boundary outline of layer 2 (middle layer) is set to green, the boundary outline of layer 3 (top layer) is set to orange, and so on, using the three colors in a cyclical manner. The layer boundary outline is constructed using EdgesGeometry and LineSegments, presenting as the rectangular boundary of the area occupied by the goods in that layer. Users can quickly identify the range of different layers by color. After the layers are unfolded, users can intuitively view the arrangement of goods in each layer, including the placement direction of heavy buckets at the bottom layer, the stacking method of cartons in the middle layer, and the distribution of lightweight small items at the top layer.

[0071] It also includes a layer-by-layer filtering submodule, which maintains layer-level filtering variables. During the rendering loop, it checks whether the layer number of each item matches the layer-level filtering variables. Items that do not match are skipped from rendering, allowing items at the specified layer to be displayed separately. Layer-by-layer filtering and layer-by-layer expansion complement each other: layer-by-layer expansion separates each layer vertically for overall observation, while layer-by-layer filtering renders only the user-specified layer, eliminating visual interference from other layers and facilitating layer-by-layer verification.

[0072] The layer-by-layer filtering function is used to display goods at a specified layer separately, eliminating visual interference from other layers. The system maintains a layer filter variable (filterLayer), which can be empty (displaying all layers) or a specified layer number (such as layer 2). In the main rendering loop, the system iterates through each goods object and checks whether the goods' layer number (layer_no) matches the layer filter variable: if the layer filter variable is empty, all goods are rendered; if the layer filter variable is a specified layer number, only goods with a layer number equal to that specified value are rendered, and goods that do not match are skipped during rendering (i.e., they are not added to the rendering queue, or their visible property is set to false).

[0073] By filtering by layer, users can view the layout of goods on a specific layer, such as only viewing the placement of heavy containers on the bottom layer or only viewing the distribution of lightweight small items on the top layer. The layer filtering and layer expansion functions can be used in combination: users can first separate the layers by expanding them, and then use layer filtering to display only the specified layers, achieving "detailed verification within each layer".

[0074] It also includes a compact stacking submodule, which sorts the goods in the 3D scene according to a three-level sorting rule of hierarchy, shape weight, and index in the merged view. It calculates the compact reference plane of each item through footprint overlap detection, so that adjacent layers of goods fit tightly together in the vertical direction, eliminating gaps between layers. Compact stacking can restore the true loading state of the goods after they are compacted in the merged view, avoiding visual misleading caused by gaps between layers in the layered unfolded state.

[0075] The Compact Stacking feature is used to recreate the actual loading state of goods after compaction in the merged view. In the layered unfolded view, the goods in each layer are separated vertically, with gaps between layers. When the user switches back to the merged view, if rendered directly according to the original coordinates, the goods in each layer will be restored to their actual loading positions, but gaps between layers may still exist due to calculation errors or data discrepancies. The Compact Stacking submodule eliminates these gaps in the following ways: First, the sorting rules. In the merged view, the system sorts the goods in the 3D scene according to a three-level sorting rule: layer_no, shape weight, and index. Shape weight is used to identify the type priority of the goods; for example, bottom-level goods such as heavy buckets and full pallets have higher shape weights, while top-level goods such as cartons have lower shape weights. The index represents the original order of the goods in the data. The purpose of sorting is to process the goods sequentially from the bottom to the top, ensuring the correctness of the stacking calculation.

[0076] Second, footprint overlap detection. The system calculates the compact reference plane of each item: for the currently processed item, it detects whether its horizontal projection (footprint) overlaps with the horizontal projection of the processed item below. If there is an overlap, the item is placed on top of the item below, so that the two are closely fitted; if there is no overlap (e.g., the item is located in a suspended area), the item is moved vertically downwards to the nearest support surface. Through footprint overlap detection, the system can determine the final position of each item in the vertical direction, eliminating gaps between layers.

[0077] Third, the reference plane is calculated. The system calculates a compact reference plane for each piece of cargo, which is the final vertical position of the cargo. The calculation of the compact reference plane takes into account the top height of the cargo below, the bottom position of the current cargo, and loading constraints (such as the principle of heavier cargo at the bottom and lighter cargo at the top), ensuring that the cargo fits tightly while meeting loading safety requirements. Through the above processing, the cargo in the merged view presents a compactly stacked true loading form, consistent with the actual state after on-site packing.

[0078] It also includes a floating interaction submodule (part of the web-based 3D rendering module), used to listen for mouse movement events and detect goods below the mouse position using a raycasting algorithm. Upon a hit, the hit goods are highlighted and a floating information window pops up, displaying at least the goods' SKU, stack number, layer, size, and shape type. Through raycasting collision detection, the system can accurately identify the individual goods the mouse is pointing at and present their detailed information in a highlighted material and floating information window, enabling item-by-item identification and information traceability within the box. In the floating interaction submodule, highlighting the hit goods is achieved by enhancing the self-illuminating properties of the hit goods' materials; the floating information window is a tooltip that follows the mouse movement and automatically hides when the mouse moves away from the hit goods. Highlighting by enhancing the material's emissive properties significantly improves the visual recognizability of the hit goods without altering the model's geometry; the floating information window follows the mouse movement, avoiding obscuring other goods and providing a smooth user experience.

[0079] The hover interaction function is used to identify and display information about each item inside the box. The system listens for mouse movement events and uses a raycasting algorithm to detect items below the mouse position during mouse movement.

[0080] The implementation principle of the ray casting algorithm is as follows: The system calculates the two-dimensional coordinates (clientX, clientY) of the mouse on the screen into a ray in three-dimensional space through the camera projection matrix. The starting point of the ray is the camera position, and the direction is towards the projection position of the mouse in the three-dimensional scene. Then, the system performs intersection detection (intersectObjects) between this ray and the bounding boxes or meshes of all goods in the scene, obtains the list of goods that the ray passes through, and takes the closest hit goods to the camera as the detection result.

[0081] When a target item is detected, the system performs the following actions: First, it highlights the target item by enhancing its material's emissive property. This is achieved by setting the emissive property of the target item's material to a highlight color (e.g., bright yellow), making the item stand out prominently in the scene. Second, a tooltip pops up, displaying detailed information about the target item. This tooltip follows the mouse cursor and includes at least: SKU (Stock Keeping Unit), stack number (e.g., P1-2-3, indicating the 1st stack, 2nd layer, 3rd item), layer number, dimensions (length × width × height or diameter × height), and shape type (carton / medium drum / large drum, etc.). The tooltip follows the mouse cursor to avoid obscuring other items. When the mouse moves away from the target item, the system cancels the highlight and automatically hides the tooltip.

[0082] Through light projection and levitation interaction, users can identify any item in the box one by one and quickly view detailed information such as SKU, stack number, layer, and size of the goods, meeting the needs of refined verification of packing plans, such as checking whether a specified SKU of goods is loaded in the correct position during the shipment inspection process.

[0083] It also includes a camera adaptation submodule and a multi-container switching submodule. The camera adaptation submodule automatically calculates the optimal viewing distance and target point based on the container size and the current view mode, which includes a merged view mode and a layered view mode. The multi-container switching submodule matches the corresponding stacking data in the 3D data according to the container identifier and triggers re-rendering when switching containers in the container list, while simultaneously driving the camera adaptation submodule to adjust the camera to the optimal viewing angle. Camera adaptation ensures that users have a complete field of view regardless of container size or whether they are in a merged or layered view; multi-container switching enables seamless browsing between multiple containers.

[0084] The camera adaptation submodule automatically calculates the optimal viewing distance and target point based on the container size and the current view mode. The system obtains the length, width, and height dimensions of the current container, as well as the current view mode (merged view or layered view), and calculates the scene's bounding sphere radius R (the minimum bounding sphere radius of all objects in the scene). In the merged view, the viewing distance D is calculated based on the container's diagonal length, i.e., D = k1 × L_diag, where k1 is the distance coefficient (e.g., 2.0), and L_diag is the container's diagonal length. In the layered view, because the layers are separated vertically, the scene height increases, and the viewing distance D = k2 × (L_diag + H_expand), where k2 is the distance coefficient, and H_expand is the total offset of the layered expansion. The target point (camera alignment point) is set to the center of the scene's bounding sphere, i.e., the geometric center of the container (or the geometric center of the expanded scene in the layered view). The camera position is calculated using spherical coordinates (distance D, elevation angle θ, azimuth angle φ). The default elevation and azimuth angles provide a 45° overhead view. Users can freely rotate and zoom the camera view using the OrbitControls interactive control.

[0085] The multi-container switching submodule is used to match the corresponding stacking data in 3D data based on the container identifier (container_id) when switching containers in the container list, and triggers the regenerate function to reconstruct the 3D scene. The switching process is as follows: when the user clicks on the target container in the container list, the system obtains the container_id of the target container, matches the corresponding stacking data (container size, cargo coordinates, rotation, shape, layer, etc.) in the 3D data, triggers the regenerate function to reconstruct the 3D scene of the target container, and simultaneously drives the camera adaptation submodule to recalculate the optimal viewing distance and target point based on the size of the target container and the current view mode, so that the camera automatically adjusts to the optimal viewing angle. The entire switching process does not require closing and reopening the panel, and the operation is smooth and seamless.

[0086] Example 2: System Operation Flow like Figure 2 As shown in the figure, this embodiment details the complete operation process of the intelligent container sorting system for the international logistics lubricant industry. This process forms a complete business link from viewing container details to process closure.

[0087] Step S1: The user enters the container sorting details page. The user accesses the container sorting system through a browser and enters the container sorting details page of the specified execution order. This page displays the basic information of the execution order and the operation entry points.

[0088] Step S2: View the execution order information and product details. The system displays the product details of the execution order, including the SKU, name, specifications (size), weight, batch number, quantity, shape type, etc., allowing users to have a comprehensive understanding of the products to be sorted. The product detail data comes from the backend system and is provided through the data / interface layer.

[0089] Step S3: Select the packing method. The system provides two packing methods for the user to choose from: automatic packing and manual packing. If the user selects automatic packing, proceed to step S4a; if the user selects manual packing, proceed to step S4b.

[0090] Step S4a: Automatic container allocation. The system calls the autoPack interface, passing the execution single ID to the server. The server automatically calculates the container allocation scheme based on parameters such as cargo specifications, weight, batch, and quantity, combined with container size constraints and loading constraints such as heavier items on the bottom and lighter items on top, larger items on the bottom and smaller items on top, and maximum load capacity. It then returns the container allocation result and the internal coordinates of each cargo. After automatic container allocation is completed, the system automatically proceeds to step S5.

[0091] Step S4b, Manual Packing. The user enters the manual packing interface, which provides a two-way interface for both the list of goods to be packed and the list of already packed goods. The user selects the goods to be packed and enters the quantity to be moved. The system uses the BigNumber high-precision calculation library to calculate the total weight of the selected goods in real time and compares it with the container's weight limit. If the weight exceeds the limit, a secondary confirmation prompt will pop up. After the user completes the goods moving operation, the system proceeds to step S5. Manual packing also supports template import for batch packing and one-click reset, allowing users to return and readjust at any time.

[0092] Step S5: Packing complete, confirm packing. After the user confirms the packing plan is correct, click "Confirm Packing," and the system calls the finishPack interface to submit the final packing plan. If there is an overweight situation during the packing process, the system will perform a load-bearing verification again before confirming packing to ensure that the weight of the submitted plan is compliant.

[0093] Step S6, 3D Visualization View (Core Function). After confirming the packing, the system automatically enters the 3D visualization view interface. This interface uses WebGL (Three.js engine) to render the 3D scene of the container and cargo in real time on the browser. Specifically, it includes: container wireframe modeling, presenting the container's boundaries in wireframe mode; multi-form cargo rendering, rendering each cargo as geometric shapes such as cartons, medium drums, and large drums; layered expansion / layer-by-layer filtering, allowing users to view the cargo arrangement of each layer by expanding the layers, or to display cargo of a specific layer by filtering by layer; and light projection hover interaction, where hovering the mouse over any cargo will highlight the cargo and pop up a floating window containing information such as SKU, stack number, layer, size, and shape type.

[0094] Step S7, Subsequent Operations (Closed-Loop Process). After completing the 3D visualization, users can perform one or more of the following subsequent operations as needed: Container editing, including container type editing and loading method settings; Fumigation arrangement, including submitting fumigation requests and stamping / certification markings; WMS synchronization, including pushing sub-container data to the WMS or canceling synchronization; TSDO management, including PDF preview and Excel export of TSDO documents; Shipment inspection, including confirmation of contract, payment, business, EUC, and other inspection items; Gift management, including adding and removing gifts; Special requirements viewing, including viewing the special requirements of the execution order and its approval records.

[0095] Step S8, process ends. After the above operations are completed, the entire container sorting business chain is closed, and the system records the operation logs of each step for subsequent traceability.

[0096] In this embodiment, the 3D visualization viewing in step S6 is the core function of the entire process. It is seamlessly connected with the binning calculation in steps S4a / S4b through a direct data drive mechanism: the result data generated by automatic or manual binning directly drives the 3D scene rendering. Users do not need to perform data export, conversion, and re-import operations, realizing an integrated process from binning calculation to 3D presentation.

[0097] Example 3: System Operating Environment and Computer-Readable Storage Medium This embodiment provides the specific operating environment implementation of the system. The system adopts a front-end and back-end separation architecture: the front-end is a browser-based web application, using the Vue.js framework to build the user interface and Three.js to build the 3D rendering scene; the back-end is a server-side application, providing interface services such as autoPack, getExec3D, getContainers, snyWms, and finishPack. The back-end can be implemented using mainstream server-side frameworks such as Java Spring Boot and Node.js, and the database can use relational databases such as MySQL and PostgreSQL to store business data such as goods, containers, and execution orders. The front-end and back-end interact with each other via HTTP / HTTPS protocol, and the data format is JSON.

[0098] In terms of hardware environment, the system runs on servers and user terminal devices: the server is responsible for bin-based computing and data storage, while the user terminal (desktop, laptop, or other devices with a browser) is responsible for rendering and interacting with the 3D scene. Since WebGL rendering is completed on the user terminal, the user terminal needs to support WebGL browsers (such as Chrome, Firefox, and Edge) and WebGL-enabled graphics card drivers.

[0099] This embodiment also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the functions of the intelligent bin-sorting system described in any of the above embodiments. The computer-readable storage medium includes, but is not limited to, media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0100] Example 4: Typical Application of the System in International Logistics Scenarios in the Lubricating Oil Industry This embodiment uses a batch of international logistics orders for lubricating oil as an example to illustrate the typical application process of the present invention. Assume a lubricating oil exporting company has an execution order containing 30 large 200L drums of lubricating oil (large drums, each weighing 180kg, with dimensions of 600mm in diameter and 900mm in height), 200 cartons of 4L small drums of lubricating oil (cartons, each weighing 18kg, with dimensions of 600mm × 400mm × 300mm), and 10 palletized goods (including pallets, each pallet weighing 800kg, with dimensions of 1200mm × 1000mm × 1000mm), planned to be loaded into a 20-foot container (internal dimensions approximately 5890mm × 2350mm × 2390mm, with a maximum load capacity of approximately 28,000kg).

[0101] After viewing the execution order information and cargo details on the container allocation details page, the user selects the automatic container allocation method. The system calls the autoPack interface. Based on the cargo specifications, weight, batch, quantity, and container size constraints, combined with the constraints of heavier items on the bottom and lighter items on top, larger items on the bottom and smaller items on top, and the maximum load capacity, the server automatically calculates the container allocation plan: 10 pallets of full-size cargo and 30 200L drums are placed at the bottom of the container (layer 1). 200 4L cartons are stacked in multiple layers on the top layers (layers 2 and 3). The total weight (10×800+30×180+200×18 = 8000+5400+3600 = 17000kg) is verified to be within the maximum load capacity of 28000kg, indicating successful container allocation. The system returns the container allocation result and internal coordinates for each cargo.

[0102] After confirmation of loading, the system enters a 3D visualization viewing interface: based on Three.js, it renders the container wireframe, bottom drums, pallet model, and top carton model in the browser. Users can use the layer expansion function to separate the 1st, 2nd, and 3rd layers vertically, using blue, green, and orange wireframes to distinguish layer boundaries, allowing for a clear view of the layer arrangement. Users can also use the layer filtering function to display only the 1st layer, verifying the placement of the bottom drums and pallets. Hovering the mouse over any item triggers a light projection detection, highlighting the item and displaying a pop-up information window showing the item's SKU, stack number, layer, size, and shape. When switching from the container list to another 40-foot container, the system automatically refreshes the 3D scene and adjusts the camera to the optimal viewing angle. The entire process requires no data export and import, achieving a seamless transition from container segmentation calculations to 3D rendering.

[0103] The embodiments described herein are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An intelligent sorting system for the international logistics lubricant industry, characterized in that, include: The cargo attribute recognition module is used to acquire cargo attribute data of the cargo to be sorted and identify the cargo form. The cargo attribute data includes at least the cargo shape type field, specifications and dimensions, weight, batch and quantity. The cargo attribute recognition module divides the cargo into three forms: cardboard box, medium drum and large drum according to the cargo shape type field, and performs packing calculations in combination with the weight of the goods and the packaging size, and configures corresponding geometric modeling parameters and rendering attributes for each form. The loading constraint engine is used to implement built-in loading constraint rules such as heavy loads on the bottom and light loads on the top, and large loads on the bottom and small loads on the top. It prioritizes heavy barrels and palletized goods with a weight greater than the preset weight threshold to be placed at the bottom of the container, while light and small cartons are placed on the top. During the sorting process, it checks in real time whether the total weight of the goods exceeds the container's load-bearing limit and triggers an overweight warning if it does. The automatic container sorting module is used to send the execution order identifier of the goods to be sorted to the server. The server automatically calculates the sorting scheme based on the specifications, weight, batch, quantity and container size constraints of the goods, and returns the container allocation result, internal coordinate position, rotation angle, layer information and shape type of each goods. The web-based 3D rendering module is used to construct a 3D scene of containers and cargo on the browser side based on the WebGL graphics rendering engine. It directly drives the 3D scene rendering based on the container segmentation results returned by the automatic container segmentation module, generating 3D models of various cargo types. The web-based 3D rendering module includes a layered expansion submodule, a layer-by-layer filtering submodule, and a floating interaction submodule, wherein: The layered unfolding submodule is used to calculate the interlayer offset based on the height of each layer of goods in the container, and to separate and display each layer of goods in the vertical direction step by step, with different colored wireframes to distinguish the boundaries of each layer. The layer-by-layer filtering submodule is used to maintain layer-by-layer filtering variables. In the rendering loop, it determines whether the layer number of each item matches the layer-by-layer filtering variables. Items that do not match are skipped from rendering so that items of the specified layer can be displayed separately. The floating interaction submodule is used to listen for mouse movement events, detect goods below the mouse position through a ray casting algorithm, highlight the hit goods and pop up a floating information window when the hit is detected. The floating information window displays at least the SKU, stack number, level, size and shape type of the goods.

2. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, In the cargo attribute recognition module, the geometric modeling parameters for the three types of shapes—carton, medium drum, and large drum—include length, width, and height dimensions or bottom diameter and height, and their respective rendering attributes include color, texture, and transparency. The system also includes a pallet modeling subunit, a pad modeling subunit, a partition modeling subunit, and a strapping rendering subunit, which are used to render the models of the pallet, pad, partition, and strapping in the three-dimensional scene respectively to restore the real loading form of the lubricating oil cargo.

3. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, The loading constraint engine is also used to compare the total weight of the goods in the current packing scheme with the container's maximum load capacity in real time during the packing process. When the total weight of the goods exceeds the container's maximum load capacity, a secondary confirmation prompt will pop up. The total weight of the goods is obtained in real time by a high-precision calculation library, where the weight of a single item is the product of the weight of the goods unit and the quantity of goods.

4. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, The automatic container splitting module completes the container splitting calculation by calling the automatic container splitting interface. The automatic container splitting interface receives the execution order identifier. The server calculates the container splitting scheme based on the cargo specifications, weight, batch, quantity, container size constraints, and loading constraints, and returns the container allocation result and internal coordinate position of each cargo through the automatic container splitting interface.

5. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, It also includes a manual box splitting auxiliary module, which provides a two-way operation interface for the list of goods to be split and the list of goods already split. After the user selects the goods to be split and fills in the quantity to be moved in, the system uses a high-precision calculation library to accumulate the total weight of the selected goods in real time and compares the accumulation result with the upper limit of the container's load capacity. If the weight exceeds the limit, a second confirmation pop-up will appear. The manual bin sorting auxiliary module also supports batch bin sorting via template import and one-click reset of bin sorting results.

6. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, In the layered unfolding submodule, the interlayer offset is calculated based on the maximum height of each layer of goods. After each layer of goods is separated vertically, the boundaries between layers are distinguished by blue, green, and orange lines to visually present the placement range of each layer of goods.

7. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, It also includes a compact stacking submodule, which sorts the goods in the 3D scene according to the three-level sorting rules of hierarchy, shape weight, and index in the merged view, and calculates the compact reference plane of each item by item through footprint overlap detection, so that the goods in adjacent layers fit tightly in the vertical direction and eliminate the gaps between layers.

8. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, In the floating interaction submodule, highlighting the hit goods is achieved by enhancing the self-illuminating properties of the hit goods' materials; the floating information window is a tooltip that follows the mouse movement and automatically hides when the mouse moves away from the hit goods.

9. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, It also includes a camera adaptive submodule and a multi-cabinet switching submodule; The camera adaptive submodule is used to automatically calculate the optimal observation distance and observation target point based on the container size and the current view mode, which includes a merged view mode and a layered view mode. The multi-container switching submodule is used to match the corresponding stacking data in the 3D data according to the container identifier and trigger re-rendering when switching containers in the container list. At the same time, it drives the camera adaptive submodule to adjust the camera to the best viewing angle.

10. The intelligent sorting system for the international logistics lubricant industry according to claim 1, characterized in that, It also includes a data direct-drive integration module, which is used to obtain container size, stack attributes, and cargo coordinates, rotation angle, shape type and layer data through a 3D data interface, and directly drive the Web-based 3D rendering module to perform scene rendering without intermediate data conversion; The data direct drive integration module also monitors data changes through a reactive listening mechanism. When data changes, it automatically performs data loading and scene reconstruction, and releases the geometry and material resources of the old model before scene reconstruction to avoid memory leaks.

Citation Information

Patent Citations

  • Article boxing method and device and computer readable storage medium

    CN109840726A

  • A three-dimensional packing method for cargo in modular containers

    CN112607234B

  • Material transfer visual intelligent system and building method

    CN114067048A

  • Load planning optimization using automated 3D packing

    US11820611B2

  • Augmented reality visualization embarkation method and system

    US20240338911A1