Method, device and equipment for determining logistics cost

Abstract

This invention discloses a method, apparatus, and equipment for determining logistics costs. The method, applied in the field of logistics technology, includes: constructing a unified data platform; collecting multi-source heterogeneous data from a logistics platform across the entire supply chain; intelligently fusing the data to obtain multi-dimensional data characterizing a parcel cost calculation strategy; predicting the physical dimensions, weight, and visual features of express parcels based on 3D vision and AI technology; determining the target category of the express parcel based on its physical dimensions, weight, and visual features; transforming unstructured, text-based contract terms and price lists into a structured, relational knowledge graph based on the multi-dimensional data; and determining the parcel cost calculation strategy based on the knowledge graph; and determining the estimated cost of the express parcel based on the parcel cost calculation strategy, the physical dimensions, weight, visual features, and target category. This invention can accurately predict the cost of express parcels.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, and equipment for determining logistics costs. Background Technology

[0002] In related technologies, different logistics carriers use different volumetric weight conversion factors, and the criteria for judging heavy goods and bulky goods differ, resulting in inconsistent billing results for the same goods at different carriers.

[0003] The lack of deep integration between Transportation Management System (TMS), Warehouse Management System (WMS), and ERP system creates a data silo effect, and the distortion of key information leads to a 5-15% deviation in freight cost calculation.

[0004] Existing system optimization algorithms overemphasize explicit transportation costs while neglecting implicit costs such as inventory turnover and order fulfillment timeliness, potentially trapping businesses in a vicious cycle of "the more you optimize, the more you lose." Frequent changes in carrier rates (monthly or even weekly) require traditional systems to undergo extensive manual configuration or code modifications to adapt to these changes, and there is a lack of fast and efficient maintenance tools.

[0005] Discrepancies exist between the automated settlement system of online freight platforms and the actual transportation execution. Issues such as errors in basic freight calculation, omissions in surcharges, and distorted mileage data result in significant annual transportation cost losses for enterprises. Traditional billing methods, which rely heavily on post-transport settlement, cannot monitor cost changes in real time during transportation, making cost prediction and optimization even more difficult. Summary of the Invention

[0006] This invention provides a method, apparatus, and equipment for determining logistics costs, which can accurately predict the cost of express parcels.

[0007] On one hand, the present invention provides a method for determining logistics costs, the method comprising: A unified data platform is built to collect multi-source heterogeneous data from the logistics platform across the entire chain, and the data is intelligently integrated to obtain multi-dimensional data that characterizes the parcel cost calculation strategy. Based on 3D vision and AI technology, predict the physical size, weight and visual features of express parcels; The target category of the express parcel is determined based on its physical size, weight, and visual characteristics. Based on the multi-dimensional data, unstructured, text-based contract terms and price lists are transformed into structured, interconnected knowledge graphs, and a package fee calculation strategy is determined based on the knowledge graphs. The estimated cost of the express package is determined based on the package cost calculation strategy, the physical dimensions, weight, visual characteristics, and target category of the express package.

[0008] On the other hand, a logistics cost determination device is provided, the device comprising: The multi-dimensional data acquisition module is used to build a unified data platform, collect multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently fuse the data to obtain multi-dimensional data that characterizes the package cost calculation strategy. The feature prediction module is used to predict the physical size, weight, and visual features of express parcels based on 3D vision and AI technology. The category determination module is used to determine the target category of the express parcel based on its physical size, weight, and visual characteristics. The strategy determination module is used to transform unstructured, text-based contract terms and price lists into a structured, related knowledge graph based on the multi-dimensional data, and to determine the package fee calculation strategy based on the knowledge graph. The cost determination module is used to determine the estimated cost of the express package based on the package cost calculation strategy, the physical size, weight, visual characteristics, and target category of the express package.

[0009] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: By acquiring heterogeneous data sources from express parcels through diverse data interfaces, multi-source heterogeneous data can be obtained. The multi-source heterogeneous data is cleaned, verified, and augmented to obtain processed data. The processed data is subjected to entity parsing, and information from different sources describing the same entity is intelligently fused to obtain multi-dimensional data representing the package cost calculation strategy.

[0010] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: The processed data is parsed to determine whether an entity has a primary key; the primary key is either the waybill number or the order number. If the primary key exists, the processed data is associated according to the primary key, and the processed data corresponding to the same primary key are merged to obtain the first merged data; If the primary key does not exist, start the fuzzy matching algorithm to determine the combination of multiple fields in each entity; and calculate the similarity score between any two entities based on the field combinations corresponding to each of the two entities. Multiple entities with similarity scores greater than a preset threshold are identified as associated entities; and the processed data corresponding to each of the associated entities are fused to obtain second fused data. Based on the first fused data and the second fused data, multi-dimensional data characterizing the package cost calculation strategy are determined. The device further includes a graph display module, used to construct and display a data association graph based on the multi-dimensional data.

[0011] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: Based on the first fused data and the second fused data, create a waybill entity object corresponding to each associated entity; Based on the timeline, determine the order information, warehousing, sorting, trunk transportation, delivery and receipt information corresponding to each waybill entity, and obtain the full process events of each waybill entity. Based on the cost elements of each event in the entire process, multi-dimensional data representing the package cost calculation strategy are determined, and a logistics digital twin is constructed.

[0012] In one exemplary embodiment, the feature prediction module is further configured to: A multimodal sensor array was constructed based on 3D vision and AI technologies; The weight, visual features, and 3D point cloud data of the express parcel are acquired based on the multimodal sensor array. The three-dimensional point cloud data is filtered and denoised to obtain processed point cloud data; Based on the processed point cloud data, the physical dimensions of the express package are determined.

[0013] In one exemplary embodiment, the feature prediction module is further configured to: Based on the processed point cloud data, the number of express parcels is determined. If there are multiple express parcels, the processed point cloud data is segmented into instances to obtain the point cloud data corresponding to each express parcel, and an instance ID is assigned to the point cloud data corresponding to each express parcel. For each express parcel, calculate the minimum bounding cube of the point cloud data; Principal component analysis is performed on the minimum bounding cube to determine the main direction of the point cloud data, and a bounding box is constructed along the main direction. The length, width and height of the bounding box are then output to obtain the physical dimensions of the express parcel corresponding to the point cloud data.

[0014] In one exemplary embodiment, the category determination module is further configured to: Based on the physical dimensions of the express package, determine the volume, aspect ratio, and surface area-to-volume ratio of the express package; The actual density of the express package is determined based on the ratio of its weight to its volume. The visual features of the express package are input into a convolutional neural network to extract texture features, shape regularity, and material appearance features. Obtain the origin information and current time of the express package as contextual features; The actual density, aspect ratio, surface area to volume ratio, texture features, shape regularity, material appearance features, and contextual features of the express package are input into the density classification model, and the target category of the express package is output; the target category represents the density category of the express package.

[0015] On the other hand, an electronic device is provided, the device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the logistics cost determination method as described above.

[0016] On the other hand, a computer storage medium is provided that stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the logistics cost determination method as described above.

[0017] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the logistics cost determination method as described above.

[0018] The logistics cost determination method, apparatus, and equipment provided by this invention have the following technical advantages: This invention constructs a unified data platform, collecting multi-source heterogeneous data from the entire logistics platform and intelligently fusing the data to obtain multi-dimensional data characterizing parcel cost calculation strategies. Based on 3D vision and AI technology, it predicts the physical dimensions, weight, and visual features of express parcels. Based on these physical dimensions, weight, and visual features, it determines the target category of the express parcel. Based on the multi-dimensional data, it transforms unstructured, text-based contract terms and price lists into a structured, relational knowledge graph, and determines the parcel cost calculation strategy based on the knowledge graph. Finally, based on the parcel cost calculation strategy, the physical dimensions, weight, visual features, and target category of the express parcel, it determines the estimated cost of the express parcel. This invention, by constructing a unified data platform, integrates multi-source heterogeneous data from the entire logistics chain, breaking down information silos and achieving end-to-end supply chain data connectivity. Based on computer vision and deep learning technologies, it automatically identifies the 3D dimensions and density features of goods, achieving accurate classification of heavy and bulky goods, and can accurately predict the cost of express parcels. Attached Figure Description

[0019] To more clearly illustrate the technical solutions and advantages in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a logistics cost determination system provided in the embodiments of this specification; Figure 2 This is a flowchart illustrating a method for determining logistics costs provided in the embodiments of this specification; Figure 3 This is a flowchart illustrating a method for determining multi-dimensional data characterizing a package cost calculation strategy based on the first fused data and the second fused data, as provided in an embodiment of this specification. Figure 4 This is a flowchart illustrating a method for determining the physical dimensions of an express parcel based on the processed point cloud data provided in an embodiment of this specification. Figure 5 This is a flowchart illustrating a method for determining the target category of a courier package based on its physical dimensions, weight, and visual characteristics, as provided in an embodiment of this specification. Figure 6 This is a schematic diagram of the structure of a logistics cost determination device provided in the embodiments of this specification; Figure 7 This is a schematic diagram of the structure of a server provided in the embodiments of this specification. Detailed Implementation

[0021] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0023] Please see Figure 1 , Figure 1 This is a schematic diagram of a logistics cost determination system provided in the embodiments of this specification, such as... Figure 1 As shown, the logistics cost determination system may include at least server 01 and client 02.

[0024] Specifically, in the embodiments of this specification, server 01 may include a standalone server, a distributed server, or a server cluster composed of multiple servers. It may also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Server 01 may include network communication units, processors, and memory, etc. Specifically, server 01 can be used to determine the target category of the express parcel and determine the estimated cost of the express parcel.

[0025] Specifically, in this embodiment of the specification, the client 02 may include physical devices such as smartphones, desktop computers, tablets, laptops, digital assistants, smart wearable devices, smart speakers, in-vehicle terminals, and smart TVs. It may also include software running on the physical device, such as web pages provided to users by service providers, or applications provided to users by such service providers. Specifically, the client 02 can be used to display the target category of the express parcel and the estimated cost of the express parcel.

[0026] The following describes a method for determining logistics costs according to the present invention. Figure 2This is a flowchart illustrating a method for determining logistics costs provided in an embodiment of this specification. This specification provides the operational steps of the method described in the embodiment or flowchart, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiment is merely one possible execution order among many and does not represent the only possible execution order. In actual system or server product execution, the method can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment) as shown in the embodiment or drawings. Specifically, as... Figure 2 As shown, the method may include: S201: Construct a unified data platform to collect multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently integrate the data to obtain multi-dimensional data representing the parcel cost calculation strategy. In the embodiments described in this specification, the goal of this step is to construct a unified, reliable, and efficient data hub that connects to and deeply cleans and integrates dispersed, heterogeneous, and dynamically changing data sources throughout the entire logistics chain via standardized pipelines. This is the cornerstone of the entire intelligent accounting system, and its quality directly determines the accuracy of all subsequent analyses.

[0027] S203: Based on 3D vision and AI technology, predict the physical size, weight and visual characteristics of express parcels.

[0028] In the embodiments described in this specification, a high-precision, automated method is used to replace the traditional, inefficient, and error-prone manual measurement, accurately capturing the physical dimensions and visual characteristics of the goods, and providing an irrefutable data foundation for subsequent refined billing and loading optimization.

[0029] S205: Determine the target category of the express parcel based on its physical size, weight, and visual characteristics.

[0030] In the embodiments of this specification, the target category of the express parcel can be determined based on its physical dimensions, weight, and visual characteristics. The target category represents the density category of the express parcel.

[0031] S207: Based on the aforementioned multi-dimensional data, the unstructured, text-based contract terms and price lists are transformed into a structured, interconnected knowledge graph, and a package fee calculation strategy is determined based on the knowledge graph.

[0032] In the embodiments described in this specification, complex, variable, and implicit logistics billing rules can be transformed into a digital rule system that is understandable, executable, and manageable by computers, and an intelligent engine capable of applying these rules quickly, accurately, and flexibly can be built. Knowledge graph-based digital modeling of billing rules transforms unstructured, text-based contract terms and price lists into structured, interconnected knowledge graphs.

[0033] S209: Determine the estimated cost of the express package based on the package cost calculation strategy, the physical dimensions, weight, visual characteristics, and target category of the express package.

[0034] As can be seen from the technical solutions provided in the embodiments of this specification above, the present invention constructs a unified data platform, collects multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently fuses the data to obtain multi-dimensional data characterizing the parcel cost calculation strategy; based on 3D vision and AI technology, it predicts the physical size, weight, and visual characteristics of express parcels; based on the physical size, weight, and visual characteristics of the express parcels, it determines the target category of the express parcels; based on the multi-dimensional data, it transforms unstructured, text-based contract terms and price lists into structured, related knowledge graphs, and determines the parcel cost calculation strategy according to the knowledge graphs; based on the parcel cost calculation strategy, the physical size, weight, visual characteristics, and target category of the express parcels, it determines the estimated cost of the express parcels. This invention, by constructing a unified data platform, integrates multi-source heterogeneous data from the entire logistics chain, breaks down information silos, and achieves end-to-end supply chain data connectivity; based on computer vision and deep learning technology, it automatically identifies the 3D size and density characteristics of goods, achieving accurate classification of heavy and bulky goods; and it can accurately predict the cost of express parcels.

[0035] In some embodiments, the multi-source heterogeneous data collected from the end-to-end logistics platform is intelligently fused to obtain multi-dimensional data characterizing the package cost calculation strategy, including: By acquiring heterogeneous data sources from express parcels through diverse data interfaces, multi-source heterogeneous data can be obtained. The multi-source heterogeneous data is cleaned, verified, and augmented to obtain processed data. The processed data is subjected to entity parsing, and information from different sources describing the same entity is intelligently fused to obtain multi-dimensional data representing the package cost calculation strategy.

[0036] In the embodiments described in this specification, the raw data from multiple interfaces generally suffers from missing, abnormal, inconsistent, and distorted issues. Therefore, the multi-source heterogeneous data can be cleaned, verified, and augmented to obtain processed data, which is then used for data analysis.

[0037] In some embodiments, the entity parsing of the processed data involves intelligently fusing information from different sources describing the same entity to obtain multi-dimensional data characterizing the package fee calculation strategy, including: The processed data is parsed to determine whether an entity has a primary key; the primary key is either the waybill number or the order number. If the primary key exists, the processed data is associated according to the primary key, and the processed data corresponding to the same primary key are merged to obtain the first merged data; If the primary key does not exist, start the fuzzy matching algorithm to determine the combination of multiple fields in each entity; and calculate the similarity score between any two entities based on the field combinations corresponding to each of the two entities. Multiple entities with similarity scores greater than a preset threshold are identified as associated entities; and the processed data corresponding to each of the associated entities are fused to obtain second fused data. Based on the first fused data and the second fused data, multi-dimensional data characterizing the package cost calculation strategy are determined. The method further includes: constructing and displaying a data association map based on the multi-dimensional data.

[0038] This embodiment first attempts to associate data using precise primary keys (such as unique waybill numbers or order numbers). When the primary key is missing or inconsistent, a fuzzy matching algorithm is activated. The algorithm comprehensively compares combinations of multiple fields, such as "recipient's name + last four digits of recipient's phone number + destination city + product category," and calculates a comprehensive similarity score. Records exceeding a threshold will be associated as the same entity.

[0039] In some embodiments, such as Figure 3 As shown, the multi-dimensional data characterizing the package fee calculation strategy, determined based on the first fused data and the second fused data, includes: S301: Based on the first fused data and the second fused data, create a waybill entity object corresponding to each associated entity; S303: Based on the timeline, determine the order information, warehousing, sorting, trunk transportation, delivery and receipt information corresponding to each waybill entity, and obtain the full process events of each waybill entity. S305: Based on the cost elements of each event in the entire process, determine multi-dimensional data representing the package cost calculation strategy, and construct a logistics digital twin.

[0040] In this embodiment, each successful association creates a rich "waybill entity" object. This object not only contains basic information, but also connects the entire process of the goods from order placement, warehousing, sorting, trunk transportation, delivery to receipt through a timeline, and integrates the cost elements generated in each link (warehousing fees, handling fees, trunk transportation fees, delivery fees, etc.), thereby constructing a logistics digital twin.

[0041] For example, multi-source data acquisition and intelligent fusion processing specifically include: 1. Ubiquitous access and protocol adaptation of heterogeneous data sources The system is designed with a modular, plug-in-the-loop general-purpose data access gateway to cope with the diverse data interface formats in the logistics ecosystem. This gateway acts like the system's "sensory nerve endings," capable of processing information from different protocols and data formats simultaneously.

[0042] Standardized API Synchronous Access: For modern systems that provide open APIs, such as transportation management systems and mainstream e-commerce platform order systems, the system has built-in clients for various authentication protocols such as OAuth 2.0 and API Key. Through pre-configured request templates and schedulers, timed or event-driven data retrieval and push are achieved. For example, whenever a new waybill is generated in the TMS, an event trigger will synchronize a snapshot of the waybill to this system in real time.

[0043] Intelligent parsing of unstructured documents: For the large number of contracts, quotations, and manual waybills still transmitted via Excel, PDF, or even scanned images, this system integrates a powerful document understanding engine. This engine combines Optical Character Recognition (OCR) technology, Named Entity Recognition (NER) technology from Natural Language Processing (NLP), and table detection algorithms to automatically locate and extract key fields from unstructured documents, such as "destination code," "weight tiered pricing," and "special handling fees," and transform them into structured data. For example, the system can understand text descriptions like "First 1kg: xx yuan, subsequent kilograms: x yuan" and convert them into computable rule objects.

[0044] Real-time ingestion of IoT and streaming data: Streaming data from warehouse camera networks, dynamic weighbridges, handheld terminals, and vehicle GPS are accessed in real time via message queues (such as Apache Kafka). The system is designed with a dedicated stream processing pipeline for this type of high-throughput, low-latency data, capable of preliminary filtering, aggregation, and windowing calculations, such as real-time calculation of cargo throughput or average vehicle loading rate at a sorting center.

[0045] 2. The raw data in the multi-dimensional data cleaning, validation and enhancement pipeline generally suffers from missing, abnormal, inconsistent and distorted problems.

[0046] This step establishes an automated, configurable data cleaning pipeline to ensure that the data flowing into the core computing engine is clean and reliable.

[0047] Rule-driven data validation: The system maintains a scalable base of business rules for performing data quality checks. These rules include: Value range rules: Check the reasonableness of the values. For example, the package length must be greater than 0 and less than the maximum allowable size of the means of transport (such as cargo plane door); the weight must be within the range of the weighing equipment.

[0048] Logical rules: Check the logical consistency between data. For example, the ratio between a product's "volume weight" (length / width / height / coefficient) and its actual weight (i.e., density) should conform to the range of common commodity physical properties. If a product labeled as "foam plastic" has a calculated density as high as 2 kg / L, it will be marked as "high confidence anomaly".

[0049] Association rules: Check cross-field associations. For example, if the shipping destination is "remote mountainous area" but the selected service product is "same-day delivery", this may be an option conflict that needs to be reviewed.

[0050] Intelligent Repair and Missing Value Filling: For identified problematic data, the system does not simply discard it, but attempts intelligent repair. For missing, non-critical numerical fields (such as minor dimensions of a package), the system may fill them in using the mean or median of the same batch or type of goods. For obvious outliers (such as mistakenly recording 15 meters instead of 1.5 meters), the system can make speculative corrections based on historical records and related information (such as typical product dimensions). All repair operations are recorded for auditing purposes.

[0051] Data standardization and formatting: The cleaned data was forcibly converted to a unified internal representation. All length units were converted to millimeters, weight units to grams, currency was standardized to RMB, and dates and times were all converted to UTC timestamps in xxx format. This eliminated fatal errors caused by unit confusion in subsequent calculations.

[0052] 3. Multi-source data fusion and association based on entity parsing The core of this step is to piece together fragmented information from different sources that describe the same entity (such as a shipment of goods) into a complete panoramic view, which is the key to achieving "end-to-end" cost analysis.

[0053] Primary Key Association and Fuzzy Matching: The system first attempts to associate data using precise primary keys (such as unique waybill numbers or order numbers). When a primary key is missing or inconsistent, a fuzzy matching algorithm is activated. The algorithm comprehensively compares combinations of multiple fields, such as "recipient's name + last four digits of recipient's phone number + destination city + product category," and calculates a comprehensive similarity score. Records exceeding a threshold will be associated as the same entity.

[0054] Building a "logistics digital twin": Each successful connection creates a rich "waybill entity" object. This object not only contains basic information but also connects the entire process of the goods from order placement, warehousing, sorting, trunk transportation, delivery to receipt via a timeline, and integrates the cost elements generated at each stage (warehousing fees, handling fees, trunk transportation fees, delivery fees, etc.). This digital twin is the foundation for in-depth cost attribution and optimization analysis.

[0055] Visualization of relationships: The system provides graphical tools to display the complete data relationship map of a specific product or order, helping managers to intuitively understand data flow and status changes, and quickly locate data breakpoints or abnormal relationships.

[0056] 4. Continuous Data Quality Monitoring and Governance: Data quality is not a one-time fix. This system has established a continuous data quality monitoring system.

[0057] Define and measure data quality dimensions: The system continuously monitors the core dimensions of data, including completeness (missing rate of key fields), accuracy (difference rate with authoritative sources or manual sampling), consistency (difference of the same indicator in different reports), and timeliness (delay from data generation to availability).

[0058] Dashboard and Proactive Alerts: All quality metrics are presented in real-time as charts on the management dashboard. The system allows setting alert and alarm thresholds for each metric. For example, if the daily data delay from a certain data source exceeds 1 hour, or the proportion of abnormal data exceeds 5% for 3 consecutive hours, the system will automatically send an alarm notification to the data governance team and may trigger a switch to a backup data source.

[0059] Closed-loop feedback mechanism: Data quality issues are recorded as "governance work orders" and assigned to responsible personnel. The processing results of the work orders (such as confirming that the source system is wrong, correcting parsing rules, etc.) are fed back to the system to optimize cleaning rules or adjust access configurations, forming a closed loop of continuous improvement.

[0060] In some embodiments, the prediction of the physical dimensions, weight, and visual features of express parcels based on 3D vision and AI technology includes: A multimodal sensor array was constructed based on 3D vision and AI technologies; The weight, visual features, and 3D point cloud data of the express parcel are acquired based on the multimodal sensor array. The three-dimensional point cloud data is filtered and denoised to obtain processed point cloud data; Based on the processed point cloud data, the physical dimensions of the express package are determined.

[0061] In the embodiments described in this specification, the collaborative deployment and calibration of multimodal sensor arrays, in order to achieve stable and reliable automated measurement, requires the deployment of integrated sensing units at key operational nodes (such as inbound and outbound sorting lines). Each sensing unit is a hardware and software integrated "intelligent measurement station." It typically includes: 2-3 high-resolution industrial area array cameras (for acquiring high-definition color texture information), 1 3D depth camera (such as a structured light or time-of-flight (ToF) camera for acquiring depth information), a high-precision dynamic weighing platform (for real-time weighing), and a set of controllable supplementary lighting sources. All these sensors are coordinated and controlled by an industrial-grade edge computing control computer to ensure the synchronization of data acquisition.

[0062] System calibration and coordinate system 1: This is a prerequisite for ensuring measurement accuracy. Using a high-precision calibration board, multiple cameras are jointly calibrated to determine their positional and orientation relationships, and the coordinate systems of all sensors (including 2D cameras, 3D cameras, and the weighing platform) are unified to the same world coordinate system. The calibration process needs to be performed periodically (e.g., monthly) or automatically after equipment movement to ensure long-term accuracy.

[0063] In some embodiments, such as Figure 4 As shown, determining the physical dimensions of the express package based on the processed point cloud data includes: S401: Based on the processed point cloud data, determine the number of express parcels. If there are multiple express parcels, perform instance segmentation on the processed point cloud data to obtain the point cloud data corresponding to each express parcel, and assign an instance ID to the point cloud data corresponding to each express parcel. S403: For the point cloud data corresponding to each express parcel, calculate the minimum bounding cube of the point cloud data; S405: Perform principal component analysis on the minimum bounding cube to determine the main direction of the point cloud data, construct a bounding box along the main direction, and output the length, width, and height of the bounding box to obtain the physical dimensions of the express parcel corresponding to the point cloud data.

[0064] In the embodiments of this specification, high-precision dimensional measurement based on deep learning and 3D reconstruction is the core technology of this step, which uses computer vision algorithms to calculate the precise outer dimensions of the goods from the raw sensor data.

[0065] Acquisition and preprocessing of 3D point cloud data: The 3D depth camera directly outputs sparse or dense point cloud data. The system first filters the point cloud to remove noise points (such as ground reflections and distant backgrounds), and applies voxel downsampling to reduce the amount of data while preserving features.

[0066] Cargo Instance Segmentation and Point Cloud Clustering: In complex scenarios containing multiple goods (such as multiple packages on a conveyor belt), segmentation is necessary. The system utilizes a deep learning instance segmentation model (such as a variant based on PointNet++) to process the data directly on the point cloud or after projecting the point cloud onto a 2D image, assigning a unique instance ID to each individual cargo point cloud. This solves the separation challenge in cases where packages are close together or stacked.

[0067] Minimum Bounding Cube Calculation and Size Output: For a segmented point cloud of a single cargo item, calculate its minimum bounding cube in 3D space. This is typically done by using Principal Component Analysis (PCA) to find the principal orientation of the point cloud and then constructing a bounding box along that orientation. The final output is the length (L), width (W), and height (H) of this bounding box. The algorithm can handle regular boxes and soft, irregular packages (such as mailbags).

[0068] In some embodiments, such as Figure 5 As shown, determining the target category of the express parcel based on its physical dimensions, weight, and visual characteristics includes: S501: Based on the physical dimensions of the express parcel, determine the volume, aspect ratio, and surface area-to-volume ratio of the express parcel; S503: Determine the actual density of the express package based on the ratio of its weight to its volume; S505: Input the visual features of the express package into a convolutional neural network to extract texture features, shape regularity, and material appearance features; S507: Obtain the source information and current time of the express package as contextual features; S509: Input the actual density, aspect ratio, surface area to volume ratio, texture features, shape regularity, material appearance features, and contextual features of the express package into the density classification model, and output the target category of the express package; the target category represents the density category of the express package.

[0069] In the embodiments of this specification, a density classification model can be pre-trained to obtain accurate physical dimensions and weight, and then combined with visual features to perform more refined classification and value assessment of goods.

[0070] Feature extraction and fusion: The system extracts multiple features to construct a feature vector. Physical characteristics: calculated actual density (weight / volume), aspect ratio, and surface area to volume ratio.

[0071] Visual features: Extract texture features (roughness / smoothness), shape regularity, and material appearance features (whether it is reflective, whether it is transparent) of the cargo surface using a pre-trained convolutional neural network (CNN).

[0072] Contextual features: origin of goods (e.g., goods shipped from an e-commerce warehouse are mostly small items), current measurement time, etc.

[0073] Density-based classification models: The feature vectors described above are input into a lightweight gradient booster (such as LightGBM) or a small neural network classifier. The model is trained to output a classification probability, for example: Heavy goods: The actual density is much greater than the volumetric density, and the charge should be based on the actual weight.

[0074] Bulk goods / lightweight goods: The actual density is much less than the volumetric density, and should be charged according to volumetric weight.

[0075] Standard goods: The actual density and the volumetric density are close, and the higher of the two is used for billing.

[0076] Formulaic determination: Volumetric weight = (L * W * H) / Volume coefficient (6000 is commonly used for air freight, 5000 or 4000 for land freight). Chargeable weight = MAX(Actual weight, Volumetric weight). The classification model's role is to predict cargo type in advance and intelligently, providing earlier decision-making basis for cost estimation and loading optimization. For example, identifying cargo as "high-volume cargo" can trigger a "recommendation for repackaging" warning.

[0077] In some embodiments, the method further includes: three-dimensional visual measurement and density intelligent recognition, specifically including: 1. Collaborative Deployment and Calibration of Multimodal Sensor Arrays: To achieve stable and reliable automated measurement, it is necessary to deploy integrated sensing units at key operational nodes (such as inbound and outbound sorting lines).

[0078] Hardware Coordinated Configuration: Each sensing unit is a hardware and software integrated "intelligent measurement station." It typically includes: 2-3 high-resolution industrial area scan cameras (for acquiring high-definition color texture information), 1 3D depth camera (such as a structured light or time-of-flight (ToF) camera for acquiring depth information), a high-precision dynamic weighing platform (for real-time weighing), and a set of controllable supplementary lighting sources. All these sensors are coordinated and controlled by an industrial-grade edge computing control computer to ensure the synchronization of data acquisition.

[0079] System calibration and coordinate system 1: This is a prerequisite for ensuring measurement accuracy. Using a high-precision calibration board, multiple cameras are jointly calibrated to determine their positional and orientation relationships, and the coordinate systems of all sensors (including 2D cameras, 3D cameras, and the weighing platform) are unified to the same world coordinate system. The calibration process needs to be performed periodically (e.g., monthly) or automatically after equipment movement to ensure long-term accuracy.

[0080] 2. High-precision dimensional measurement based on deep learning and 3D reconstruction: This is the core technology of this step, which uses computer vision algorithms to calculate the precise outer dimensions of the goods from the raw sensor data.

[0081] Acquisition and preprocessing of 3D point cloud data: The 3D depth camera directly outputs sparse or dense point cloud data. The system first filters the point cloud to remove noise points (such as ground reflections and distant backgrounds), and applies voxel downsampling to reduce the amount of data while preserving features.

[0082] Cargo Instance Segmentation and Point Cloud Clustering: In complex scenarios containing multiple goods (such as multiple packages on a conveyor belt), segmentation is necessary. The system utilizes a deep learning instance segmentation model (such as a variant based on PointNet++) to process the data directly on the point cloud or after projecting the point cloud onto a 2D image, assigning a unique instance ID to each individual cargo point cloud. This solves the separation challenge in cases where packages are close together or stacked.

[0083] Minimum Bounding Cube Calculation and Size Output: For a segmented point cloud of a single cargo item, calculate its minimum bounding cube in 3D space. This is typically done by using Principal Component Analysis (PCA) to find the principal orientation of the point cloud and then constructing a bounding box along that orientation. The final output is the length (L), width (W), and height (H) of this bounding box. The algorithm can handle regular boxes and soft, irregular packages (such as mailbags).

[0084] 3. Density-based intelligent classification and value assessment using multi-feature fusion: Based on accurate physical dimensions and weight, visual features are combined to perform more refined classification and value assessment of goods.

[0085] Feature extraction and fusion: The system extracts multiple features to construct a feature vector. Physical characteristics: calculated actual density (weight / volume), aspect ratio, and surface area to volume ratio.

[0086] Visual features: Extract texture features (roughness / smoothness), shape regularity, and material appearance features (whether it is reflective, whether it is transparent) of the cargo surface using a pre-trained convolutional neural network (CNN).

[0087] Contextual features: origin of goods (e.g., goods shipped from an e-commerce warehouse are mostly small items), current measurement time, etc.

[0088] Density-based classification models: The feature vectors described above are input into a lightweight gradient booster (such as LightGBM) or a small neural network classifier. The model is trained to output a classification probability, for example: Heavy goods: The actual density is much greater than the volumetric density, and the charge should be based on the actual weight.

[0089] Bulk goods / lightweight goods: The actual density is much less than the volumetric density, and should be charged according to volumetric weight.

[0090] Standard goods: The actual density and the volumetric density are close, and the higher of the two is used for billing.

[0091] 4. Continuous optimization and closed-loop calibration of the measurement system Online learning and incremental training: The system establishes a feedback loop. When operators review, confirm, or correct the automatic measurement results in the system, this "human-machine verification" data is automatically collected and used as new training samples to periodically incrementally train the visual measurement and classification model, enabling it to adapt to new types of goods and packaging materials.

[0092] Automated accuracy assessment and reporting: The system periodically samples goods passing through the measurement station and compares the results with those from higher-precision offline measurement equipment (such as 3D scanners), automatically generating measurement accuracy reports (such as average error, maximum error, and error distribution) to ensure that system performance is always maintained above acceptable standards.

[0093] In the embodiments of this specification, complex, variable, and implicit logistics billing rules are transformed into a digital rule system that is understandable, executable, and manageable by computers, and an intelligent engine capable of applying these rules quickly, accurately, and flexibly is constructed.

[0094] 1. Digital modeling of billing rules based on knowledge graphs transforms unstructured, text-based contract terms and price lists into structured, interconnected knowledge graphs, which forms the basis for intelligent rule generation.

[0095] Ontology Construction: First, the core concepts (entities) and relationships in the logistics billing domain are defined. Entities include: carriers, service products (such as "same-day delivery" and "economy express"), geographical regions (countries, provinces, cities, districts / counties, postal code segments), cargo attributes (weight segments, volume segments, product categories), and fee types (basic freight, fuel surcharge, long-distance delivery fee, insurance fee, etc.). Relationships include: providing, applicable, including, excluding, and overlapping.

[0096] Knowledge Extraction and Graph Construction: Utilizing natural language processing technology, triples of (entity, relation, entity) are automatically extracted from historical contracts, PDF quotations, and web pages. For example, from the text "East China (Jiangsu, Zhejiang, and Shanghai) mutual shipping, first kilogram 12 yuan, additional kilogram 2 yuan," relations such as (service product: standard express, applicable region, East China) and (service product: standard express, included fees, basic shipping fee: first kilogram 12 yuan) can be extracted. These triples are stored in a graph database, forming a vast, interconnected knowledge network of billing rules.

[0097] Visualized rule management: Provides business personnel with a visual knowledge graph editing interface. They can intuitively define and modify billing rules by dragging and dropping nodes and lines, much like drawing a mind map, without writing any code. For example, a new "holiday surcharge" node can be easily created and associated with a specific date range and a specific service product.

[0098] 2. Design and execution of a high-performance rule engine: Based on the constructed knowledge graph, we design an inference engine that can efficiently handle massive and complex rule matching.

[0099] Rule Expression and Compilation: The engine supports rule writing in natural language-like formats or DSLs (Domain-Specific Languages). Each rule is compiled into efficient internal executable code. Rules typically include: Conditions section: Composed of multiple atomic conditions combined using AND, OR, or NOT. Conditions can be applied to any attribute of the waybill.

[0100] Action section: Defines the cost calculation operation to be performed when the conditions are met, such as setting a cost item, applying a discount, or triggering another rule.

[0101] Rete Algorithm Optimization for Matching: The core of the engine uses the Rete algorithm or its variants. This algorithm constructs a network to perform pattern matching between facts (waybill data) and rule conditions, and caches some matching results, thus avoiding a full match every time new data or rules are added. This allows the system to complete the matching of all applicable rules for a single waybill within milliseconds, even with tens of thousands of rules.

[0102] Conflict Detection and Resolution Strategies: When multiple rules are triggered simultaneously and their actions may conflict (e.g., one rule grants customer A a discount, while another rule prohibits the discount due to the special nature of the goods), the engine will adjudicate based on predefined conflict resolution strategies. These strategies include: priority strategies (each rule has a clear priority), specificity strategies (rules with more specific conditions prevail), and ordering strategies. The adjudication process and results will be recorded to ensure transparent and auditable billing.

[0103] 3. Dynamic management and intelligent learning of rules Rule lifecycle management: The system manages the entire lifecycle of rules, including creation, testing, release, deployment, monitoring, and decommissioning. It supports canary releases of rules, allowing new rules to be applied to a small number of waybills for testing before full deployment. It provides robust version control capabilities, enabling users to revert to rule sets at any historical point in time for simulation or reconciliation.

[0104] Rule mining and recommendation: The system analyzes historical billing data and manual adjustment records, and uses association rule learning (such as the Apriori algorithm) or decision tree algorithm to automatically discover potential, undefined billing patterns. For example, the system may discover that "all electronic products shipped from a certain warehouse to a certain region have actually had a 'fragile item handling fee' manually added by the operator," and thus recommend to the administrator whether to solidify this into a formal rule.

[0105] 4. Panoramic billing simulation and intelligent reconciliation The "sand table" simulation allows users to input or select a batch of waybill parameters at the decision-making front end. The system then calls the rules engine to perform simulation calculations, instantly displaying detailed cost breakdowns and totals for different carriers, products, and route combinations. This provides strong data support for market quotations and customer bidding scheme development.

[0106] Automated discrepancy reconciliation: During the monthly billing cycle, the system compares its calculated "expected bill" with the "actual bill" provided by the carrier line by line. This comparison goes beyond just the total amount; it delves into each shipment and each expense item. The engine automatically analyzes the reasons for discrepancies, such as: "An extra 'excess length fee' appears on the bill, but according to our records and rules, the cargo length did not exceed the limit." These discrepancies are categorized and summarized into reconciliation reports, greatly improving financial reconciliation efficiency and plugging cost loopholes.

[0107] In some embodiments, the method further includes: dynamic optimization and predictive analysis; This step aims to move beyond passive billing and proactively leverage data insights and algorithmic models to achieve pre-emptive cost prediction, in-process optimization, and continuous cost reduction, transitioning from "accounting" to "intelligent control."

[0108] 1. Multidimensional feature engineering and cost driver factor analysis: In-depth analysis of internal and external factors affecting logistics costs to build a feature foundation for prediction and optimization.

[0109] Feature library construction: Hundreds of features are extracted from historical data, real-time data, and external data, including: Core cargo characteristics: such as chargeable weight, volume, density classification, cargo value, and category code obtained in previous steps.

[0110] Spatiotemporal routing characteristics: origin and destination latitude and longitude, administrative level, transportation distance, route type (trunk line / branch line / terminal line), transit points, and expected time period (whether it is peak / night).

[0111] Market environment characteristics: real-time fuel price index, highway toll standards, congestion index of specific areas (such as industrial parks and ports), and seasonal fluctuation factors (such as the "Double Eleven" coefficient).

[0112] Commercial and performance characteristics: customer contract price, carrier tender price, promised delivery time, whether cash on delivery is required, and whether a signed return document is required.

[0113] Feature importance analysis: Using the built-in features of tree models (such as Random Forest and XGBoost) or model interpretation tools like SHAP, quantify the impact of each feature on the final cost. This helps to focus on key cost drivers; for example, it may be found that "whether the destination is remote" is a more important cost factor than "distance".

[0114] 2. A multi-objective collaborative intelligent optimization model is established to find the optimal balance point among multiple objectives such as cost, timeliness, and service under multiple business constraints.

[0115] Problem Modeling: Given a batch of orders awaiting shipment, the optimization objective is to select the optimal combination of carrier, service products, and routes. This is a typical combinatorial optimization problem, which can be formally described as follows: Decision variable: X_{ij} ∈ {0, 1}, representing whether order i chooses option j (option j defines a combination of carrier, product, and route).

[0116] Objective function: Minimize total cost while minimizing average delivery time and improving performance stability. This is a multi-objective optimization problem, commonly solved using the weighted sum method or by finding the Pareto front.

[0117] Model Solving: For large-scale problems, exact solutions (such as integer programming) can be too time-consuming. This system employs metaheuristic algorithms (such as genetic algorithms and simulated annealing) or reinforcement learning to efficiently find high-quality approximate solutions. The algorithm explores a solution space comprised of multiple objectives such as "cost," "timeliness," and "reliability," ultimately outputting a set of Pareto optimal solutions for decision-makers to select based on current business priorities.

[0118] 3. Machine Learning-Based Cost Prediction and Anomaly Detection Cost forecasting model: This model predicts logistics costs for a future period (such as next week or next month) to support budgeting and resource planning. The system can employ an ensemble learning approach that combines time-series models (such as Prophet, which excels at capturing trends, seasonality, and holiday effects) with machine learning regression models (such as XGBoost Regressor, which excels at handling multi-feature nonlinear relationships).

[0119] Real-time anomaly detection and early warning: During transportation operations, the system monitors cost-related indicators in real time. It utilizes unsupervised learning algorithms such as Statistical Process Control (SPC) or Isolation Forest to detect abnormal fluctuations. For example, if the average daily cost per ton-kilometer on a particular route suddenly spikes, exceeding historical normal fluctuations, the system will immediately issue an alert, indicating a possible abnormal event (such as temporary road closures or detours due to accidents), prompting management to intervene and investigate promptly.

[0120] 4. Strategy Simulation and Closed-Loop Optimization The "If-Then" scenario simulator provides management with an interactive simulation tool. Users can flexibly adjust various "control variables," such as "if fuel prices rise by 10%," "if we sign an annual framework agreement with carrier A and receive an additional 5% discount," or "if some goods from the North China warehouse are diverted to the Central China warehouse." Based on historical data and optimization algorithms, the system will quickly simulate the potential impact of these strategy changes on overall costs, timeliness, and network efficiency, providing data insights for strategic decision-making.

[0121] Closed-loop feedback and model iteration: The actual execution results of all optimization suggestions (whether they are adopted, and the real costs and effects after adoption) are collected by the system and compared with the prediction results. These "strategy-result" pairs serve as new training data and are continuously fed back to the optimization and prediction models, enabling them to learn from the actual feedback of the business and become increasingly intelligent with use, forming a reinforced closed loop from "analysis" to "decision-making" and then to "learning".

[0122] In some embodiments, the method further includes: visual decision support and system integration; This step presents the data, insights, and capabilities generated from all previous steps to users in different roles through an intuitive, interactive interface and standardized, open interfaces, while ensuring that the system can be flexibly and robustly integrated into the enterprise's existing IT ecosystem.

[0123] 1. A role-based and scenario-based data visualization and interactive analysis system provides a multi-level, customizable data portal to meet the diverse needs of everyone from operators to CEOs.

[0124] Strategic Dashboard (for Executives): Provides a company-wide panoramic view of logistics. At its core are several key "North Star Metrics," such as "Single-Piece Logistics Cost Trends," "Proportion of High-Value Cargo and Optimization Potential," and "Cost Achievement Rate (vs. Budget)." Interactive charts illustrate how these metrics change over time and by business unit, supporting drill-down analysis. For example, clicking on an area on the map where costs are surging allows drill-down analysis to see which specific routes, customers, and products caused the cost increase.

[0125] Tactical Layer Analyzer (for Operations / Finance Managers): Offers a wealth of multidimensional analysis and reporting tools. Users can freely drag and drop dimensions (such as time, region, product, carrier) and metrics (such as cost, weight, number of tickets) to perform cross-analysis. The system comes pre-loaded with commonly used analysis templates, such as "Scatter Plot Comparing Costs and Delivery Times of Each Carrier," "Monthly Cost Structure Waterfall Chart," and "Sankey Diagram for Analyzing Abnormal Costs." It supports saving analysis charts to personal dashboards or automatically generating and pushing PDF reports periodically.

[0126] Execution-level workbench (for operations / customer service personnel): The interface design is task-centric, clearly presenting to-do items. Examples include: "List of packages with size discrepancies to be reviewed," "Billing rule conflicts to be confirmed," and "Carrier billing discrepancies to be processed." One-click operations are provided, such as "Confirm measurement results," "View billing details," and "Initiate a dispute." Instant messaging is integrated for convenient and rapid communication regarding abnormal orders.

[0127] 2. Intelligent alerts, automated workflows, and collaboration transform system insights into automated actions, improving operational efficiency.

[0128] Configurable intelligent alarm center: Users can customize alarm rules based on business concerns. For example: "When the volumetric weight of a single shipment exceeds 100% of the actual weight, send a 'suspected over-packaging' alarm to the warehouse manager," or "When the daily cost of a certain route exceeds the forecast by 20%, send an alert to the route planner." Alarms can be delivered through multiple channels such as email, SMS, and instant messaging tools.

[0129] Automated approval and processing workflow: Alerts can trigger predefined approval flows. For example, when the system detects a "special operation fee" application, but the goods information does not meet the free criteria, an approval form can be automatically generated and routed to the relevant supervisor for approval. After approval, the system automatically updates the billing rules. This achieves a closed-loop process from "problem detection" to "problem resolution."

[0130] 3. Open architecture and ecosystem integration: The system adopts a microservice architecture to ensure high availability, scalability, and ease of integration.

[0131] API-first design: All core business capabilities, such as "cost calculation," "waybill synchronization," and "report generation," are exposed through a set of well-designed and well-documented RESTful APIs or gRPC interfaces. This allows external systems (such as ERP, OMS, and customer service systems) to easily integrate the system's capabilities.

[0132] Pre-built connectors and out-of-the-box functionality: For mainstream ERP systems, as well as common TMS and WMS systems on the market, the system provides pre-developed and pre-configured connectors, greatly reducing the workload and complexity of initial integration. It also supports data synchronization via ETL tools and message brokers.

[0133] Hybrid cloud deployment support: The system supports deployment in public cloud, private cloud or hybrid cloud environments to meet the data security and architecture requirements of different enterprises.

[0134] 4. System observability, operation and maintenance, and continuous learning Comprehensive observability: The system has a built-in, comprehensive monitoring, logging, and tracing system. Dashboards display the real-time health status (UP / DOWN), performance metrics (request latency, QPS), and resource utilization (CPU, memory) of each microservice. Any anomalies can be quickly located to the specific service, module, or even line of code.

[0135] Model Operations Management: Establish independent model registry, version management, and release pipelines for the AI ​​models (visual measurement, density classification, cost prediction, etc.) involved in steps two and four. Monitor model performance in the production environment (e.g., whether prediction accuracy declines), and automatically trigger retraining and deployment processes when performance falls below a threshold to ensure the continued effectiveness of AI capabilities.

[0136] User behavior analysis and experience optimization: Anonymously collect interaction data from the user interface to analyze which functions are most frequently used and which operation paths are most complex. Leverage these insights to continuously optimize the user interface and workflows, lowering the barrier to entry and improving user experience and operational efficiency. Ultimately, the entire system becomes an intelligent agent capable of self-awareness, self-optimization, and continuous learning from business feedback.

[0137] This solution's multimodal data fusion architecture: builds a unified data platform, integrates multi-source heterogeneous data from the entire logistics chain, breaks down information silos, and achieves end-to-end supply chain data connectivity.

[0138] Intelligent density recognition engine: Based on computer vision and deep learning technology, it automatically identifies the three-dimensional dimensions and density characteristics of goods, enabling accurate classification of heavy / bulky goods.

[0139] Dynamic rule adaptation mechanism: Using knowledge graph and rule engine technology, it dynamically adapts to changes in billing rules of different carriers, reducing the workload of manual configuration.

[0140] Predictive cost optimization: Utilizing time series analysis and machine learning algorithms, it predicts transportation cost trends and provides optimal transportation solutions.

[0141] Blockchain-based evidence storage and intelligent reconciliation: Combining blockchain technology to achieve tamper-proof evidence storage of waybill data and building an AI-driven intelligent reconciliation system.

[0142] This technical solution provides an AI-based intelligent cost calculation method for heavy cargo in express logistics. By constructing a unified data platform, it integrates multi-source heterogeneous data from the entire logistics chain, breaking down information silos and achieving end-to-end supply chain data connectivity. Based on computer vision and deep learning technologies, it automatically identifies the three-dimensional dimensions and density characteristics of goods, achieving accurate classification of heavy / bulky cargo. Employing knowledge graph and rule engine technologies, it dynamically adapts to changes in billing rules from different carriers, reducing manual configuration workload. Using time series analysis and machine learning algorithms, it predicts transportation cost trends and provides optimal transportation solution suggestions. Combined with blockchain technology, it achieves tamper-proof storage of waybill data, constructing an AI-driven intelligent reconciliation system. This improves billing accuracy and operational efficiency.

[0143] This specification also provides a device for determining logistics costs, such as... Figure 6 As shown, the device includes: The multi-dimensional data acquisition module 610 is used to build a unified data platform, collect multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently fuse the data to obtain multi-dimensional data characterizing the package cost calculation strategy. The feature prediction module 620 is used to predict the physical size, weight, and visual features of express parcels based on 3D vision and AI technology. The category determination module 630 is used to determine the target category of the express parcel based on its physical size, weight, and visual characteristics. The strategy determination module 640 is used to transform unstructured, text-based contract terms and price lists into a structured, related knowledge graph based on the multi-dimensional data, and to determine the package fee calculation strategy based on the knowledge graph. The cost determination module 650 is used to determine the estimated cost of the express package based on the package cost calculation strategy, the physical size, weight, visual characteristics, and target category of the express package.

[0144] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: By acquiring heterogeneous data sources from express parcels through diverse data interfaces, multi-source heterogeneous data can be obtained. The multi-source heterogeneous data is cleaned, verified, and augmented to obtain processed data. The processed data is subjected to entity parsing, and information from different sources describing the same entity is intelligently fused to obtain multi-dimensional data representing the package cost calculation strategy.

[0145] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: The processed data is parsed to determine whether an entity has a primary key; the primary key is either the waybill number or the order number. If the primary key exists, the processed data is associated according to the primary key, and the processed data corresponding to the same primary key are merged to obtain the first merged data; If the primary key does not exist, start the fuzzy matching algorithm to determine the combination of multiple fields in each entity; and calculate the similarity score between any two entities based on the field combinations corresponding to each of the two entities. Multiple entities with similarity scores greater than a preset threshold are identified as associated entities; and the processed data corresponding to each of the associated entities are fused to obtain second fused data. Based on the first fused data and the second fused data, multi-dimensional data characterizing the package cost calculation strategy are determined. The device further includes a graph display module, used to construct and display a data association graph based on the multi-dimensional data.

[0146] In one exemplary embodiment, the multidimensional data acquisition module is further configured to: Based on the first fused data and the second fused data, create a waybill entity object corresponding to each associated entity; Based on the timeline, determine the order information, warehousing, sorting, trunk transportation, delivery and receipt information corresponding to each waybill entity, and obtain the full process events of each waybill entity. Based on the cost elements of each event in the entire process, multi-dimensional data representing the package cost calculation strategy are determined, and a logistics digital twin is constructed.

[0147] In one exemplary embodiment, the feature prediction module is further configured to: A multimodal sensor array was constructed based on 3D vision and AI technologies; The weight, visual features, and 3D point cloud data of the express parcel are acquired based on the multimodal sensor array. The three-dimensional point cloud data is filtered and denoised to obtain processed point cloud data; Based on the processed point cloud data, the physical dimensions of the express package are determined.

[0148] In one exemplary embodiment, the feature prediction module is further configured to: Based on the processed point cloud data, the number of express parcels is determined. If there are multiple express parcels, the processed point cloud data is segmented into instances to obtain the point cloud data corresponding to each express parcel, and an instance ID is assigned to the point cloud data corresponding to each express parcel. For each express parcel, calculate the minimum bounding cube of the point cloud data; Principal component analysis is performed on the minimum bounding cube to determine the main direction of the point cloud data, and a bounding box is constructed along the main direction. The length, width and height of the bounding box are then output to obtain the physical dimensions of the express parcel corresponding to the point cloud data.

[0149] In one exemplary embodiment, the category determination module is further configured to: Based on the physical dimensions of the express package, determine the volume, aspect ratio, and surface area-to-volume ratio of the express package; The actual density of the express package is determined based on the ratio of its weight to its volume. The visual features of the express package are input into a convolutional neural network to extract texture features, shape regularity, and material appearance features. Obtain the origin information and current time of the express package as contextual features; The actual density, aspect ratio, surface area to volume ratio, texture features, shape regularity, material appearance features, and contextual features of the express package are input into the density classification model, and the target category of the express package is output; the target category represents the density category of the express package.

[0150] The apparatus and method embodiments described herein are based on the same inventive concept.

[0151] This specification provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program, which is loaded and executed by the processor to implement the logistics cost determination method provided in the above method embodiments.

[0152] Embodiments of the present invention also provide a computer storage medium, which can be disposed in a terminal to store at least one instruction or at least one program related to implementing a logistics cost determination method in the method embodiments. The at least one instruction or at least one program is loaded and executed by the processor to implement the logistics cost determination method provided in the above method embodiments.

[0153] Embodiments of the present invention also provide a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the logistics cost determination method provided in the above-described method embodiments.

[0154] Optionally, in the embodiments of this specification, the storage medium may be located at at least one of the multiple network servers in a computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0155] The memory described in the embodiments of this specification can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for the functions, etc.; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory may also include a memory controller to provide the processor with access to the memory.

[0156] The logistics cost determination method provided in the embodiments of this specification can be executed on a mobile terminal, computer terminal, server, or similar computing device. Taking running on a server as an example, Figure 7 This is a hardware structure block diagram of a server for a logistics cost determination method provided in the embodiments of this specification. (Example) Figure 7As shown, the server 700 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 710 (CPUs 710 may include, but are not limited to, microprocessors (MCUs) or programmable logic devices (FPGAs), a memory 730 for storing data, and one or more storage media 720 (e.g., one or more mass storage devices) for storing application programs 723 or data 722. The memory 730 and storage media 720 may be temporary or persistent storage. The program stored in the storage media 720 may include one or more modules, each module may include a series of instruction operations on the server. Furthermore, the CPU 710 may be configured to communicate with the storage media 720 and execute the series of instruction operations stored in the storage media 720 on the server 700. Server 700 may also include one or more power supplies 760, one or more wired or wireless network interfaces 750, one or more input / output interfaces 740, and / or one or more operating systems 721, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0157] The input / output interface 740 can be used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of server 700. In one example, the input / output interface 740 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the input / output interface 740 may be a radio frequency (RF) module used for wireless communication with the Internet.

[0158] Those skilled in the art will understand that Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, server 700 may also include... Figure 7 The more or fewer components shown, or having the same Figure 7 The different configurations shown.

[0159] As can be seen from the embodiments of the logistics cost determination method, apparatus, electronic device, or storage medium provided by the present invention, the present invention constructs a unified data platform, collects multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently fuses the data to obtain multi-dimensional data characterizing the parcel cost calculation strategy; based on 3D vision and AI technology, it predicts the physical size, weight, and visual characteristics of express parcels; based on the physical size, weight, and visual characteristics of the express parcels, it determines the target category of the express parcels; based on the multi-dimensional data, it transforms unstructured, text-based contract terms and price lists into structured, related knowledge graphs, and determines the parcel cost calculation strategy according to the knowledge graphs; based on the parcel cost calculation strategy, the physical size, weight, visual characteristics, and target category of the express parcels, it determines the estimated cost of the express parcels. The present invention, by constructing a unified data platform, integrates multi-source heterogeneous data from the entire logistics chain, breaks down information silos, and achieves end-to-end supply chain data connectivity; based on computer vision and deep learning technology, it automatically identifies the 3D size and density characteristics of goods, achieving accurate classification of heavy goods and bulky goods; and it can accurately predict the cost of express parcels.

[0160] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0161] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0162] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer storage medium, such as a read-only memory, a disk, or an optical disk.

[0163] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining logistics costs, characterized in that, The method includes: A unified data platform is built to collect multi-source heterogeneous data from the logistics platform across the entire chain, and the data is intelligently integrated to obtain multi-dimensional data that characterizes the parcel cost calculation strategy. Based on 3D vision and AI technology, predict the physical size, weight and visual features of express parcels; The target category of the express parcel is determined based on its physical size, weight, and visual characteristics. Based on the multi-dimensional data, unstructured, text-based contract terms and price lists are transformed into structured, interconnected knowledge graphs, and a package fee calculation strategy is determined based on the knowledge graphs. The estimated cost of the express package is determined based on the package cost calculation strategy, the physical dimensions, weight, visual characteristics, and target category of the express package.

2. The method according to claim 1, characterized in that, The entire logistics chain collects multi-source heterogeneous data from the platform and intelligently fuses the data to obtain multi-dimensional data characterizing the package cost calculation strategy, including: By acquiring heterogeneous data sources from express parcels through diverse data interfaces, multi-source heterogeneous data can be obtained. The multi-source heterogeneous data is cleaned, verified, and augmented to obtain processed data. The processed data is subjected to entity parsing, and information from different sources describing the same entity is intelligently fused to obtain multi-dimensional data representing the package cost calculation strategy.

3. The method according to claim 1, characterized in that, The processed data undergoes entity parsing, intelligently fusing information from different sources describing the same entity to obtain multi-dimensional data characterizing the package fee calculation strategy, including: The processed data is parsed to determine whether an entity has a primary key; the primary key is either the waybill number or the order number. If the primary key exists, the processed data is associated according to the primary key, and the processed data corresponding to the same primary key are merged to obtain the first merged data; If the primary key does not exist, start the fuzzy matching algorithm to determine the combination of multiple fields in each entity; and calculate the similarity score between any two entities based on the field combinations corresponding to each of the two entities. Multiple entities with similarity scores greater than a preset threshold are identified as associated entities; and the processed data corresponding to each of the associated entities are fused to obtain second fused data. Based on the first fused data and the second fused data, multi-dimensional data characterizing the package cost calculation strategy are determined. The method further includes: constructing and displaying a data association map based on the multi-dimensional data.

4. The method according to claim 3, characterized in that, The determination of multi-dimensional data characterizing the package fee calculation strategy based on the first fused data and the second fused data includes: Based on the first fused data and the second fused data, create a waybill entity object corresponding to each associated entity; Based on the timeline, determine the order information, warehousing, sorting, trunk transportation, delivery and receipt information corresponding to each waybill entity, and obtain the full process events of each waybill entity. Based on the cost elements of each event in the entire process, multi-dimensional data representing the package cost calculation strategy are determined, and a logistics digital twin is constructed.

5. The method according to claim 1, characterized in that, The method for predicting the physical dimensions, weight, and visual features of express parcels based on 3D vision and AI technology includes: A multimodal sensor array was constructed based on 3D vision and AI technologies; The weight, visual features, and 3D point cloud data of the express parcel are acquired based on the multimodal sensor array. The three-dimensional point cloud data is filtered and denoised to obtain processed point cloud data; Based on the processed point cloud data, the physical dimensions of the express package are determined.

6. The method according to claim 5, characterized in that, Determining the physical dimensions of the express package based on the processed point cloud data includes: Based on the processed point cloud data, the number of express parcels is determined. If there are multiple express parcels, the processed point cloud data is segmented into instances to obtain the point cloud data corresponding to each express parcel, and an instance ID is assigned to the point cloud data corresponding to each express parcel. For each express parcel, calculate the minimum bounding cube of the point cloud data; Principal component analysis is performed on the minimum bounding cube to determine the main direction of the point cloud data, and a bounding box is constructed along the main direction. The length, width and height of the bounding box are then output to obtain the physical dimensions of the express parcel corresponding to the point cloud data.

7. The method according to claim 1, characterized in that, The process of determining the target category of the express parcel based on its physical dimensions, weight, and visual characteristics includes: Based on the physical dimensions of the express package, determine the volume, aspect ratio, and surface area-to-volume ratio of the express package; The actual density of the express package is determined based on the ratio of its weight to its volume. The visual features of the express package are input into a convolutional neural network to extract texture features, shape regularity, and material appearance features. Obtain the origin information and current time of the express package as contextual features; The actual density, aspect ratio, surface area to volume ratio, texture features, shape regularity, material appearance features, and contextual features of the express package are input into the density classification model, and the target category of the express package is output; the target category represents the density category of the express package.

8. A device for determining logistics costs, characterized in that, The device includes: The multi-dimensional data acquisition module is used to build a unified data platform, collect multi-source heterogeneous data from the logistics platform across the entire chain, and intelligently fuse the data to obtain multi-dimensional data that characterizes the package cost calculation strategy. The feature prediction module is used to predict the physical size, weight, and visual features of express parcels based on 3D vision and AI technology. The category determination module is used to determine the target category of the express parcel based on its physical size, weight, and visual characteristics. The strategy determination module is used to transform unstructured, text-based contract terms and price lists into a structured, related knowledge graph based on the multi-dimensional data, and to determine the package fee calculation strategy based on the knowledge graph. The cost determination module is used to determine the estimated cost of the express package based on the package cost calculation strategy, the physical size, weight, visual characteristics, and target category of the express package.

9. An electronic device, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or at least one program being loaded and executed by the processor to implement the logistics cost determination method as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the logistics cost determination method as described in any one of claims 1-7.

Images