End-to-end material unmanned aerial vehicle logistics link traceability management method and system

Abstract

The application provides an end-to-end unmanned aerial vehicle logistics link traceability management method and system, relates to the technical field of link traceability, and comprises the following steps: extracting a logistics waybill number and a multi-modal evidence package of material damage; obtaining a loading position change record of target material in an unmanned aerial vehicle cargo hold through query; calling multi-source heterogeneous original data streams, including a material state sensing stream, an unmanned aerial vehicle flight control stream and an environmental parameter stream; after damage feature deconstruction to obtain a logistics damage feature, performing damage attribution probability modeling, constructing a damage responsibility judgment matrix; performing millisecond-level space-time alignment traceability, positioning a high-probability damage-causing space-time node sequence; performing double-mode joint root cause analysis, and outputting a material damage responsibility allocation report. The application solves the technical problem that the prior art lacks an effective multi-source data fusion and space-time accurate traceability mechanism, resulting in inaccurate traceability of material damage, fuzzy responsibility attribution and influence on the reliability of medical material transportation.

Description

Technical Field

[0001] This invention relates to the field of logistics traceability technology, specifically to an end-to-end logistics traceability management method and system for unmanned aerial vehicles (UAVs). Background Technology

[0002] Drones, as an efficient, fast, and flexible delivery tool, are particularly suitable for delivering medical supplies in emergency situations. They can rapidly deliver supplies to areas with traffic congestion or inaccessibility, significantly reducing delivery time. This is especially true in remote or disaster-stricken areas, where they can dramatically improve logistics efficiency and the timeliness of supply acquisition. However, medical supplies are highly sensitive to environmental conditions and external factors during transportation. Damage during delivery can render the supplies unusable. Therefore, ensuring the safety, integrity, and timeliness of supplies during delivery is of paramount importance.

[0003] Most current material distribution systems rely on a single data source or relatively simple sensor data, which cannot comprehensively and systematically integrate multimodal information from different sensors. They also lack high-precision spatiotemporal traceability capabilities and cannot effectively synchronize and align elements such as time, location, and sensor data. Consequently, the spatiotemporal nodes where material damage occurs cannot be accurately identified. This not only makes it difficult to quickly locate the root cause of material damage but also makes it impossible to clarify the responsibility for the damage, affecting the reliability of drone delivery systems in the transportation of medical supplies. Summary of the Invention

[0004] This application provides an end-to-end method and system for tracing and managing the logistics links of unmanned aerial vehicles (UAVs) for supplies. It aims to solve the technical problems of existing technologies lacking effective multi-source data fusion and spatiotemporal precise traceability mechanisms, which leads to inaccurate tracing of damaged supplies, unclear attribution of responsibility, and affects the reliability of medical supply transportation.

[0005] The first aspect disclosed in this application provides an end-to-end method for tracing and managing the logistics link of unmanned aerial vehicles (UAVs) for materials. The method includes: receiving and parsing a damage reporting information stream uploaded by the consignee to extract a logistics waybill number and a multimodal evidence package of material damage; querying a material warehouse binding relationship database based on the logistics waybill number to obtain a record of changes in the loading position of the target material within the UAV cargo hold; using the loading position change record as a spatiotemporal index condition to retrieve a multi-source heterogeneous raw data stream from the black box, wherein the multi-source heterogeneous raw data stream includes a material status sensing stream, a UAV flight control stream, and an environmental parameter stream; after deconstructing the multimodal evidence package of material damage to obtain logistics damage features, performing damage attribution probability modeling to construct a damage responsibility determination matrix; performing millisecond-level spatiotemporal alignment tracing on the material status sensing stream based on the damage responsibility determination matrix to locate a high-probability damage-causing spatiotemporal node sequence; and performing dual-modal joint root cause analysis on the UAV flight control stream and the environmental parameter stream based on the high-probability damage-causing spatiotemporal node sequence to output a material damage responsibility allocation report.

[0006] The second aspect of this application discloses an end-to-end material drone logistics link traceability management system. This system is used in the aforementioned end-to-end material drone logistics link traceability management method. The system includes: an evidence package extraction module, used to receive and parse damage reporting information streams uploaded by the consignee to extract logistics waybill numbers and multimodal evidence packages of material damage; a query module, used to query a material warehouse binding relationship database based on the logistics waybill number to obtain records of changes in the loading position of the target material within the drone's cargo hold; and a raw data retrieval module, used to retrieve multi-source heterogeneous raw data streams from the black box using the loading position change records as spatiotemporal index conditions. The multi-source heterogeneous raw data stream includes a material status sensing stream, a UAV flight control stream, and an environmental parameter stream. A modeling module is used to perform damage attribution probability modeling and construct a damage responsibility determination matrix after deconstructing the multimodal evidence package of material damage to obtain logistics damage characteristics. A spatiotemporal alignment tracing module is used to perform millisecond-level spatiotemporal alignment tracing on the material status sensing stream based on the damage responsibility determination matrix to locate high-probability damage-causing spatiotemporal node sequences. A root cause analysis module is used to perform dual-modal joint root cause analysis on the UAV flight control stream and the environmental parameter stream based on the high-probability damage-causing spatiotemporal node sequences, and output a material damage responsibility allocation report.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects:

[0008] By receiving and parsing the damage reporting information stream uploaded by the consignee, the logistics tracking number and multimodal evidence package of material damage are extracted, providing an accurate data source for subsequent liability determination. By using the loading location change record as a spatiotemporal index condition, multi-source heterogeneous raw data streams are retrieved from the black box, enabling backtracking of material damage in a precise spatiotemporal context, avoiding information loss and errors, especially in the high-precision delivery of medical supplies. By deconstructing the damage features of the multimodal evidence package of material damage, logistics damage characteristics are obtained, and a damage liability determination matrix is ​​constructed through damage attribution probability modeling. This provides a quantitative assessment of different damage modes and their causes, enabling efficient and accurate location of the root cause of damage, especially in the process of medical supply delivery, ensuring the accurate allocation of damage liability. By performing millisecond-level spatiotemporal alignment tracing of the material status sensor flow based on the damage liability determination matrix, high-probability damage-causing spatiotemporal nodes can be accurately located. This ensures that the time and space of each material position change and status change are precisely correlated, avoiding deviations in time and space information and guaranteeing the high accuracy of the tracing process. By refining the damage time to the millisecond level and the damage location to the millimeter level, accurate spatiotemporal data is provided for subsequent root cause analysis. Based on the high-probability damage-causing spatiotemporal node sequence, combined with the UAV flight control flow and environmental parameter flow, a dual-modal joint root cause analysis is performed to output a material damage liability allocation report. By jointly analyzing the impact of UAV flight operations and the external environment on material damage, the root cause of each type of damage can be clearly determined, providing a more comprehensive and accurate damage liability allocation report.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the end-to-end material drone logistics link traceability management method provided in the embodiments of this application.

[0011] Figure 2 A schematic diagram of the end-to-end material unmanned aerial vehicle logistics link traceability management system provided in this application embodiment.

[0012] Figure labeling: Evidence package extraction module 10, query module 20, raw data retrieval module 30, modeling module 40, spatiotemporal alignment tracing module 50, root cause analysis module 60. Detailed Implementation

[0013] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0014] Example 1, as Figure 1 As shown in the figure, this application provides an end-to-end material drone logistics link traceability management method, the method including:

[0015] A100: Receives and parses the damage report information stream uploaded by the consignee to extract the logistics tracking number and the multimodal evidence package of material damage.

[0016] Upon receiving the goods, if the recipient discovers any damage, they can upload a damage reporting information stream through an online platform or API. This information stream includes details of the damage and key evidence to aid in subsequent tracing and liability determination. The damage reporting information stream undergoes data parsing and processing to extract key fields, including: the logistics tracking number, a unique identifier for each delivery task used to link the entire logistics process; and a multimodal evidence package of the damage, comprising various sensor data such as images, sound, video, or physical sensor data, describing the damage encountered by the goods during transportation.

[0017] A200: Based on the logistics waybill number, query the material warehouse binding relationship database to obtain the loading position change record of the target material in the drone's cargo hold.

[0018] Based on the extracted logistics waybill number, a query request is initiated to the material warehouse binding relationship database. This database records the loading information of each material in the drone's cargo hold, including the loading position and status of the material at various times. The database is then used to search for the loading position change records of the target material under that waybill number. Whenever the material's position changes within the drone's cargo hold, a loading position change record is generated. This record contains the following information: the initial and final loading positions of the material; the time of the change; and the status or path the material experienced during the change.

[0019] A300: Using the loading location change record as a spatiotemporal index condition, retrieve multi-source heterogeneous raw data streams from the black box, wherein the multi-source heterogeneous raw data streams include material status sensing streams, UAV flight control streams, and environmental parameter streams.

[0020] Since each change in the loading location of supplies has its specific temporal and spatial information, this information, serving as a spatiotemporal index, can help accurately locate and trace the transportation path of supplies in subsequent steps. Drones are used for the delivery of medical supplies. The drone black box refers to the device or system on the drone that records various data. The drone black box includes multiple sensor data streams, recording information such as drone flight control, supply status, and environmental parameters. Using the loading location change record as an index, the following multi-source heterogeneous raw data streams can be retrieved: Supply status sensor stream, including data on temperature, humidity, vibration, and pressure during transportation. These sensors can monitor the status of supplies in real time, especially the stability of medical supplies during transportation; Drone flight control stream, recording drone flight control data such as flight altitude, speed, and attitude. This data helps analyze the potential impact of the flight process on the supplies; Environmental parameter stream, recording environmental data such as temperature, humidity, air pressure, and wind speed. Changes in these external environments can also affect the damage of supplies, especially sensitive medical supplies.

[0021] A400: After deconstructing the damage features of the multimodal evidence package of material damage to obtain the logistics damage features, damage attribution probability modeling is performed to construct a damage responsibility determination matrix.

[0022] The multimodal evidence package of material damage here is a data package. The data contained therein is obtained through data collection on the material entity. The specific process is as follows: using image processing technology, damage features such as cracks, dents, and surface wear are extracted from the image of the damaged material; through physical sensor data, such as vibration sensors and temperature sensors, abnormal changes in the material during transportation, such as impact, vibration, and temperature anomalies, are extracted; based on structural scanning technology, deformation information of the material surface or structure is extracted to construct the topological features of the damage, such as deformation modes and crack propagation paths.

[0023] After receiving and deconstructing the multimodal evidence package of material damage, we can obtain various data from visual and physical sensors and structural scans. By extracting and analyzing this data, we can identify different types of damage characteristics.

[0024] After extracting various damage characteristics, attribution analysis is performed on these characteristics to identify potential causes of the damage. The attribution process is based on a Bayesian inference model with equal probability, taking into account the impact of multiple factors on the damage, such as impact, temperature changes, and flight process. Through attribution probability modeling, a damage liability determination matrix is ​​generated to quantify the allocation of responsibility for material damage caused by various damaging factors.

[0025] A500: Based on the damage liability determination matrix, perform millisecond-level spatiotemporal alignment tracing on the material status sensor stream to locate the spatiotemporal node sequence with high probability of damage.

[0026] Based on the constructed damage liability determination matrix, spatiotemporal alignment is performed on the material state sensor stream. Specifically, based on the damage probability model and the timestamps and location data in the material state sensor stream, the exact time and location of the damage are precisely located. This process requires millisecond-level accuracy to ensure that each spatiotemporal node accurately reflects the state changes of the material during transportation. For example, if the material experiences significant vibration or temperature change at a certain moment, this information is recorded and correlated with the damage characteristics. Through spatiotemporal alignment, a series of high-probability damage-causing nodes are identified. These nodes correspond to the critical moments and locations that led to the material damage. Ultimately, these high-probability nodes form a spatiotemporal sequence, revealing the spatiotemporal path of the damage.

[0027] A600: Based on the high-probability damage spatiotemporal node sequence, perform dual-modal joint root cause analysis on the UAV flight control flow and environmental parameter flow, and output a material damage responsibility allocation report.

[0028] Based on a high-probability spatiotemporal sequence of damage events, the system jointly analyzes the UAV flight control flow and environmental parameter flow to precisely identify the root cause of the damage. For example, at a certain moment, improper flight control operations (such as sudden acceleration or sharp turns) combined with a sudden drop in ambient temperature may lead to damage to supplies. Finally, a detailed report on the allocation of responsibility for the damage is output. This report will specify: the exact spatiotemporal nodes, indicating the precise time and location of the damage, such as "goods at coordinates XX suffered type XX damage at XX seconds due to operation XX"; and the root cause analysis, including UAV operation and environmental factors, indicating the attribution of responsibility.

[0029] Furthermore, after deconstructing the multimodal evidence package of material damage to obtain logistics damage features, damage attribution probability modeling is performed to construct a damage liability determination matrix. The method includes:

[0030] A410: Deconstruct the damage features of the multimodal evidence package of material damage to obtain multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors; A420: Based on damage spatial correlation, perform cross-modal coupling analysis on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors to obtain a damage synergy tensor; A430: After performing causation probability Bayesian inversion modeling on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, perform material sensitivity correction in conjunction with the damage synergy tensor, and output the damage responsibility determination matrix.

[0031] The multimodal evidence package for material damage contains data from various sensors, reflecting the damage to materials during transportation. Based on data type, it can be categorized into three types: visual damage features, extracted from image or video data, such as cracks, breaks, dents, and deformations; physical damage features, physical data acquired through sensors, such as vibration sensors, accelerometers, and temperature sensors, used to describe factors such as physical impacts, temperature fluctuations, and pressure changes experienced by materials during transportation; and structural damage topology, data obtained through structural scanning, describing the deformation and damage patterns of the material's surface or structure, such as crack propagation paths and localized deformations.

[0032] Each data type in the damage evidence package is processed and its features are extracted: features of each damaged area are extracted using image processing technology and converted into visual damage feature vectors. Each feature vector corresponds to a different damage pattern or type, such as cracks, scratches, and dents. Sensor data is analyzed to extract physical damage feature fragments related to the damage, such as impact energy, vibration amplitude, and temperature changes. The damage topology of the material surface or structure is reconstructed through 3D scanning to generate structural damage topology descriptors, such as deformation modes and crack propagation.

[0033] Each damage type has different spatial distributions of features, so it is necessary to map these data into a unified spatial coordinate system through spatial correlation. Specifically, visual damage features, physical damage features, and structural damage descriptors are mapped to the same coordinate system so that they can be compared and analyzed in the same spatial framework. The damage area is divided into a voxel grid, which is similar to three-dimensional pixels, thus facilitating subsequent coupled analysis.

[0034] Cross-modal coupling of damage features from different modalities combines these features into a global damage feature representation. Damage features from each modality influence the damage behavior of other modalities. For example, temperature changes may cause crack propagation, and visual damage combined with physical impact may lead to more severe damage. Coupling analysis analyzes the relationships and interactions between these features to more accurately describe and predict material damage. Cross-modal analysis yields a damage synergy tensor, which describes the interactions and weighted effects between different types of damage features.

[0035] Bayesian inference is used to model various damage characteristics, analyzing their probability of occurrence and causes. The Bayesian inversion method combines different types of damage characteristics with potential damaging factors (such as impact, environmental changes, flight operations, etc.) to perform probabilistic modeling, outputting the causative probability of each type of damage. After obtaining the causative probabilities of various damage types, the damage synergy tensor is applied to material sensitivity correction. The damage synergy tensor describes the interaction between different damage types and can therefore be used to adjust the material's response model. Through this correction, the sensitivity of the material to different damages can be more accurately reflected, thereby optimizing damage liability determination. Finally, a damage liability determination matrix is ​​generated, listing the liability attribution and probability for each damage type.

[0036] Furthermore, the damage feature deconstruction of the multimodal evidence package of material damage yields multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors. The method includes:

[0037] A411: Extract visual evidence stream, physical sensor evidence, and structural scanning evidence from the multimodal evidence package of material damage; A412: Extract multi-scale damage features from the visual evidence stream to obtain multiple visual damage feature vectors; A413: Perform time-frequency joint domain anomaly detection on the physical sensor evidence to obtain multiple physical damage feature fragments; A414: Reconstruct the three-dimensional deformation field gradient from the structural scanning evidence to obtain multiple structural damage topological descriptors.

[0038] Visual evidence streams are image or video data obtained from cameras or image sensors, used to detect visible damage to the surface of materials, such as cracks, breaks, and scratches; physical sensor evidence is data collected through physical sensors, such as vibration sensors, temperature sensors, and pressure sensors, which helps analyze the impact of physical shocks, pressures, or environmental factors on materials during transportation; structural scanning evidence is data obtained using 3D scanning equipment, such as laser scanning or ultrasonic scanning, providing detailed structural information about materials, including surface or internal deformation, crack propagation, etc.

[0039] Image processing techniques, such as edge detection, texture analysis, and segmentation algorithms, are used to process images in the visual evidence stream, extracting damage features at multiple levels: At the macro scale, large-scale damage can be detected, such as cracks, breaks, or deformed areas; at the micro scale, finer damage can be captured, such as microcracks and tiny scratches on the material surface. Through multi-scale analysis, a comprehensive understanding of different levels of damage can be obtained, yielding multiple visual damage feature vectors that describe different damage patterns on the material surface.

[0040] Given the superposition characteristics of multi-source damage in physical sensing signals, such as multiple impacts coupled in the time-domain waveform, it is necessary to use empirical mode decomposition (EMD) techniques to separate damage components, obtain instantaneous frequency characteristics by applying Hilbert-Huang transform to the intrinsic mode functions, analyze the temporal correlation of damage events in the time domain, locate the energy peak of abnormal frequency bands in the frequency domain using power spectral density, and establish a quantitative mapping between vibration energy distribution and damage intensity, thereby analyzing the damage accumulation effect and impact persistence characteristics.

[0041] Through time-frequency analysis, multiple physical damage characteristic segments were extracted, which indicate whether the materials experienced abnormal physical phenomena during transportation.

[0042] 3D scanning yields a series of 3D point cloud data, containing detailed morphology of the material's surface or interior. To better describe the damage, gradient reconstruction is performed on this point cloud data. This involves calculating the gradient of the deformation field to reveal the local deformation and damage characteristics of the material at different locations. Gradient reconstruction yields the local deformation field of the material's surface or structure, describing the degree of deformation at the damaged areas. By comparing the original structure with the scanned structure, it's possible to analyze which areas have deformed or experienced crack propagation. Multiple structural damage topological descriptors obtained through 3D deformation field gradient reconstruction describe the damage patterns of the material. These descriptors include information such as crack morphology and the distribution of deformed areas, helping to accurately locate the damage location and type.

[0043] Furthermore, based on damage spatial correlation, cross-modal coupling analysis is performed on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors to obtain the damage synergy tensor. The method includes:

[0044] A421: Construct a material reference coordinate system for the target material, perform spatial registration of the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, and then discretize the material into a three-dimensional voxel mesh to obtain multiple multimodal damage feature vectors for the voxel mesh. The multimodal damage feature vectors include visual damage intensity, physical damage intensity, and structural damage intensity. A422: Perform cross-modal coupling calculations on the multiple multimodal damage feature vectors to obtain multiple coupled feature tensor components. A423: Construct the damage synergy tensor based on the multiple coupled feature tensor components. The damage synergy tensor includes a mechanical damage index, an environmental damage index, and an interaction intensity.

[0045] In material damage analysis, a reference coordinate system is defined for the target material, that is, the reference spatial frame of the material is determined. This coordinate system is the material's own geometric coordinate system, including the material's length, width, height, and axis of symmetry. This reference coordinate system serves as the standard space for subsequent analysis, ensuring a consistent representation of all data. For example, the material's length direction is defined as the x-axis, width direction as the y-axis, and height direction as the z-axis, so that all damage features can be analyzed uniformly within this coordinate system.

[0046] Spatial registration is the process of mapping damage features from different modalities (visual, physical, and structural) to a material reference coordinate system. Since different sensors collect data in different coordinate systems, these data need to be aligned so they can share the same spatial coordinate system. The spatially registered data is further processed by discretization using a 3D voxel grid. The voxel grid, similar to 3D pixels, divides the space of the material into small cubic units, with each voxel representing a small region within the material. Discretization refines the damage features to each voxel, making them more precise. The damage features within each voxel are represented by a multimodal damage feature vector, which includes visual damage intensity, physical damage intensity, and structural damage intensity.

[0047] Cross-modal coupling computation combines visual damage features, physical damage features, and structural damage features. Since each modality represents a different dimension of material damage, they can interact or synergistically influence each other. Therefore, it is necessary to computationally unify these modal features into a higher-level model. Coupling computation involves finding the relationships between these different types of damage features and quantifying their impact. For example, a physical impact can cause visual cracks and structural deformation, thus they are coupled. Through cross-modal coupling computation, multiple coupling feature tensor components are generated. These tensor components represent the coupling relationships between different modalities. The magnitude and direction of each component represent the degree of mutual influence between different modal damages. For example, one coupling component represents the interaction between visual and physical damage, while another component represents the coupling effect between structural damage and environmental changes.

[0048] By comprehensively analyzing multiple coupled feature tensor components, a damage synergy tensor is constructed to represent the synergistic effects between various damage types and multimodal damage features. Specifically, the mechanical damage index describes the degree of mechanical damage to materials caused by physical impacts, vibrations, and compression, reflecting the damage caused by external forces during transportation; the environmental damage index describes the degree of damage caused by changes in external environmental factors, reflecting the impact of environmental factors on materials; and the interaction strength represents the intensity of interaction between different damage features. For example, the interaction between environmental changes and mechanical damage can lead to more severe damage to materials; the interaction strength measures the combined impact of these factors.

[0049] Furthermore, after performing causative probability Bayesian inversion modeling on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, and combining the damage synergy tensor for material sensitivity correction, the damage liability determination matrix is ​​output. The method includes:

[0050] A431: Extract multiple crack geometric parameters from the multiple visual damage feature vectors, perform mechanical damage type pattern matching, and obtain multiple mechanical damage likelihood probabilities; A432: Based on Bayes' theorem, fuse the prior probability of mechanical damage and the multiple mechanical damage likelihood probabilities to obtain multiple mechanical damage posterior probabilities; A433: Extract multiple impact energy features from the multiple physical damage feature segments, perform environmental damage type response correlation analysis, and obtain multiple environmental damage likelihood probabilities; A434: Based on Bayes' theorem, fuse the prior probability of environmental damage and the multiple environmental damage likelihood probabilities to obtain multiple environmental damage posterior probabilities; A 435: Extract multiple stress concentration features from the multiple structural damage topological descriptors, perform material failure characteristic coupling effect modeling, and obtain multiple interaction likelihood probabilities; A436: Based on Bayes' theorem, fuse the prior probability of interaction and the multiple interaction likelihood probabilities to obtain multiple interaction posterior probabilities; A437: Use the damage synergy tensor to perform material sensitivity correction on the multiple mechanical damage posterior probabilities, multiple environmental damage posterior probabilities, and multiple interaction posterior probabilities, to obtain the mechanical damage probability distribution, environmental damage probability distribution, and interaction probability distribution, which constitute the damage liability determination matrix.

[0051] Geometric parameters associated with the crack are extracted from the visual damage feature vector. These parameters include crack length, width, depth, shape, and orientation. Image processing techniques, such as edge detection and morphological processing, allow for accurate identification and quantification of these geometric parameters from the image. After extraction, these parameters are pattern matched against known mechanical damage types, including different forms of cracks, plastic deformation, and fracture. By comparing these patterns with known damage types, the crack type and its association with specific mechanical damage forms can be identified. Based on the matching results of crack geometric parameters and damage types, the mechanical damage likelihood probability is calculated—the probability of each possible mechanical damage type occurring. These probabilities reflect the likelihood of different damage modes occurring under the current crack characteristics. For example, a certain crack mode may be more easily caused by impact forces, while other modes are associated with excessive bending or tension.

[0052] In Bayesian inference, prior probability represents a preliminary estimate of a damage type before any data is observed. For mechanical damage types, prior probability is based on historical data, theoretical models, or expert experience. For example, if a certain damage type is typically associated with high impact force, then the prior probability of this damage type would be high in the absence of specific data. Bayes' theorem provides a method to combine prior probabilities with new evidence (in this case, crack geometry parameters) to derive posterior probabilities, which represent the updated probability of a mechanical damage type occurring after crack features are observed. According to Bayes' theorem, by fusing the prior probabilities of mechanical damage and the likelihood probabilities of crack geometry features, multiple posterior probabilities of mechanical damage are calculated. These posterior probabilities reflect the likelihood of each mechanical damage type occurring after considering crack features.

[0053] Impact energy characteristics are extracted from vibration, acceleration, or other physical sensor data across multiple physical damage feature segments. Impact energy refers to the energy accumulated when materials are subjected to external impact forces during transportation. These characteristics include the magnitude, frequency, and duration of the impact. Environmental factors may cause damage by affecting the physical properties of materials. Environmental damage type response correlation analysis identifies the relationship between impact energy and environmental factors. For example, high temperatures increase the fragility of certain materials, making them more susceptible to impact damage. Based on response correlation analysis of environmental damage types, multiple environmental damage likelihood probabilities are obtained, representing the probability of environmental damage occurring under different environmental conditions. For instance, under high temperature and humidity conditions, materials are more prone to cracking or deformation, thus resulting in a higher environmental damage likelihood probability.

[0054] Similar to the analysis of mechanical damage types, environmental damage also has a prior probability. This prior probability is based on historical data of the material, environmental conditions, and material properties. For example, if a material is frequently transported in high-humidity environments, its probability of damage in such environments is considered higher. Using Bayes' theorem, the prior probability of environmental damage is fused with the likelihood probability obtained through environmental damage type response correlation analysis to obtain the posterior probability of environmental damage. The formula is the same as for calculating the posterior probability of mechanical damage. The updated environmental damage probability is then calculated using Bayes' theorem.

[0055] Stress concentration features are extracted from structural damage topological descriptors. These features represent the degree of stress concentration in certain areas of a material under external force. Stress concentration typically occurs on the surface or in defective areas of the material's internal structure, such as crack tips or joints. These areas bear greater stress than other areas and are more prone to further damage or failure. Since materials are often subjected to multiple factors in actual use, such as external force, temperature, and humidity, it is necessary to model the coupling effects of material failure characteristics. This modeling approach can simulate the failure behavior of materials under multiple factors, especially the interactions between different damage mechanisms. Through coupling effect modeling, the likelihood probability of interaction is calculated, which is the probability of material failure under the combined action of multiple damage factors. For example, mechanical impact in a high-temperature environment increases the probability of material fracture. The probability of these interactions is quantified through the model.

[0056] In Bayesian inference, the prior probability of an interaction represents a preliminary hypothesis about the interaction of different damage factors (such as mechanical damage and environmental damage) before any observed data is available. The prior probability of an interaction is based on historical data of the material, material properties, or theoretical models. For example, in some materials, the interaction between temperature and impact is more significant than the effect of a single factor, and therefore the prior probability of such an interaction may be higher. Using Bayes' theorem, the prior probability of an interaction is fused with the likelihood probability of an interaction obtained through modeling coupling effects to calculate the posterior probability of an interaction. This process updates the probability of the interaction occurring and incorporates newly identified damage features. Through Bayesian inference, multiple posterior probabilities of interactions are calculated, representing the probability of different damage factors interacting after combining prior probabilities and likelihood data.

[0057] Under the damage synergy tensor, material sensitivity corrections are applied to the multiple posterior probabilities. This correction adjusts the probability distribution of damage by considering the differences in the responses of different materials to different damage types. For example, for some materials, environmental factors have a greater impact on damage than mechanical impact; therefore, the posterior probability of environmental damage is weighted. After material sensitivity correction, mechanical damage probability distributions, environmental damage probability distributions, and interaction probability distributions are generated. These distributions reflect the actual probabilities of different damage types and interactions in material damage, providing important evidence for subsequent liability determination. Finally, based on these probability distributions, a damage liability determination matrix is ​​generated. This matrix lists the probabilities of various damage types, including mechanical damage, environmental damage, and interactions, and assigns liability based on these probabilities.

[0058] Furthermore, the method also includes:

[0059] A610: Deconstruct the damage liability determination matrix by liability distribution to obtain multiple damage liability items; A620: Based on the multiple dominant liability probabilities of the multiple damage liability items, perform millisecond-level spatiotemporal alignment tracing on the material status sensing flow to locate multiple high-probability damage-causing spatiotemporal nodes, forming the high-probability damage-causing spatiotemporal node sequence; A630: Based on the high-probability damage-causing spatiotemporal node sequence, perform dual-modal joint root cause analysis on the UAV flight control flow and environmental parameter flow to obtain a responsibility root cause event sequence; A640: Aggregate the responsibility root cause event sequence based on the responsible subject and output the material damage liability allocation report.

[0060] The damage liability determination matrix is ​​deconstructed. Specifically, based on the probability of each damage liability, each liability item in the matrix is ​​decomposed, which includes the liability probability of each damage type.

[0061] Within each damage liability item, based on the dominant liability probability of each damage type, it can be determined which damage types contribute the most to the material damage. For example, when the probability of mechanical damage exceeds a set threshold, mechanical damage is considered the dominant factor leading to material damage. The material status sensor stream contains various sensor data from the UAV during transportation, reflecting the environmental and physical state changes experienced by the materials during transport. First, the material status stream is examined because it reflects the impact of the UAV flight control and environmental parameters on the materials, providing a temporal and spatial data foundation for subsequent analysis. Millisecond-level spatiotemporal alignment tracing is performed on the material status sensor stream. This process includes: establishing a unified spatiotemporal coordinate system to accurately map damage features to the real-time acquired sensor data spatially and temporally; and using feature matching algorithms to scan and identify abnormal events in the material status stream, especially those highly consistent with damage features. This process can accurately locate the spatiotemporal position of the damage. By comparing the state flow and damage characteristics of materials, multiple high-probability spatiotemporal nodes leading to damage are identified. These nodes represent the most likely time and spatial locations where material damage is likely to occur, determined with millisecond-level and millimeter-level location accuracy. These key spatiotemporal nodes are arranged chronologically to form a high-probability damage-causing spatiotemporal node sequence. This sequence records in detail the damage events during material transportation and provides spatiotemporal evidence for subsequent root cause analysis.

[0062] Based on the high-probability damage-causing spatiotemporal node sequence, the drone flight control flow and environmental parameter flow are jointly analyzed. The flight control flow includes flight data of the drone during transportation, such as flight speed, attitude, and acceleration, which affects the condition of the supplies. The environmental parameter flow includes data on environmental factors, such as temperature, humidity, and air pressure, which can also affect the supplies due to changes in the external environment. By comparing and analyzing these data flows with the high-probability damage-causing spatiotemporal node sequence, the root cause event sequence is identified. This sequence lists the fundamental causes of the supplies damage, which may be flight control operations, changes in environmental factors, or a combination of both. For example, vibrations may occur when the drone makes a sharp turn, or a sudden temperature rise may cause structural damage.

[0063] Based on the sequence of root cause events, the responsible parties are aggregated to identify the source of responsibility for each damage event. For example, some damage may be caused by improper operation of the drone's flight control, while others may be caused by environmental factors (such as high temperatures). The purpose of aggregating responsible parties is to clarify the attribution of responsibility for each type of damage. Finally, a material damage liability allocation report is generated, which details each type of damage and its attribution of responsibility.

[0064] Furthermore, the method involves retrieving the loading location change record of the target material within the drone's cargo hold from the material warehouse binding relationship database based on the logistics waybill number.

[0065] A210: Cargo hold displacement events are tracked using micro-MEMS sensors mounted on multiple transport materials. When the displacement offset exceeds a preset threshold, a displacement event record is triggered, resulting in multiple displacement data packet sequences. A220: Incremental off-chain / on-chain collaborative storage is performed on the multiple displacement data packet sequences using the aforementioned cargo warehouse binding relationship library, where the cargo warehouse binding relationship library is a verifiable storage index engine for the logistics alliance chain. A230: Collision-resistant hash conversion is performed on the logistics waybill number to obtain a standardized waybill hash value. A240: A zero-knowledge range query is initiated in the cargo warehouse binding relationship library based on the standardized waybill hash value to obtain the displacement data packet sequence of the target material, which serves as the loading location change record.

[0066] MEMS (Micro-Electro-Mechanical Systems) sensors are sensors that integrate miniature sensors and micro-mechanical components. They offer advantages such as high precision and small size, enabling real-time monitoring of materials. During transportation, MEMS sensors are installed on the surface of multiple materials to monitor displacement events within the cargo hold. Whenever materials shift within the cargo hold, the MEMS sensors record the amount and direction of displacement in real time to determine if significant movement has occurred. A displacement offset threshold is set; when the material displacement exceeds a certain value, it is considered a critical event and triggers displacement event recording. For example, if materials experience significant movement within the cargo hold, such as vibration or impact, and the displacement exceeds the preset threshold, the event is recorded. Each time a displacement event is triggered, the displacement data is recorded, including the displacement amount, time of occurrence, and direction of displacement. This data ultimately generates multiple displacement data packet sequences, which provide precise location information for subsequent changes in loading location and accountability.

[0067] The Material Warehouse Binding Relationship Database is a database used to store the binding information between materials and warehouse locations. It supports efficient data storage and retrieval, and features a verifiable storage index engine, meaning the stored content can be verified and queried, ensuring data immutability—all achieved through blockchain technology. Using off-chain / on-chain collaborative storage technology, displacement data packet sequences are stored in the Material Warehouse Binding Relationship Database. For on-chain storage, important data requiring long-term verification is stored on the blockchain to ensure immutability; for example, the hash value of displacement events or important metadata is stored on the blockchain. For off-chain storage, large amounts of real-time data are typically stored in off-chain databases. Off-chain storage offers higher storage capacity and lower storage costs, while still allowing verification via on-chain hash values. This incremental storage scheme ensures data security and traceability while optimizing storage and retrieval performance.

[0068] To ensure the uniqueness of logistics tracking numbers and avoid collisions between different tracking numbers, a collision-resistant hash transformation is employed. Hash transformation is a technique that converts input data (such as logistics tracking numbers) into a fixed-length output called a hash value. Collision-resistant hash algorithms ensure that even minor changes in the input will generate completely different hash values, thus preventing hash collisions. Through hash transformation, each logistics tracking number is converted into a standardized tracking hash value. This hash value serves as a unique identifier for the tracking number, facilitating subsequent storage and retrieval operations. Standardized tracking hash values ​​effectively avoid problems caused by differences in tracking number formats or other factors during data storage and retrieval.

[0069] Zero-knowledge proof is a cryptographic technique used to verify the authenticity of data without revealing any additional information. In this step, zero-knowledge range queries are initiated using zero-knowledge proofs to query data related to logistics waybill numbers in a warehouse-materials binding relationship database. Through zero-knowledge range queries, the existence of data can be verified without exposing the actual data content. For example, without revealing the specific content of the displacement data packet, it can be used to query whether there is a displacement record matching a standardized waybill hash value.

[0070] Upon successful query, a sequence of displacement data packets for the target goods, associated with the standardized waybill hash value, is obtained. These data packets record changes in the loading location of the goods during transportation, including the specific location of the goods within the cargo hold, the time of occurrence, and the amount of change. This sequence of displacement data packets serves as a record of the goods' loading location changes, providing detailed spatiotemporal data for subsequent logistics tracking and accountability.

[0071] Furthermore, the displacement data packet is structured and encapsulated with the material waybill hash value, displacement update coordinates, and offset start and end timestamps.

[0072] The waybill hash value is a unique identifier obtained by encrypting the waybill number using a collision-resistant hash algorithm. It ensures the uniqueness of the waybill number and avoids data collisions, guaranteeing accurate and tamper-proof identification of the waybill number during subsequent data storage and transmission. This hash value enables secure transmission of logistics information across multiple systems while ensuring data integrity. Displacement update coordinates represent the spatial location of the goods during transportation. Each time the goods are displaced, sensors record the coordinates of the current position relative to the cargo hold. This coordinate data provides accurate spatial information for subsequent location and tracking of the goods' movement path. Offset start and end timestamps record the specific time period of each position change, accurate to the millisecond. The start and end times of each displacement event are marked to ensure precise time synchronization, supporting subsequent spatiotemporal alignment, data matching, and damage tracing analysis.

[0073] Furthermore, using the loading location change record as a spatiotemporal index condition to retrieve multi-source heterogeneous raw data streams from the black box, the method includes:

[0074] A310: Extract K displacement update coordinates and K offset start and end timestamps from the loading position change record; A320: Match K cargo hold grid area codes according to the K displacement update coordinates; A330: Using the K offset start and end timestamps as time slice references, extract spatially constrained sensing data according to the K cargo hold grid area codes to obtain K material status sensing segments, which serve as the material status sensing stream; A340: Retrieve the UAV flight control stream and environmental parameter stream based on the global time boundary of the K offset start and end timestamps.

[0075] The loading position change log includes every change in the material's position during transportation, including the time, displacement, and direction of each change. Each change record is triggered by a displacement event, including the material's starting and ending positions, as well as the specific time period during which the change occurred. From these loading position change records, K displacement update coordinates and K offset start and end timestamps are extracted. The displacement update coordinates record the spatial coordinates of each change in the material's position, with each coordinate corresponding to a specific change. The offset start and end timestamps record the specific time period during each displacement, accurate to the millisecond level. K represents an integer, indicating the number of position updates the material underwent during transportation.

[0076] The cargo hold is divided into multiple grid regions, each identified by a cargo hold grid region code. Each grid region represents a specific spatial area within the cargo hold, used for precise location of supplies. The grid region division method can be flexibly adjusted according to the cargo hold size and the type of supplies. The extracted K displacement update coordinates are matched with the grid regions within the cargo hold. Each updated supply coordinate is mapped to a specific grid region within the cargo hold and identified by the cargo hold grid region code. This ensures that each supply displacement corresponds to a specific location within the cargo hold. Ultimately, K cargo hold grid region codes are obtained, describing the spatial location corresponding to each supply displacement.

[0077] K offset start and end timestamps are used as the time slicing reference to divide the sensing data into multiple time periods. Each time period corresponds to a positional change process of the materials. In the material state stream, the precision of the time slice is at the millisecond level. Based on the K cargo hold grid area codes, the specific location of the materials within each time period can be located. Through spatial constraints, only sensing data related to these specific grid areas is extracted. This data comes from the material state sensing stream and reflects the state changes of the materials during transportation. In this way, K material state sensing segments are obtained, each segment containing sensing data at a specific location within a specific time period.

[0078] Using the time information from K offset start and end timestamps, the time range in the material status sensing stream is determined. These timestamps serve as global time boundaries, ensuring the synchronization of subsequent data retrieval with time. During the analysis of the material status sensing stream, in addition to the material's own status data, the impact of UAV flight control data and environmental data also needs to be considered. Therefore, the following data are retrieved based on the global time boundaries: UAV flight control stream, including data such as the UAV's attitude, speed, and acceleration during flight. Changes in the flight control stream may affect the material; for example, sharp turns or acceleration may cause displacement or damage to the material. Environmental parameter stream, including data such as ambient temperature, humidity, and air pressure, can affect the stability or status changes of the material. By retrieving this data, the impact on the material during transportation can be fully described, ensuring the comprehensiveness of the analysis.

[0079] Example 2, based on the same inventive concept as the end-to-end material drone logistics link traceability management method in the aforementioned examples, such as... Figure 2 As shown in the figure, this application provides an end-to-end material unmanned aerial vehicle (UAV) logistics link traceability management system, the system including:

[0080] The evidence package extraction module 10 is used to receive and parse the damage reporting information stream uploaded by the consignee to extract the logistics waybill number and the multimodal evidence package of material damage; the query module 20 is used to query the material warehouse binding relationship database based on the logistics waybill number to obtain the loading position change record of the target material in the drone cargo hold; the raw data retrieval module 30 is used to retrieve the loading position change record as a spatiotemporal index condition from the black box to retrieve the multi-source heterogeneous raw data stream, wherein the multi-source heterogeneous raw data stream includes the material status sensing stream, the drone flight control stream, and the environmental parameter stream; The modeling module 40 is used to perform damage attribution probability modeling and construct a damage responsibility determination matrix after deconstructing the damage features of the multimodal evidence package of material damage to obtain logistics damage features; the spatiotemporal alignment tracing module 50 is used to perform millisecond-level spatiotemporal alignment tracing of the material status sensing flow based on the damage responsibility determination matrix to locate the high-probability damage-causing spatiotemporal node sequence; the root cause analysis module 60 is used to perform dual-modal joint root cause analysis of the UAV flight control flow and environmental parameter flow based on the high-probability damage-causing spatiotemporal node sequence and output a material damage responsibility allocation report.

[0081] Furthermore, the modeling module 40 is used to perform the following operation steps:

[0082] The damage feature deconstruction of the multimodal evidence package of material damage yields multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors. Based on damage spatial correlation, cross-modal coupling analysis is performed on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors to obtain a damage synergy tensor. After performing causation probability Bayesian inversion modeling on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, material sensitivity correction is performed in conjunction with the damage synergy tensor to output the damage liability determination matrix.

[0083] Furthermore, the modeling module 40 is used to perform the following operation steps:

[0084] Visual evidence stream, physical sensor evidence, and structural scanning evidence are extracted from the multimodal evidence package of material damage; multi-scale damage feature extraction is performed on the visual evidence stream to obtain multiple visual damage feature vectors; time-frequency joint domain anomaly detection is performed on the physical sensor evidence to obtain multiple physical damage feature fragments; and three-dimensional deformation field gradient reconstruction is performed on the structural scanning evidence to obtain multiple structural damage topological descriptors.

[0085] Furthermore, the modeling module 40 is used to perform the following operation steps:

[0086] A reference coordinate system for the target material is constructed. After spatial registration of the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, three-dimensional voxel mesh discretization is performed to obtain multiple multimodal damage feature vectors of multiple voxel meshes. The multimodal damage feature vectors include visual damage intensity, physical damage intensity, and structural damage intensity. Cross-modal coupling calculation is performed on the multiple multimodal damage feature vectors to obtain multiple coupled feature tensor components. The damage synergy tensor is constructed based on the multiple coupled feature tensor components. The damage synergy tensor includes mechanical damage index, environmental damage index, and interaction intensity.

[0087] Furthermore, the modeling module 40 is used to perform the following operation steps:

[0088] Multiple crack geometric parameters are extracted from the multiple visual damage feature vectors, and mechanical damage type pattern matching is performed to obtain multiple mechanical damage likelihood probabilities. Based on Bayes' theorem, the prior probability of mechanical damage and the multiple mechanical damage likelihood probabilities are fused to obtain multiple mechanical damage posterior probabilities. Multiple impact energy features are extracted from the multiple physical damage feature fragments, and environmental damage type response correlation analysis is performed to obtain multiple environmental damage likelihood probabilities. Based on Bayes' theorem, the prior probability of environmental damage and the multiple environmental damage likelihood probabilities are fused to obtain multiple environmental damage posterior probabilities. Multiple stress concentration features are extracted from the multiple structural damage topological descriptors, and material failure characteristic coupling effect modeling is performed to obtain multiple interaction likelihood probabilities. Based on Bayes' theorem, the prior probability of interaction and the multiple interaction likelihood probabilities are fused to obtain multiple interaction posterior probabilities. The damage synergy tensor is used to perform material sensitivity correction on the multiple mechanical damage posterior probabilities, multiple environmental damage posterior probabilities, and multiple interaction posterior probabilities to obtain the mechanical damage probability distribution, environmental damage probability distribution, and interaction probability distribution, which constitute the damage liability determination matrix.

[0089] Furthermore, the root cause analysis module 60 is used to perform the following steps:

[0090] The damage liability determination matrix is ​​deconstructed by liability distribution to obtain multiple damage liability items. Based on the multiple dominant liability probabilities of the multiple damage liability items, millisecond-level spatiotemporal alignment tracing is performed on the material status sensing flow to locate multiple high-probability damage-causing spatiotemporal nodes, forming the high-probability damage-causing spatiotemporal node sequence. Based on the high-probability damage-causing spatiotemporal node sequence, a dual-modal joint root cause analysis is performed on the UAV flight control flow and environmental parameter flow to obtain a responsibility root cause event sequence. Based on the responsible subject, the responsibility root cause event sequence is aggregated to output the material damage liability allocation report.

[0091] Furthermore, the query module 20 is used to perform the following operation steps:

[0092] Cargo hold displacement events are tracked using miniature MEMS sensors mounted on multiple transported materials. When the material displacement exceeds a preset threshold, a displacement event record is triggered, resulting in multiple displacement data packet sequences. Incremental off-chain-on-chain collaborative storage is performed on these multiple displacement data packet sequences using the material warehouse binding relationship database, which serves as a verifiable storage index engine for the logistics alliance chain. Collision-resistant hashing is applied to the logistics waybill number to obtain a standardized waybill hash value. Based on this standardized waybill hash value, a zero-knowledge range query is initiated in the material warehouse binding relationship database to obtain the displacement data packet sequence of the target material, which serves as the loading location change record.

[0093] Furthermore, the displacement data packet is structured and encapsulated with the material waybill hash value, displacement update coordinates, and offset start and end timestamps.

[0094] Furthermore, the raw data retrieval module 30 is used to perform the following operation steps:

[0095] K displacement update coordinates and K offset start and end timestamps are extracted from the loading position change record; K cargo hold grid area codes are matched according to the K displacement update coordinates; using the K offset start and end timestamps as time slice references, spatial constraint perception data is extracted according to the K cargo hold grid area codes to obtain K material status sensing segments, which are used as the material status sensing stream; the UAV flight control stream and environmental parameter stream are retrieved according to the global time boundary of the K offset start and end timestamps.

[0096] Through the foregoing detailed description of the end-to-end material drone logistics link traceability management method, those skilled in the art can clearly understand the end-to-end material drone logistics link traceability management system in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section.

[0097] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An end-to-end unmanned aerial vehicle (UAV) logistics link traceability management method, characterized in that, The method includes: Receive and parse the damage report information stream uploaded by the consignee to extract the logistics tracking number and the multimodal evidence package of material damage; Based on the logistics waybill number, the change record of the loading position of the target material in the drone's cargo hold can be obtained by querying the material warehouse binding relationship database. Using the loading location change record as a spatiotemporal index condition, multi-source heterogeneous raw data streams are retrieved from the black box. The multi-source heterogeneous raw data streams include material status sensing streams, UAV flight control streams, and environmental parameter streams. After deconstructing the multimodal evidence package of material damage to obtain the logistics damage characteristics, damage attribution probability modeling is performed to construct a damage responsibility determination matrix; Based on the damage liability determination matrix, millisecond-level spatiotemporal alignment tracing is performed on the material status sensor stream to locate the high-probability damage-causing spatiotemporal node sequence; Based on the high-probability spatiotemporal node sequence of damage, a dual-modal joint root cause analysis is performed on the UAV flight control flow and environmental parameter flow to output a material damage responsibility allocation report.

2. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 1, characterized in that, After deconstructing the multimodal evidence package of material damage to obtain logistics damage features, damage attribution probability modeling is performed to construct a damage liability determination matrix. The method includes: The damage feature deconstruction of the multimodal evidence package of material damage yields multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors. Based on the damage space correlation, cross-modal coupling analysis is performed on the multiple visual damage feature vectors, multiple physical damage feature fragments and multiple structural damage topological descriptors to obtain the damage synergy tensor. After performing causation probability Bayesian inversion modeling on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, the material sensitivity correction is performed in conjunction with the damage synergy tensor, and the damage responsibility determination matrix is ​​output.

3. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 2, characterized in that, The damage feature deconstruction of the multimodal evidence package of material damage yields multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors. The method includes: Visual evidence stream, physical sensor evidence, and structural scan evidence were extracted from the aforementioned multimodal evidence package of material damage. Multi-scale damage feature extraction is performed on the visual evidence stream to obtain the multiple visual damage feature vectors; Time-frequency joint domain anomaly detection is performed on the physical sensing evidence to obtain the multiple physical damage feature fragments; The structural scanning evidence is reconstructed using a three-dimensional deformation field gradient to obtain the multiple structural damage topological descriptors.

4. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 2, characterized in that, Based on damage spatial correlation, cross-modal coupling analysis is performed on multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors to obtain a damage synergy tensor. The method includes: A reference coordinate system for the target material is constructed. After spatial registration of the multiple visual damage feature vectors, multiple physical damage feature segments, and multiple structural damage topological descriptors, three-dimensional voxel mesh discretization is performed to obtain multiple multimodal damage feature vectors of multiple voxel meshes. The multimodal damage feature vectors include visual damage intensity, physical damage intensity, and structural damage intensity. Cross-modal coupling calculations are performed on the multiple multimodal damage feature vectors to obtain multiple coupled feature tensor components; The damage synergy tensor is constructed based on the multiple coupled feature tensor components, wherein the damage synergy tensor includes a mechanical damage index, an environmental damage index, and an interaction intensity.

5. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 4, characterized in that, After performing causation probability Bayesian inversion modeling on the multiple visual damage feature vectors, multiple physical damage feature fragments, and multiple structural damage topological descriptors, and combining the damage synergy tensor for material sensitivity correction, the damage liability determination matrix is ​​output. The method includes: Multiple crack geometric parameters are extracted from the multiple visual damage feature vectors, and mechanical damage type pattern matching is performed to obtain multiple mechanical damage likelihood probabilities. Based on Bayes' theorem, the prior probability of mechanical damage and the multiple likelihood probabilities of mechanical damage are fused to obtain multiple posterior probabilities of mechanical damage. Multiple impact energy features of the multiple physical damage feature fragments are extracted and environmental damage type response correlation analysis is performed to obtain multiple environmental damage likelihood probabilities; Based on Bayes' theorem, the prior probability of environmental damage and the multiple likelihood probabilities of environmental damage are fused to obtain multiple posterior probabilities of environmental damage. Multiple stress concentration features are extracted from the multiple structural damage topological descriptors, and the coupling effect of material failure characteristics is modeled to obtain multiple interaction likelihood probabilities. Based on Bayes' theorem, the prior probabilities of the interactions and the likelihood probabilities of the multiple interactions are fused to obtain the posterior probabilities of the multiple interactions. The material sensitivity correction of the multiple mechanical damage posterior probabilities, multiple environmental damage posterior probabilities, and multiple interaction posterior probabilities is performed using the damage synergy tensor to obtain the mechanical damage probability distribution, environmental damage probability distribution, and interaction probability distribution, which constitute the damage liability determination matrix.

6. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 1, characterized in that, The method further includes: The damage liability determination matrix is ​​deconstructed by liability distribution to obtain multiple damage liability items; Based on the multiple dominant liability probabilities of the multiple damage liability items, millisecond-level spatiotemporal alignment tracing is performed in the material status sensing stream to locate multiple high-probability damage-causing spatiotemporal nodes, thus forming the high-probability damage-causing spatiotemporal node sequence. Based on the high-probability damage-causing spatiotemporal node sequence, a dual-modal joint root cause analysis is performed on the UAV flight control flow and environmental parameter flow to obtain the root cause event sequence. Based on the responsible parties, the sequence of root cause events is aggregated, and a report on the allocation of responsibility for material damage is output.

7. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 1, characterized in that, The method involves querying the material warehouse binding relationship database based on the logistics waybill number to obtain the loading location change record of the target material in the drone's cargo hold. Cargo hold displacement events are tracked by attaching miniature MEMS sensors to multiple transport materials. When the displacement of the materials exceeds a preset threshold, displacement event recording is triggered, resulting in multiple displacement data packet sequences. The material warehouse binding relationship database is used to perform incremental off-chain-on-chain collaborative storage on the multiple displacement data packet sequences, wherein the material warehouse binding relationship database is a verifiable storage index engine of the logistics alliance chain; The logistics waybill number is subjected to a collision-resistant hash transformation to obtain a standardized waybill hash value; Based on the standardized waybill hash value, a zero-knowledge range query is initiated in the material warehouse binding relationship database to obtain the displacement data packet sequence of the target material, which serves as the loading location change record.

8. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 7, characterized in that, The displacement data packet is structured and encapsulated with the material waybill hash value, displacement update coordinates, and offset start and end timestamps.

9. The end-to-end material unmanned aerial vehicle logistics link traceability management method as described in claim 8, characterized in that, Using the loading location change record as a spatiotemporal index condition, the method retrieves multi-source heterogeneous raw data streams from the black box, comprising: Extract K displacement update coordinates and K offset start and end timestamps from the loading position change record; The coordinates are updated based on the K displacements to match the K cargo hold grid region codes; Using the K offset start and end timestamps as time slice references, spatially constrained sensing data is extracted according to the K cargo hold grid area codes to obtain K material status sensing segments, which serve as the material status sensing stream. The UAV flight control flow and environmental parameter flow are retrieved based on the global time boundaries of the K offset start and end timestamps.

10. An end-to-end unmanned aerial vehicle (UAV) logistics link traceability management system, characterized in that, For implementing the end-to-end material unmanned aerial vehicle logistics link traceability management method according to any one of claims 1-9, the system comprises: The evidence package extraction module is used to receive and parse the damage report information stream uploaded by the consignee to extract the logistics waybill number and the multimodal evidence package of material damage; The query module is used to query the material warehouse binding relationship database based on the logistics waybill number to obtain the loading position change record of the target material in the drone cargo hold; The raw data retrieval module is used to retrieve multi-source heterogeneous raw data streams from the black box by using the loading location change record as a spatiotemporal index condition. The multi-source heterogeneous raw data streams include material status sensing streams, UAV flight control streams, and environmental parameter streams. The modeling module is used to perform damage attribution probability modeling and construct a damage liability determination matrix after deconstructing the damage features of the multimodal evidence package of material damage to obtain logistics damage features; The spatiotemporal alignment and tracing module is used to perform millisecond-level spatiotemporal alignment and tracing on the material status sensor stream based on the damage responsibility determination matrix, and to locate the spatiotemporal node sequence with high probability of damage. The root cause analysis module is used to perform dual-modal joint root cause analysis on the UAV flight control flow and environmental parameter flow based on the high-probability damage spatiotemporal node sequence, and output a material damage responsibility allocation report.

Images