Search

CN-121981636-A - Cargo supply chain transportation management system and method

CN121981636ACN 121981636 ACN121981636 ACN 121981636ACN-121981636-A

Abstract

The invention relates to the technical field of supply chain management, and particularly discloses a cargo supply chain transportation management system and method, which realize deep fusion analysis of three-dimensional point cloud and vehicle motion states by adopting a CNN-transporter-based hybrid learning model, can accurately capture fine displacement, deformation and stacking stability changes of cargoes, and remarkably improve the accuracy, instantaneity and early warning capability of in-transit cargo state monitoring; the dynamic early warning module is used for adaptively calling the early warning threshold according to the cargo type and combining a multi-level early warning mechanism, so that the span from single threshold judgment to multi-dimensional intelligent evaluation is realized, and the real-time positioning data, the vehicle motion state and the cargo three-dimensional visual model are dynamically synchronized through the digital twin module, so that a virtual-real mapping panoramic monitoring interface is constructed.

Inventors

  • WU SHIQUN

Assignees

  • 重庆至和供应链有限公司

Dates

Publication Date
20260505
Application Date
20260206

Claims (10)

  1. 1. The cargo supply chain transportation management system is characterized by comprising a plurality of vehicle-mounted sensing units and a control terminal; each of the on-vehicle sensing units includes: The radar scanner is used for scanning the inside of the container at a preset frequency to acquire three-dimensional point cloud data representing the spatial distribution and state of goods; the locator is used for acquiring real-time locating data of the vehicle; The communication module is used for uploading the three-dimensional point cloud data and the real-time positioning data to the control terminal; The control terminal includes: the analysis module is internally provided with a hybrid learning model based on a CNN-converter and is used for fusion analysis of the three-dimensional point cloud data and the real-time positioning data and judgment of the real-time state of goods; The dynamic early warning module is used for adaptively calling a corresponding early warning threshold according to the type of the goods, comparing the early warning threshold with the real-time state of the goods output by the analysis module and generating an early warning grade; and the digital twin module is used for constructing a digital twin model according to the real-time positioning data and the cargo state judged by the analysis module.
  2. 2. The cargo supply chain transportation management system of claim 1, The CNN-transducer-based hybrid learning model comprises a convolutional neural network module and a transducer time sequence coding module; The convolutional neural network module is used for extracting multi-level local spatial structural features of cargoes from single-frame three-dimensional point cloud data; The transducer time sequence coding module is used for coding a time sequence formed by continuous multi-frame point cloud characteristics and vehicle motion states at corresponding moments and capturing long-term dependency and dynamic modes of cargo state changes.
  3. 3. The cargo supply chain transportation management system of claim 2 wherein, The convolutional neural network extracting spatial structural features comprises the following steps: inputting single-frame three-dimensional point cloud data, and extracting initial characteristics of each point through a multi-layer perceptron network sharing weights; constructing a multi-scale local area of the point cloud by using the furthest point sampling and ball query algorithm; and in each local area, extracting and fusing the characteristics of points in the area through the characteristic aggregation layer to form a multi-level characteristic diagram representing the local geometrical shape and spatial relationship of the goods.
  4. 4. The cargo supply chain transportation management system of claim 2 wherein, The transform timing encoding module captures the dynamic pattern of state changes, comprising the steps of: inputting a time sequence formed by splicing point cloud feature vectors of continuous N frames and corresponding vehicle motion state feature vectors; adding a position code for each time in the time sequence, and reserving time sequence information of the time sequence; Inputting the coded time sequence into a multi-head self-attention layer, and capturing long-term dependence of cargo states at different moments by calculating correlation between any two time step characteristics in the sequence; And carrying out nonlinear transformation and characteristic enhancement on the output of the self-attention layer through a feedforward neural network layer to form a final coded time sequence characteristic vector.
  5. 5. The cargo supply chain transportation management system of claim 1, The analysis module outputs the real-time state of the goods through the state quantization and judgment steps, and the method comprises the following steps: inputting the time sequence feature vector output by the transducer time sequence coding module into a fully-connected network; the fully-connected network outputs one or more scalar values as a quantization index of the cargo state, wherein the quantization index comprises a displacement coefficient representing the overall deviation of the cargo, a deformation coefficient representing the deformation degree of the cargo and a stability coefficient representing the stacking stability; the analysis module compares the displacement coefficient, the deformation coefficient and the stability coefficient with the dynamic early warning threshold value acquired by the dynamic early warning module in real time, so as to judge whether the real-time state of the goods is abnormal.
  6. 6. The cargo supply chain transportation management system of claim 2 wherein, The vehicle motion state is obtained by calculation of the real-time positioning data, comprises real-time acceleration, angular velocity and attitude inclination angle of the vehicle, and is aligned and associated with a three-dimensional point cloud data frame with the same time stamp.
  7. 7. The cargo supply chain transportation management system of claim 1, The dynamic early warning module pre-stores or can dynamically receive a plurality of groups of early warning threshold parameters corresponding to different cargo types, wherein the early warning threshold parameters comprise numerical thresholds for judging cargo displacement, deformation and stacking stability abnormality.
  8. 8. The cargo supply chain transportation management system of claim 7, The cargo type is obtained by the electronic acquisition of the waybill information or is obtained by the identification of the initial three-dimensional point cloud data by the analysis module; and the dynamic early warning module loads corresponding early warning threshold parameters according to the determined cargo type.
  9. 9. The cargo supply chain transportation management system of claim 1, The digital twin module creates a virtual transportation environment based on a geographic information system map, drives the virtual vehicle model to synchronously move according to real-time positioning data, and dynamically updates the three-dimensional visualization state of goods in the container on the virtual vehicle model according to the goods state analysis result output by the analysis module.
  10. 10. A cargo supply chain transportation management method applied to the cargo supply chain transportation management system according to any one of claims 1 to 9, comprising the steps of: Scanning the inside of a container according to a preset frequency by a radar scanner to obtain three-dimensional point cloud data of the container; Acquiring real-time positioning data of the vehicle through a positioner; Uploading the three-dimensional point cloud data and the real-time positioning data to the control terminal through the communication module; Judging the real-time state of the goods through an analysis module and a dynamic early warning module and generating early warning; And driving a digital twin model in a digital twin module through the analysis results of the three-dimensional point cloud data and the real-time positioning data, so as to realize real-time synchronous mapping and management of the cargo transportation process.

Description

Cargo supply chain transportation management system and method Technical Field The invention relates to the technical field of supply chain management, in particular to a cargo supply chain transportation management system and method. Background At present, with the continuous development of logistics industry, the logistics market has higher requirements on the logistics company industry, particularly on logistics transportation in a valuable cargo supply chain, the logistics market requires logistics companies to have wider physical flow network point coverage so as to be capable of covering wider product collecting addresses, and the logistics network points of the logistics companies are required to be fully covered. In the prior art, the cargo supply chain transportation management mainly relies on manual inspection and a single video sensor to monitor the state of the cargo in transit in real time, the monitoring dimension is single, the perception precision is limited, and the real-time state of the cargo is difficult to monitor comprehensively and accurately. Accordingly, there is a need for a cargo supply chain transportation management system and method that addresses the above-described problems in the prior art. Disclosure of Invention The invention aims to provide a cargo supply chain transportation management system and method, and aims to solve the technical problems that cargo supply chain transportation management in the prior art mainly depends on manual inspection, a single video sensor is used for monitoring the state of in-transit cargoes in real time, the monitoring dimension is single, the perception precision is limited, and the real-time state of the cargoes is difficult to monitor comprehensively and accurately. In order to achieve the above purpose, the invention adopts a cargo supply chain transportation management system, which comprises a plurality of vehicle-mounted sensing units and a control terminal; each of the on-vehicle sensing units includes: The radar scanner is used for scanning the inside of the container at a preset frequency to acquire three-dimensional point cloud data representing the spatial distribution and state of goods; the locator is used for acquiring real-time locating data of the vehicle; The communication module is used for uploading the three-dimensional point cloud data and the real-time positioning data to the control terminal; The control terminal includes: the analysis module is internally provided with a hybrid learning model based on a CNN-converter and is used for fusion analysis of the three-dimensional point cloud data and the real-time positioning data and judgment of the real-time state of goods; The dynamic early warning module is used for adaptively calling a corresponding early warning threshold according to the type of the goods, comparing the early warning threshold with the real-time state of the goods output by the analysis module and generating an early warning grade; and the digital twin module is used for constructing a digital twin model according to the real-time positioning data and the cargo state judged by the analysis module. The hybrid learning model based on the CNN-transducer comprises a convolutional neural network module and a transducer time sequence coding module; The convolutional neural network module is used for extracting multi-level local spatial structural features of cargoes from single-frame three-dimensional point cloud data; The transducer time sequence coding module is used for coding a time sequence formed by continuous multi-frame point cloud characteristics and vehicle motion states at corresponding moments and capturing long-term dependency and dynamic modes of cargo state changes. The convolutional neural network extracting spatial structural features comprises the following steps: inputting single-frame three-dimensional point cloud data, and extracting initial characteristics of each point through a multi-layer perceptron network sharing weights; constructing a multi-scale local area of the point cloud by using the furthest point sampling and ball query algorithm; and in each local area, extracting and fusing the characteristics of points in the area through the characteristic aggregation layer to form a multi-level characteristic diagram representing the local geometrical shape and spatial relationship of the goods. The method for capturing the dynamic mode of the state change by the transducer time sequence coding module comprises the following steps of: inputting a time sequence formed by splicing point cloud feature vectors of continuous N frames and corresponding vehicle motion state feature vectors; adding a position code for each time in the time sequence, and reserving time sequence information of the time sequence; Inputting the coded time sequence into a multi-head self-attention layer, and capturing long-term dependence of cargo states at different moments by calculating correlation between