Intelligent airport luggage cross-platform collaborative retrieval system and method

Abstract

The application discloses a smart airport luggage cross-platform collaborative retrieval system and method. The retrieval system comprises a business system layer, an interface adaptation layer, a data standardization layer, a data core layer, a retrieval index layer, a retrieval service layer, a unified portal layer and a third-party platform terminal. The retrieval method comprises the steps of cross-platform data access and standardization processing, multi-source data association fusion and retrieval index construction, collaborative retrieval and result verification, retrieval result distribution, etc. The mapping relationship between the luggage unique identifier and other multi-dimensional information is established by using the association rule algorithm, the association fusion of multi-dimensional data is realized, and the composite B+ tree index is constructed. Through the index, data pre-screening is completed, and then precise screening is realized by combining image feature comparison and luggage state consistency verification. Finally, the retrieval result is safely distributed, thereby realizing the integrated processing of cross-platform access, multi-source data fusion and image collaborative retrieval.

Description

Technical Field

[0001] This invention belongs to the field of transportation management and smart airport technology, and specifically relates to a smart airport baggage cross-platform collaborative retrieval system and method. Background Technology

[0002] With the continuous advancement of smart airport construction, airport baggage management now covers multiple business processes, including check-in, baggage tracking, handling of irregular baggage, and transit transportation. Each process is typically supported by an independent business system. However, existing airport business systems are mostly built independently and in a decentralized manner, resulting in heterogeneous system architectures and inconsistent data standards. This makes it difficult for data to be effectively communicated and shared between systems, creating "data silos."

[0003] In baggage retrieval applications, most existing technologies can only retrieve data within a single business system, failing to achieve collaborative queries across platforms and multiple data sources. For example, check-in systems only store passenger check-in and baggage check-in information, baggage tracking systems only record baggage transport routes, and abnormal baggage systems retain records of abnormal handling. When staff need to query the complete flow information of a piece of baggage, they have to log into multiple systems sequentially, search separately, and manually compare data, resulting in cumbersome procedures and low retrieval efficiency.

[0004] Furthermore, current baggage retrieval methods largely rely on single-dimensional textual information (such as baggage number) as search criteria, lacking the integration and utilization of multi-dimensional information such as multi-faceted images, weight, and process nodes of baggage. This results in insufficient accuracy and comprehensiveness of search results, making it difficult to quickly locate target baggage. Especially in emergency situations such as lost or misdelivered baggage, existing technologies cannot meet the needs for rapid retrieval and collaborative handling, directly impacting airport service quality and passenger experience.

[0005] On the other hand, the existing system lacks a unified data interaction portal, and the search results cannot be quickly and securely distributed to third-party platform terminals such as airport operations, ground services, and airlines, resulting in information asynchrony among the participants and low efficiency in collaborative processing.

[0006] Therefore, there is an urgent need for a smart airport baggage cross-platform collaborative retrieval method that can solve the above-mentioned technical problems and realize cross-platform data access, multi-source information fusion and image collaborative retrieval. Summary of the Invention

[0007] The purpose of this invention is to provide a smart airport baggage cross-platform collaborative retrieval system and method, which realizes integrated processing of cross-platform access, multi-source data fusion and image collaborative retrieval, improves the baggage query and collaborative handling capabilities of smart airports, and optimizes airport operation efficiency and passenger experience.

[0008] To achieve the above objectives, this application provides a smart airport baggage cross-platform collaborative retrieval method, comprising the following steps: S1. Cross-platform data access and standardized processing: Construct a standardized data platform to access multi-source data from multiple heterogeneous business systems within the airport. The accessed data includes baggage number, passenger information, check-in time, baggage weight, multi-view images of baggage, baggage transportation status, and process node information. The process node information includes the time and location information of baggage check-in node, security check node, transfer node, loading node, and arrival node. Then, the multi-source data is standardized by unifying data format, mapping fields, cleaning and completing data. The standardized multi-source data is then uniformly stored in the data platform. S2. Multi-source data association, fusion, and retrieval index construction: Multi-dimensional information is extracted from standardized multi-source data, including unique baggage identifiers, multi-faceted images of baggage, passenger information, check-in time, baggage weight, baggage status, and process nodes. The unique baggage identifier is a mapping to the baggage number, and the baggage status is a mapping to the baggage transportation status. Then, an association rule algorithm is used to establish the mapping relationship between the unique baggage identifier and other multi-dimensional information, realizing the association and fusion of multi-dimensional data, and constructing a composite B+ tree index. This composite B+ tree index includes a primary index with the unique baggage identifier as the main key, and auxiliary indexes with other multi-dimensional information as keys, including passenger information, check-in time, baggage weight, baggage status, process nodes, and image feature quantization identifiers. The image feature quantization identifiers are obtained by vector quantization processing of the image feature vectors extracted from the multi-faceted images of baggage. S3. Collaborative retrieval and result verification: Receive baggage retrieval requests containing at least one retrieval condition, use the composite B+ tree index to perform retrieval execution data pre-screening to obtain preliminary retrieval results, and then combine image feature comparison and baggage status consistency verification to verify and filter the preliminary retrieval results, and output accurate collaborative retrieval results. S4. Search Result Distribution: The collaborative search results are formatted and distributed to the corresponding third-party platform terminals through a unified portal according to their permission requirements.

[0009] Furthermore, in step S1, the data platform accesses data with various heterogeneous business systems through corresponding interface adapters. The heterogeneous business systems include the airport's internal check-in system, baggage tracking system, irregular baggage system, and transfer system.

[0010] Furthermore, in step S1, the standardization process for multi-source data includes the following steps: S1.1 Data Format Unification: Convert time-based, numerical, text-based, image-based, and label-based data from multiple sources into a preset unified format; S1.2 Field Mapping: For the different fields of various heterogeneous business systems, establish a unified field mapping table to map the different fields of various heterogeneous business systems to the standard fields of the data platform; S1.3 Data Cleaning and Completion: Data cleaning algorithms are used to remove redundant and abnormal data, and missing data is filled in using completion strategies based on type and scenario.

[0011] Furthermore, in step S2, the association rule algorithm is the Apriori algorithm, and the steps for establishing the mapping relationship include: S2.1.1 Data Preprocessing: The extracted multi-dimensional information is converted into a format recognizable by the Apriori algorithm. Taking the unique baggage identifier as the core associated item, image feature values, check-in time, baggage weight, baggage status and process node information are used as associated auxiliary items of the unique baggage identifier to construct an initial associated dataset. S2.1.2 Frequent Itemset Mining: Frequent itemsets in the initial associated dataset are mined using the Apriori algorithm. A minimum support is set to mine frequent itemsets between baggage unique identifiers and multi-dimensional information. S2.1.3, Association rule generation and mapping relationship solidification: Based on the frequently mined itemsets, association rules that meet the minimum confidence level are generated, and the association rules are solidified into the association mapping table of the data platform to establish a one-to-one mapping relationship between the unique identifier of luggage and information of each dimension.

[0012] Furthermore, in step S2, the steps for constructing the composite B+ tree index include: S2.2.1 Store the physical storage address of the multi-dimensional data corresponding to the unique identifier of each baggage into the leaf node of the main B+ tree to complete the construction and balancing adjustment of the main index; S2.2.2. Construct multiple auxiliary B+ tree indexes using passenger information, check-in time, baggage weight, baggage status, process node, and image feature quantification identifier as keywords. The leaf nodes of each auxiliary index store the unique baggage identifier associated with the corresponding keyword and the pointer to the main index, thereby realizing the association with the main index. S2.2.3 Establish an index update mechanism. When multi-source data in the data platform changes, the associated mapping table is updated synchronously, and the index is dynamically adjusted to ensure the consistency between the index and the data. S2.2.4 Perform performance testing on the completed B+ tree index, optimize the index structure, and finally complete the construction of the composite B+ tree index.

[0013] Furthermore, in step S2, the vector quantization processing of the image feature vectors includes: using the K-means clustering algorithm to cluster the image feature vectors of all luggage, mapping similar image feature vectors to unique one-dimensional integer codes as image feature quantization identifiers, and storing a mapping table between the quantization identifiers and the original image feature vectors in the data platform.

[0014] Further, in step S3, the search conditions include at least one of baggage number, passenger information, baggage characteristics, baggage weight, check-in time, and baggage status. During verification, the initial search results are first checked for consistency of status, and then image feature comparison is performed. The consistency of status check is used to verify whether the current status of the baggage in the initial search results matches the process node information. If they do not match, the result is discarded. The image feature comparison verification includes the following steps: S3.1 First-level filtering: For search results with inconsistent status that have been removed, cross-validate the preliminary search results based on non-image search conditions in the search request to narrow down the search scope; S3.2 Secondary Filtering: For the results after primary filtering, perform image feature comparison, extract the feature vector of the luggage image in the search request, compare it with the feature vector of the luggage image after primary filtering, calculate the similarity, set the comparison threshold, and remove luggage with similarity lower than the comparison threshold. S3.3 Three-level screening: If only one piece of luggage remains after the second-level screening, it is determined to be the target luggage; if there are still two or more pieces of luggage with similarity higher than the comparison threshold after the second-level screening, they are sorted according to similarity priority and process node priority, and the differences between each similar piece of luggage are marked, triggering manual review, and the target luggage is confirmed by the staff.

[0015] Furthermore, in step S3, the image feature comparison uses a CNN deep learning algorithm to extract feature vectors from the multi-faceted images of luggage and calculates the similarity with the image feature vectors in the retrieval request. The comparison threshold is 80%-95%.

[0016] Furthermore, in step S4, a permission model is set according to the permission requirements of the third-party platform terminal, and the search results are distributed to the corresponding third-party platform terminals in a hierarchical manner through the corresponding distribution protocol. The permission model is a two-layer permission model that combines role-based access control and data-level permissions. The roles include administrator roles, operation roles, and view roles. Data-level permissions are filtered according to the rule that the third-party platform terminal can only access data within its own business scope. The distribution protocol adopts an encrypted distribution protocol that combines HTTPS protocol with JWT token. The unified portal provides a standardized RESTful API interface for bidirectional interaction with the third-party platform terminal. The standardized RESTful API interface includes a search result push interface, a confirmation interface, a query interface, and a feedback interface.

[0017] This application also provides a smart airport baggage cross-platform collaborative retrieval system for performing any of the above-mentioned retrieval methods, including: The business system layer, as the core data source module, consists of multiple heterogeneous business systems within the airport, and is used to provide multi-source data for all stages of the baggage process; The interface adaptation layer, as a data access adaptation module, is connected to the business system layer and is used to configure the corresponding interface adapter to achieve compatible access to data from various heterogeneous business systems. The accessed data includes baggage number, passenger information, check-in time, baggage weight, multi-view images of baggage, baggage transportation status, and process node information. The process node information includes the time and location information of baggage check-in node, security check node, transfer node, loading node, and arrival node. The data standardization layer, as a data processing module, is connected to the interface adaptation layer. It is used to standardize the incoming multi-source data by unifying the data format, mapping fields, cleaning and completing the data, and outputting standardized multi-source data. The standardized multi-source data includes baggage unique identifier, baggage multi-faceted images, passenger information, check-in time, baggage weight, baggage status and process node information. The baggage unique identifier is a mapping of the baggage number, and the baggage status is a mapping of the baggage transportation status. The core layer of the data platform, as a data storage and fusion module, is connected to the data standardization layer. It is used to uniformly store standardized multi-source data and to establish a mapping relationship between the unique identifier of luggage and information of various dimensions using association rule algorithms, so as to complete the association and fusion of multi-dimensional data. The retrieval index layer, as an index management module, is connected to the core layer of the data platform. It is used to receive the fused data and construct a composite B+ tree index. The composite B+ tree index includes a main index with the unique baggage identifier as the primary key and an auxiliary index with other multi-dimensional information as the key, in addition to the unique baggage identifier. The other multi-dimensional information includes passenger information, check-in time, baggage weight, baggage status, process nodes, and image feature quantization identifiers. The image feature quantization identifiers are obtained by vector quantization processing of the image feature vectors extracted from the multi-faceted images of the baggage. The retrieval service layer, as the core module for retrieval and verification, is connected to the retrieval index layer. It is used to receive retrieval requests, parse retrieval conditions, and perform pre-screening of retrieval execution data by calling the composite B+ tree index of the retrieval index layer to obtain preliminary retrieval results. Then, it combines image feature comparison and baggage status consistency verification to verify and filter the preliminary retrieval results, and output accurate collaborative retrieval results. The unified portal layer, serving as a result distribution and interaction module, is connected to the retrieval service layer. It is used to format the collaborative retrieval results and distribute the retrieval results to the third-party platform terminals in a hierarchical manner based on the permission requirements of the third-party platform terminals. The third-party platform terminal, as a terminal display and processing module, is connected to the unified portal layer to receive and display search results and achieve collaborative processing.

[0018] After adopting the above solution, the beneficial effects of the present invention are as follows: 1. This application, by constructing a standardized data platform, enables the data access and standardized processing of heterogeneous business systems such as airport check-in, baggage tracking, irregular baggage, and transfer, breaking down the "data silos" between systems and providing a unified, complete, and reliable data foundation for cross-platform collaborative retrieval.

[0019] 2. This application innovatively integrates multi-dimensional information such as multi-faceted images of baggage, check-in time, baggage weight, baggage status, and process nodes to construct a composite B+ tree retrieval index consisting of a "main index + auxiliary index." Combined with a multi-level filtering mechanism involving pre-screening of business data, image feature comparison, and status consistency verification, it significantly improves the accuracy (up to over 98%) and response speed (reduced to 1-3 seconds), enabling rapid and accurate location of target baggage, especially suitable for emergency scenarios such as lost or misdelivered baggage. Furthermore, in constructing the composite B+ tree retrieval index, image feature vectors are quantized into one-dimensional identifiers to participate in the auxiliary index, solving the problem that multi-dimensional image features cannot be directly used in traditional B+ tree indexes.

[0020] 3. This application uses a unified portal and a two-tiered permission model distribution protocol to securely distribute search results to third-party platform terminals in a hierarchical manner, enabling cross-platform sharing and collaborative processing of search results. This enhances the collaborative capabilities of smart airport baggage management, reduces operational procedures for staff, and optimizes airport operational efficiency and passenger experience.

[0021] 4. The method described in this application is simple and easy to operate, and it is compatible with the upgrade and transformation of existing airport business systems. It does not require large-scale reconstruction of existing systems, thus reducing implementation costs and having broad application prospects. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention.

[0023] Figure 2 This is a system architecture block diagram of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application, and the range values ​​mentioned in this application all include endpoint values.

[0025] Key references Figure 1 This invention provides a smart airport baggage cross-platform collaborative retrieval method, comprising the following steps: S1. Cross-platform data access and standardized processing: Build a standardized data platform to access multi-source data from multiple heterogeneous business systems within the airport, perform standardized processing on the multi-source data, and store the standardized multi-source data in a unified manner in the data platform.

[0026] The data middle platform implements data access from various heterogeneous business systems through interface adapters, and configures corresponding interface adapters for business systems with different architectures. The interface adapters realize compatible access of data from various heterogeneous business systems through means such as RESTful interfaces (a software architecture style based on the HTTP protocol, used for designing interfaces of network applications) and message queues. The heterogeneous business systems include at least the airport's internal check-in system, baggage tracking system, irregular baggage system and transfer system. The accessed data includes baggage number, passenger information, check-in time, baggage weight, multi-surface images of baggage, baggage transportation status, process node information, etc. The baggage number is used as the unique identifier of the baggage, the passenger information includes flight number, seat number, check-in sequence number, passenger name, etc. The multi-surface images of baggage are collected by check-in counter cameras, security check channel cameras and transfer area cameras to ensure coverage of all surfaces of the baggage. The process node information includes the time and position information of the baggage's check-in node, security check node, transfer node, loading node and arrival node. The data middle platform is configured to uniformly store the standardized multi-source data (distributed storage architecture can be adopted for storage), and complete the association and fusion of multi-dimensional data in step S2 through an association rule algorithm. The standardization processing of multi-source data can be completed in the data standardization layer of the retrieval system, and the specific steps are as follows: S1.1 Unification of data format: Convert time-type, numerical-type, text-type, image-type and identifier-type data in the multi-source data into a preset uniform format respectively, define the uniform format of all types of data, to ensure that data output by different systems can be directly fused. Wherein, the time-type data uniformly adopts the format "YYYY-MM-DD HH:MM:SS" (e.g., 2026-04-30 14:30:00), which needs to cover check-in time, process node time, etc.; the numerical-type data uniformly adopts the decimal format, and the baggage weight retains 1 decimal place (unit: kg), such as 20.5kg; the text-type data uniformly adopts UTF-8 encoding, and text fields such as baggage transportation status and passenger name are uniformly processed to remove leading and trailing spaces and special characters (for example, there is a space in "al rea dy loa ded", the space shall be removed and corrected to "already loaded"); the image-type data is uniformly converted to JPG format, and the resolution is uniformly set to 1080P to ensure the uniformity of image feature extraction; identification-type data such as barcodes and RFID tags are uniformly converted into string format, and redundant separators are removed (for example, "123-456-789" is corrected to "123456789").

[0027] S1.2 Field mapping: For different fields of various heterogeneous business systems, establish a unified field mapping comparison table, map different fields of various heterogeneous business systems to standard fields of the data middle platform, realize the one-to-one correspondence between fields of different business systems and standard fields of the data middle platform, and eliminate field ambiguity. The specific mapping rules are as follows: In the check-in system, "passenger ID number" and "baggage number" are mapped to "passenger ID number" and "baggage unique identifier" in the data platform, respectively; in the baggage tracking system, "baggage location" and "baggage transportation status" are mapped to "current node location" and "baggage status" in the data platform, respectively; in the abnormal baggage system, "abnormality type" and "handling record" are mapped to "baggage abnormality type" and "abnormality handling details" in the data platform, respectively; in the transfer system, "transfer flight number" and "transfer node time" are mapped to "transfer flight information" and "transfer node time" in the data platform, respectively; and in all business systems, "baggage weight" and "operator" are uniformly mapped to "actual baggage weight" and "data entry person" in the data platform, ensuring that the same type of data in different business systems corresponds to the same standard field.

[0028] S1.3 Data Cleaning and Completion: Data cleaning algorithms are used to remove redundant and abnormal data. Missing data is filled using a type-specific and scenario-specific completion strategy to avoid affecting the accuracy of subsequent searches. Specifically, for key identifier fields (such as unique baggage identifiers and passenger ID numbers), if missing, they are directly marked as "to be supplemented" and manual review is triggered to ensure uniqueness. For time-related fields (such as check-in time and security check time), if missing, they are filled using the average time of corresponding baggage on the same flight and in the same batch (e.g., if a piece of baggage is missing a security check time, the average security check time of baggage checked in during the same time period on the same flight is used). For text-related fields (such as baggage status), if missing, they are inferred from the status of the preceding and following process nodes (e.g., if baggage has completed check-in but has not yet reached security check, the missing status is filled as "pending security check"). For non-critical auxiliary fields (such as operator contact information), if missing, they are uniformly filled as "not recorded," without affecting the use of core data.

[0029] S2. Multi-source data association, fusion, and retrieval index construction: Extract multi-dimensional information including at least unique baggage identifiers, multi-faceted images of baggage, passenger information, check-in time, baggage weight, baggage status, and process nodes. Use association rule algorithms to establish mapping relationships between unique baggage identifiers and other multi-dimensional information, achieving multi-dimensional data association and fusion. Construct a composite B+ tree index, which includes a primary index with the unique baggage identifier as the primary key and auxiliary indexes with other multi-dimensional information as keys, including passenger information, check-in time, baggage weight, baggage status, process nodes, and quantified image feature identifiers. The multi-dimensional data association and fusion can be completed within the data platform, while the construction of the composite B+ tree index can be performed at the retrieval index layer of the retrieval system.

[0030] The image feature quantization identifier is obtained by vector quantization of the image feature vectors extracted from the multi-faceted images of the luggage. The image feature vectors are obtained by feature extraction from the multi-faceted images of the luggage using a CNN deep learning algorithm. However, since the image feature values ​​extracted by the CNN are multi-dimensional vectors (usually 128-dimensional or 256-dimensional), they cannot directly participate in the one-dimensional B+ tree index. Therefore, this application uses the K-means clustering algorithm to quantize the image feature vectors of all luggage, mapping similar image feature vectors to unique quantization identifiers (i.e., one-dimensional integer codes). Then, the quantization identifier is used as the key of the auxiliary index to construct an image feature-assisted B+ tree index. At the same time, a mapping table of "quantization identifier - original image feature vector" is stored in the data platform. During retrieval, the image feature vectors (if any) in the retrieval request are first subjected to the same quantization process to obtain the corresponding quantization identifier. The unique identifier of the corresponding luggage is retrieved through the image feature-assisted B+ tree index. Then, the complete data is obtained through the main index. Finally, the original image feature vectors are compared (image feature comparison in step S3) for further filtering. This solves the problem that multi-dimensional image feature vectors cannot directly participate in the B+ tree index and ensures the accuracy of image retrieval.

[0031] The association rule algorithm adopts the Apriori algorithm. The core advantage of choosing the Apriori algorithm as the association rule algorithm is that it can efficiently mine the association relationship between multi-dimensional discrete and numerical data, and the computational complexity is moderate. It is suitable for the association mining needs of multi-source data of airport baggage (including textual baggage status and process nodes, numerical weight and time, and eigenvalue image features). It does not require complex hardware support, can quickly realize the association matching of multi-dimensional data, and facilitates the rapid call of subsequent retrieval indexes.

[0032] The steps for establishing a mapping relationship include: S2.1.1 Data Preprocessing: The extracted multi-dimensional information is converted into a format recognizable by the Apriori algorithm. Taking the unique baggage identifier as the core association item, the image feature values ​​(multi-dimensional feature vectors extracted by CNN, converted into string format for easy association), check-in time (in a unified format of "YYYY-MM-DD HH:MM:SS", as a discrete time item), baggage weight (retaining one decimal place, as a numerical association item), baggage status (standardized "check-in", "security check", etc., as discrete status items), and process node information (node ​​time + node location combined into a string format of "node location - time", as a discrete node item) are used as associated auxiliary items of the unique baggage identifier to construct the initial association dataset. S2.1.2 Frequent Itemset Mining: Frequent itemsets in the initial associated dataset are mined using the Apriori algorithm. A minimum support is set (preferably 5% based on the amount of airport baggage data, meaning that a certain type of association is considered valid if its frequency of occurrence in the total dataset is not less than 5%). Core frequent itemsets such as "baggage unique identifier - image feature value" and "baggage unique identifier - process node information" are prioritized for mining. Then, other multi-dimensional cross-frequent itemsets (such as "baggage unique identifier - image feature value - baggage weight" and "baggage unique identifier - process node information - baggage status") are gradually mined. Invalid and low-frequency association items are removed, and core associations related to baggage retrieval and status tracking are retained. S2.1.3 Association Rule Generation and Mapping Relationship Consolidation: Based on the mined frequent itemsets, association rules that meet the minimum confidence level (preferably set to 85% to ensure the reliability of the association relationship) are generated. The core rule is "Unique Baggage Identifier → [Image Feature Value, Check-in Time, Baggage Weight, Baggage Status, Process Node Information]". That is, through a single unique baggage identifier, all multi-dimensional information corresponding to that baggage can be linked. This association rule is consolidated into the association mapping table of the data platform, establishing a one-to-one correspondence between the unique baggage identifier and each dimension of information. The mapping table can be updated synchronously in real time (i.e., when the process node or status of a baggage changes, the mapping relationship is automatically updated), ensuring the relevance and timeliness of multi-dimensional data.

[0033] Preferably, to avoid redundancy in association rules, a pruning strategy is adopted to remove duplicate and invalid association rules. At the same time, a dynamic update mechanism is set up so that when new business system data or new baggage dimension information (such as baggage material) is added, the Apriori algorithm is called again to perform frequent itemset mining and update the association mapping table. This ensures that the mapping relationship always adapts to the needs of multi-source data fusion and provides a reliable association foundation for the construction of subsequent collaborative retrieval indexes.

[0034] Based on the multi-dimensional data fused using the Apriori algorithm, a B+ tree index is constructed as a collaborative baggage retrieval index. The index keywords include baggage number, passenger information, image feature quantification identifiers, check-in time, baggage weight, baggage status, and process nodes, consistent with the core related items in the association mapping table. It supports multi-condition combined retrieval. The primary index uses baggage number (unique baggage identifier) ​​as the main keyword, while secondary indexes use other information as keywords. The leaf nodes of each secondary index do not store complete data, only the corresponding unique baggage identifier and a pointer to the primary index. After retrieving the corresponding unique baggage identifier through the secondary index, the primary index can be quickly linked through the primary index pointer to obtain complete multi-dimensional data, thus enabling multi-condition combined retrieval. The specific steps for constructing the composite B+ tree index include: S2.2.1 Store the physical storage address of the multi-dimensional data corresponding to the unique identifier of each baggage into the leaf node of the main B+ tree to complete the construction and balancing adjustment of the main index; S2.2.2. Construct multiple auxiliary B+ tree indexes using passenger information, check-in time, baggage weight, baggage status, process node, and image feature quantification identifier as keywords. The leaf nodes of each auxiliary index store the unique baggage identifier associated with the corresponding keyword and the pointer to the main index, thereby realizing the association with the main index. S2.2.3 Establish an index update mechanism. When multi-source data (such as process nodes, luggage status, etc.) in the data platform changes, the associated mapping table is updated synchronously, and the index is dynamically adjusted (such as insertion, deletion, balancing, etc.) to ensure the consistency between the index and the data. S2.2.4. Perform performance testing on the completed B+ tree index and optimize the index structure (e.g., adjust the order of the B+ tree and control the number of leaf nodes) to ensure that the response time for multi-condition combined retrieval is controlled within 1-3 seconds. Finally, complete the construction of the luggage collaborative retrieval index. This index can be directly adapted to the retrieval process in step S3, and the auxiliary index keywords can be directly used as retrieval conditions.

[0035] S3. Collaborative Search and Result Verification: Receive baggage search requests submitted by users (such as airport staff or airline staff). These requests include one or more search criteria such as baggage number, passenger name, baggage characteristics (e.g., baggage color, baggage brand), baggage weight, check-in time, and baggage status. Based on the search criteria in the request, the composite B+ tree index is used to perform pre-screening of the search execution data to obtain preliminary search results. Then, the preliminary search results are verified and filtered by combining image feature comparison and baggage status consistency verification to output accurate collaborative search results.

[0036] During the verification process, the initial search results are first checked for consistency in status, and then the image feature comparison is performed. The consistency in status check is used to verify whether the current status of the baggage in the initial search results matches the process node information. If they do not match, the result is removed. For example, if the baggage status is "loaded", it is checked whether there is a "loading node" record in its process nodes, and whether the loading time is later than the check-in and security check time. If not, the result is removed.

[0037] Image feature comparison and verification includes the following steps: S3.1 First-level filtering: For search results with inconsistent statuses that have already been removed, cross-validation is performed on the preliminary search results based on non-image search conditions in the search request (such as passenger information, check-in time, baggage weight, baggage brand / size). For example, if the search request includes passenger information (such as passenger name, ID number), baggage that matches the passenger information is prioritized; if it includes check-in time and baggage weight, baggage with check-in time within ±30 minutes of the search range and weight deviation ≤1kg is filtered out; if it includes baggage brand and size, precise matching is performed using baggage feature information stored in the data platform, and baggage that does not meet the cross-validation conditions is removed, which greatly reduces the workload of subsequent image feature comparison and improves filtering efficiency. S3.2 Secondary Filtering: After the results of the primary filtering, image feature comparison is performed to extract the feature vector of the luggage image in the retrieval request and compare it accurately with the feature vector of the luggage image after the primary filtering. A comparison threshold is set to remove luggage with similarity lower than the comparison threshold. The similarity with the image feature vector in the retrieval request can be calculated using a CNN deep learning algorithm. The comparison threshold is 80%-95%.

[0038] S3.3 Three-level screening: If only one piece of luggage remains after the second-level screening, it is determined to be the target luggage, and the match is successful. It can be sent to the third-party platform terminal. If there are still two or more pieces of luggage with similarity higher than the comparison threshold after the second-level screening, they are sorted according to similarity priority and process node priority. The higher the image feature similarity and the higher the correlation between the current process node and the search request (e.g., when searching for "already loaded" luggage, the luggage with the most recent loading time is prioritized), the higher the ranking. At the same time, the differences between each similar piece of luggage are marked (e.g., minor damage to luggage, differences in accessories, minor weight deviations, etc.), triggering manual review. The staff confirms the target luggage, and only after the target luggage is confirmed is it distributed to the third-party platform terminal to avoid blindly distributing all similar luggage and causing chaos in the handling.

[0039] If a unique target baggage cannot be identified after the above multi-level screening, all similar baggage (similar baggage with marked differences and similarities) will be distributed to the corresponding third-party platform terminal along with a prompt that "the target baggage needs to be manually confirmed". This ensures that the relevant staff can complete the final screening based on the actual scenario, which not only ensures search efficiency but also avoids missing the target baggage. The final output is accurate collaborative search results, laying the foundation for subsequent result distribution.

[0040] S4. Search Result Distribution: The collaborative search results are formatted and can be displayed in a visual format combining tables and graphics. Specific details may include basic baggage information (baggage number, passenger information, weight), multi-view images of the baggage, process trajectory (time and location at each node), and current status. The search results are then distributed hierarchically to corresponding third-party platform terminals through a unified portal. This unified portal uses a standardized interface design to connect with third-party platform terminals such as airport operations management platforms, ground service platforms, airline systems, and passenger service platforms. Based on the permission requirements of the third-party platform terminals, hierarchical distribution of search results is implemented. For example, airline terminals can only view baggage search results for their own flights, ensuring data security.

[0041] Specifically, based on the permission requirements of third-party platform terminals, a permission model is set up to distribute search results to the corresponding third-party platform terminals in a hierarchical manner through the corresponding distribution protocol, thereby achieving hierarchical control. The permission model is a two-tier permission model combining role-based access control (RBAC) and data-level permissions (i.e., the RBAC two-tier permission model). Roles are divided into three core categories based on the business scenarios of the third-party platform terminals, including: Administrator role: Corresponding to the business scenario of the airport operation management platform, with the highest level of authority, including the ability to view all baggage search results, modify permission configurations, and view search logs; Operational Role: Corresponds to the business scenarios of the ground service platform and airline system. They can only view baggage search results related to their own business (e.g., ground service terminals can only view baggage under their responsibility in this area, and airline terminals can only view baggage of their own flights). They can submit search requests and receive processing feedback. Viewing Role: Corresponds to the passenger service platform terminal. Can only view the baggage search results associated with themselves (passenger identity information needs to be verified), and has no processing authority.

[0042] The dataset permissions, based on role-based permissions, perform data-level filtering on search results. The core rule is that "third-party platform terminals can only access data within their own business scope." Specifically: the airport operations management platform can access baggage data for all flights and all areas; the ground service platform has permissions divided by area, only accessing baggage data for its own area (e.g., Terminal 1, transit area); the airline system has permissions divided by flight number, only accessing baggage data for flights operated by the airline itself; and the passenger service terminal has permissions divided by passenger ID number, only accessing baggage data associated with that passenger. Permission configurations are stored in the unified portal's permission management module and support dynamic adjustments (e.g., adding airlines, adjusting regional permissions).

[0043] The distribution protocol employs an encrypted distribution protocol combining HTTPS and JWT tokens to ensure the security and reliability of the retrieval result transmission. The specific protocol specifications are as follows: Transmission protocol: HTTPS protocol is used for data transmission, and SSL / TLS encrypted channel is used to prevent data from being stolen or tampered with during transmission; Identity authentication: JWT (JSON Web Token) tokens are used for identity verification. When a third-party platform terminal accesses the unified portal, it needs to submit a unique identifier (terminal ID) and a key. After the unified portal verifies the identity, it generates a JWT token (valid for 2 hours). The third-party platform terminal must include this token when sending subsequent distribution requests; otherwise, the distribution will be rejected. Data format: The search results are encapsulated in JSON format, including fields such as result identifier, baggage information, distribution time, and permission identifier, which facilitates parsing and display by third-party platform terminals; Anomaly Handling: The protocol has a built-in retransmission mechanism. If the third-party platform terminal fails to receive the search results (e.g., due to network interruption), the unified portal will automatically retransmit within 10 minutes. After three failed retransmissions, the anomaly log will be recorded and the administrator will be notified to ensure distribution reliability.

[0044] The unified portal interacts with third-party platform terminals using a bidirectional interaction mode of "standardized interface calls + real-time push," achieving efficient hierarchical distribution of search results. Specifically, the unified portal provides standardized RESTful API interfaces for bidirectional interaction with third-party platform terminals. These standardized RESTful API interfaces include interfaces for pushing search results, receiving confirmation, querying, and providing feedback. All interface parameters and return formats are standardized and compatible with third-party platforms of different architectures. The distribution process is as follows: After the collaborative search is completed, the unified portal's permission management module filters the search results that the terminal can access based on the terminal's role and data-level permissions. Then, the unified portal pushes the encrypted search results to the third-party platform terminal via the push interface, carrying a JWT token, through the HTTPS protocol. After receiving the results, the third-party platform terminal returns confirmation information to the unified portal through the confirmation interface. After receiving the confirmation information, the unified portal records the distribution log (including distribution time, terminal ID, and result identifier). If the unified portal does not receive confirmation information, it triggers a retransmission mechanism.

[0045] Third-party platform terminals can actively query related search results through the query interface, or submit handling feedback to the unified portal through the feedback interface (such as "target baggage found" or "further screening required"). After receiving the feedback, the unified portal will update the baggage status and search logs synchronously to achieve closed-loop management of collaborative handling.

[0046] In this application, the above method enables cross-platform data collaboration among multiple airport business systems, reduces baggage retrieval response time to 1-3 seconds, improves retrieval accuracy to over 98%, significantly enhances airport baggage inquiry and collaborative handling capabilities, reduces handling time for lost or mishandled baggage, and optimizes passenger experience.

[0047] To more clearly illustrate the practical application process of the method in this application, and in conjunction with the common airport scenario of "baggage misdelivery retrieval and collaborative handling," a complete embodiment is provided for further explanation, as follows: Scenario: At Terminal 2 of an airport, passenger Zhang boarded airline A's flight MU1234 (Beijing → Shanghai). During check-in, he checked in a black 28-inch suitcase (weighing 22.3kg, brand Samsonite, with a minor scratch on the side). After the flight arrived at Shanghai Pudong Airport, Zhang could not find his luggage in the baggage claim area and sought assistance from airport ground staff. The ground staff initiated the cross-platform collaborative baggage retrieval process described in this invention, relying on a data platform, a composite B+ tree search index, and a unified portal to conduct retrieval and collaborative processing.

[0048] Retrieval and Processing: Ground service personnel submit retrieval requests through a unified portal, entering search criteria such as passenger name, ID number, baggage characteristics, check-in time, baggage weight, and flight number. The retrieval service layer parses the retrieval request, quantifies the baggage characteristics, matches them with a composite B+ tree auxiliary index, and quickly obtains three preliminary retrieval results in conjunction with the main index. First, a consistency check is performed. Then, according to the "first-level filtering" rule, cross-validation is performed using non-image criteria such as passenger information and baggage weight, eliminating two mismatched records, leaving one target baggage record (baggage number 3379235689). Subsequently, CNN image feature comparison confirms that the baggage image feature similarity is 96% (with a threshold of 90%), exceeding the 90% threshold, confirming it as the target baggage. Based on the retrieval results, it is confirmed that the baggage has arrived at conveyor belt 3 in the baggage sorting area of ​​Terminal 1 at Shanghai Pudong Airport, and its status is "awaiting retrieval."

[0049] Tiered distribution and closed-loop processing: The unified portal, based on the RBAC two-layer permission model, matches the permissions of the ground service terminals at Shanghai Pudong Airport T1 and the A-Airline system terminals. Using HTTPS+JWT encryption, it pushes the visual search results (including multi-view images of baggage and process tracking) to the two third-party platform terminals. Ground service personnel locate the baggage based on the search results and submit processing feedback to the unified portal through their terminals. The A-Airline system simultaneously updates the baggage status. After Mr. Zhang verifies his identity, he retrieves his baggage. The unified portal archives the entire process log, completing the collaborative processing closed loop.

[0050] Key references Figure 2This application also provides a smart airport baggage cross-platform collaborative retrieval system for performing the above-mentioned retrieval method, including: The business system layer, as the core data source module, consists of multiple heterogeneous business systems within the airport, such as the check-in system, baggage tracking system, irregular baggage system, and transfer system. It is used to provide raw data for each stage of the baggage process, namely the multi-source data accessed in step S1, including baggage number, passenger information, check-in time, baggage weight, multi-view images of baggage, baggage transportation status, process node information, etc. It is the data source foundation of the entire retrieval system, ensuring that the data covers the entire business scenario of baggage management. The interface adaptation layer, as a data access adaptation module, is connected to the business system layer. Its main function is to configure corresponding interface adapters (such as RESTful interface adapters and message queue adapters) for the architectural differences of various heterogeneous business systems, so as to achieve compatible access of data from different business systems, solve the problem of data interoperability between heterogeneous systems, and unify the access of raw data from various business systems to the subsequent data processing module, ensuring the stability and compatibility of data access. The data standardization layer, as a data processing module, is connected to the interface adaptation layer and is used to standardize the incoming raw data. Specifically, it includes operations such as abnormal data removal, redundant data deletion, missing data completion, data format unification, and field mapping in steps S1.1-S1.3. This transforms heterogeneous and messy raw data into standardized data with unified standards, providing clean and unified data support for subsequent data fusion and index construction. After processing, it outputs standardized multi-source data, which includes at least baggage unique identifier, baggage multi-faceted images, passenger information, check-in time, baggage weight, baggage status, and process node information. The core layer of the data platform, as the data storage and fusion module, also known as the data platform mentioned above, is connected to the data standardization layer. It is used to uniformly store the standardized multi-source data and to establish the mapping relationship between the unique identifier of luggage and information of various dimensions using the Apriori association rule algorithm. It stores the "quantized identifier-original image feature vector" mapping table to complete the association and fusion of multi-dimensional data, providing data support for the construction of retrieval indexes and retrieval services. It is the data core of the system. The retrieval index layer, acting as an index management module, connects to the core layer of the data platform. It receives the fused data transmitted from the core layer and constructs, dynamically updates, and balances a composite B+ tree collaborative retrieval index. This composite B+ tree index includes a primary index with unique baggage identifiers as the main keyword, and auxiliary indexes with other multi-dimensional information as keywords. This other multi-dimensional information includes at least passenger information, check-in time, baggage weight, baggage status, process nodes, and image feature quantization identifiers. The image feature quantization identifiers are obtained by vector quantization of image feature vectors extracted from multi-faceted images of baggage. After the index is established, the completed retrieval index is stored, providing efficient index support for the retrieval service and improving retrieval response speed. The retrieval service layer, as the core module for retrieval and verification, is connected to the retrieval index layer. It is used to receive retrieval requests submitted by users, parse retrieval conditions, and perform pre-screening of retrieval execution data by calling the composite B+ tree index of the retrieval index layer to obtain preliminary retrieval results. Then, it combines image feature comparison and baggage status consistency verification to verify and filter the preliminary retrieval results, remove invalid and erroneous results, and output accurate collaborative retrieval results. This layer is the core execution module of the entire retrieval process. The unified portal layer, serving as the result distribution and interaction module (i.e., the same portal mentioned above), connects to the retrieval service layer. It is used to format collaborative retrieval results and, based on the permission requirements of third-party platform terminals (specifically, using a two-layer RBAC permission model and HTTPS+JWT encryption protocol), distributes retrieval results to third-party platform terminals in a tiered manner. Simultaneously, it provides standardized interfaces to enable bidirectional interaction with third-party platform terminals (retrieval result query and processing feedback), records distribution logs, and ensures the security, efficiency, and closed-loop management of retrieval result distribution.

[0051] The third-party platform terminal, as a terminal display and processing module, is connected to the unified portal layer to receive and display the search results distributed by the unified portal layer. Specifically, it can display basic baggage information, multi-faceted images, process trajectory and current status for staff or passengers to view. At the same time, it submits processing feedback to the unified portal layer to realize collaborative baggage processing and is the terminal carrier for the search results.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0053] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart airport baggage cross-platform collaborative retrieval method, characterized in that, Includes the following steps: S1. Cross-platform data access and standardized processing: Construct a standardized data platform to access multi-source data from multiple heterogeneous business systems within the airport. The accessed data includes baggage number, passenger information, check-in time, baggage weight, multi-view images of baggage, baggage transportation status, and process node information. The process node information includes the time and location information of baggage check-in node, security check node, transfer node, loading node, and arrival node. Then, the multi-source data is standardized by unifying the data format, mapping fields, cleaning and completing data. The standardized multi-source data is then uniformly stored in the data platform. S2. Multi-source data association, fusion, and retrieval index construction: Multi-dimensional information is extracted from standardized multi-source data, including unique baggage identifiers, multi-faceted images of baggage, passenger information, check-in time, baggage weight, baggage status, and process nodes. The unique baggage identifier is a mapping to the baggage number, and the baggage status is a mapping to the baggage transportation status. Then, an association rule algorithm is used to establish the mapping relationship between the unique baggage identifier and other multi-dimensional information, realizing the association and fusion of multi-dimensional data, and constructing a composite B+ tree index. This composite B+ tree index includes a primary index with the unique baggage identifier as the main key, and auxiliary indexes with other multi-dimensional information as keys, including passenger information, check-in time, baggage weight, baggage status, process nodes, and image feature quantization identifiers. The image feature quantization identifiers are obtained by vector quantization processing of the image feature vectors extracted from the multi-faceted images of baggage. S3. Collaborative retrieval and result verification: Receive baggage retrieval requests containing at least one retrieval condition, use the composite B+ tree index to perform retrieval execution data pre-screening to obtain preliminary retrieval results, and then combine image feature comparison and baggage status consistency verification to verify and filter the preliminary retrieval results, and output accurate collaborative retrieval results. S4. Search Result Distribution: The collaborative search results are formatted and distributed to the corresponding third-party platform terminals through a unified portal according to their permission requirements.

2. The smart airport baggage cross-platform collaborative retrieval method as described in claim 1, characterized in that: In step S1, the data platform accesses data with various heterogeneous business systems through corresponding interface adapters. The heterogeneous business systems include the airport's internal check-in system, baggage tracking system, irregular baggage system, and transfer system.

3. The smart airport baggage cross-platform collaborative retrieval method as described in claim 2, characterized in that: In step S1, the standardization process for multi-source data includes the following steps: S1.1 Data Format Unification: Convert time-based, numerical, text-based, image-based, and label-based data from multiple sources into a preset unified format; S1.2 Field Mapping: For the different fields of various heterogeneous business systems, establish a unified field mapping table to map the different fields of various heterogeneous business systems to the standard fields of the data platform; S1.3 Data Cleaning and Completion: Data cleaning algorithms are used to remove redundant and abnormal data, and missing data is filled in using completion strategies based on type and scenario.

4. The smart airport baggage cross-platform collaborative retrieval method as described in claim 3, characterized in that: In step S2, the association rule algorithm is the Apriori algorithm, and the steps for establishing the mapping relationship include: S2.1.1 Data Preprocessing: The extracted multi-dimensional information is converted into a format recognizable by the Apriori algorithm. Taking the unique baggage identifier as the core associated item, image feature values, check-in time, baggage weight, baggage status and process node information are used as associated auxiliary items of the unique baggage identifier to construct an initial associated dataset. S2.1.2 Frequent Itemset Mining: Frequent itemsets in the initial associated dataset are mined using the Apriori algorithm. A minimum support is set to mine frequent itemsets between baggage unique identifiers and multi-dimensional information. S2.1.3, Association rule generation and mapping relationship solidification: Based on the frequently mined itemsets, association rules that meet the minimum confidence level are generated, and the association rules are solidified into the association mapping table of the data platform to establish a one-to-one mapping relationship between the unique identifier of luggage and information of each dimension.

5. The smart airport baggage cross-platform collaborative retrieval method as described in claim 4, characterized in that: In step S2, the steps for constructing the composite B+ tree index include: S2.2.1 Store the physical storage address of the multi-dimensional data corresponding to the unique identifier of each baggage into the leaf node of the main B+ tree to complete the construction and balancing adjustment of the main index; S2.2.

2. Construct multiple auxiliary B+ tree indexes using passenger information, check-in time, baggage weight, baggage status, process node, and image feature quantification identifier as keywords. The leaf nodes of each auxiliary index store the unique baggage identifier associated with the corresponding keyword and the pointer to the main index, thereby realizing the association with the main index. S2.2.3 Establish an index update mechanism. When multi-source data in the data platform changes, the associated mapping table is updated synchronously, and the index is dynamically adjusted to ensure the consistency between the index and the data. S2.2.4 Perform performance testing on the completed B+ tree index, optimize the index structure, and finally complete the construction of the composite B+ tree index.

6. The smart airport baggage cross-platform collaborative retrieval method as described in claim 4, characterized in that: In step S2, the vector quantization processing of the image feature vectors includes: using the K-means clustering algorithm to cluster the image feature vectors of all luggage, mapping similar image feature vectors to unique one-dimensional integer codes as image feature quantization identifiers, and storing a mapping table between the quantization identifiers and the original image feature vectors in the data platform.

7. The smart airport baggage cross-platform collaborative retrieval method as described in claim 1, characterized in that: In step S3, the search conditions include at least one of baggage number, passenger information, baggage characteristics, baggage weight, check-in time, and baggage status. During verification, the initial search results are first checked for consistency of status, and then image feature comparison is performed. The consistency of status is used to verify whether the current status of the baggage in the initial search results matches the process node information. If they do not match, the result is discarded. The image feature comparison verification includes the following steps: S3.1 First-level filtering: For search results with inconsistent status that have been removed, cross-validate the preliminary search results based on non-image search conditions in the search request to narrow down the search scope; S3.2 Secondary Filtering: For the results after primary filtering, perform image feature comparison, extract the feature vector of the luggage image in the search request, compare it with the feature vector of the luggage image after primary filtering, calculate the similarity, set the comparison threshold, and remove luggage with similarity lower than the comparison threshold. S3.3 Three-level screening: If only one piece of luggage remains after the second-level screening, it is determined to be the target luggage; if there are still two or more pieces of luggage with similarity higher than the comparison threshold after the second-level screening, they are sorted according to similarity priority and process node priority, and the differences between each similar piece of luggage are marked, triggering manual review, and the target luggage is confirmed by the staff.

8. The smart airport baggage cross-platform collaborative retrieval method as described in claim 7, characterized in that: In step S3, the image feature comparison uses a CNN deep learning algorithm to extract feature vectors from the multi-faceted images of luggage and calculates the similarity with the image feature vectors in the retrieval request. The comparison threshold is 80%-95%.

9. The smart airport baggage cross-platform collaborative retrieval method as described in claim 1, characterized in that: In step S4, a permission model is set according to the permission requirements of the third-party platform terminal, and the search results are distributed to the corresponding third-party platform terminals in a hierarchical manner through the corresponding distribution protocol. The permission model is a two-tiered model combining role-based access control and data-level permissions. Roles include administrator, operator, and viewer roles. Data-level permissions are filtered based on rules that allow third-party platform terminals to access data only within their own business scope. The distribution protocol uses an encrypted distribution protocol combining HTTPS and JWT tokens. The unified portal provides standardized RESTful API interfaces for bidirectional interaction with third-party platform terminals. These standardized RESTful API interfaces include interfaces for pushing search results, receiving confirmation, querying, and providing feedback.

10. A smart airport baggage cross-platform collaborative retrieval system, used to execute the retrieval method according to any one of claims 1-9, characterized in that, include: The business system layer, as the core data source module, consists of multiple heterogeneous business systems within the airport, and is used to provide multi-source data for all stages of the baggage process; The interface adaptation layer, as a data access adaptation module, is connected to the business system layer and is used to configure the corresponding interface adapter to achieve compatible access to data from various heterogeneous business systems. The accessed data includes baggage number, passenger information, check-in time, baggage weight, multi-view images of baggage, baggage transportation status, and process node information. The process node information includes the time and location information of baggage check-in node, security check node, transfer node, loading node, and arrival node. The data standardization layer, as a data processing module, is connected to the interface adaptation layer. It is used to standardize the incoming multi-source data by unifying the data format, mapping fields, cleaning and completing data, and outputting standardized multi-source data. The standardized multi-source data includes baggage unique identifier, baggage multi-faceted images, passenger information, check-in time, baggage weight, baggage status and process node information. The baggage unique identifier is a mapping of the baggage number, and the baggage status is a mapping of the baggage transportation status. The core layer of the data platform, as a data storage and fusion module, is connected to the data standardization layer. It is used to uniformly store standardized multi-source data and to establish a mapping relationship between the unique identifier of luggage and information of various dimensions using association rule algorithms, so as to complete the association and fusion of multi-dimensional data. The retrieval index layer, as an index management module, is connected to the core layer of the data platform. It is used to receive the fused data and construct a composite B+ tree index. The composite B+ tree index includes a main index with baggage unique identifier as the primary key and an auxiliary index with other multi-dimensional information as the key, in addition to baggage unique identifier. Other multi-dimensional information includes passenger information, check-in time, baggage weight, baggage status, process nodes, and image feature quantization identifier. The image feature quantization identifier is obtained by vector quantization processing of the image feature vector extracted from the multi-faceted images of baggage. The retrieval service layer, as the core module for retrieval and verification, is connected to the retrieval index layer. It is used to receive retrieval requests, parse retrieval conditions, and perform pre-screening of retrieval execution data by calling the composite B+ tree index of the retrieval index layer to obtain preliminary retrieval results. Then, it combines image feature comparison and baggage status consistency verification to verify and filter the preliminary retrieval results, and output accurate collaborative retrieval results. The unified portal layer, serving as a result distribution and interaction module, is connected to the retrieval service layer. It is used to format the collaborative retrieval results and distribute the retrieval results to the third-party platform terminals in a hierarchical manner based on the permission requirements of the third-party platform terminals. The third-party platform terminal, as a terminal display and processing module, is connected to the unified portal layer to receive and display search results and achieve collaborative processing.

Images