Periodic Reporting for period 1 - BAG-INTEL (Intelligent system for improved efficiency and effectiveness Intelligent system for improved efficiency and effectiveness of the customs control of passenger baggage from international flight arrivals)
Okres sprawozdawczy: 2023-09-01 do 2025-02-28
The BAG-INTEL project aims to enhance customs controls of incoming travellers’ baggage at inland border airports. Currently, customs inspections rely primarily on officers’ experience and their assessment of travellers' behaviour and luggage appearance. BAG-INTEL introduces an AI-supported risk assessment process to help customs officers to focus manual inspections only on flagged high-risk baggage. This will increase contraband detection while reducing time spent on inspecting bags that do not contain contraband.
The overall aim of BAG-INTEL is to provide robust AI-based data utilisation and decision support tools to enhance customs operations, increasing the effectiveness and efficiency of incoming baggage controls at inland border airports, without increasing the number of human resources needed for this operation.
This aim addresses the challenge of maintaining effective and efficient customs controls of incoming passenger baggage amid the substantial growth in air traveller numbers at inland border airports, despite limited human customs resources. To achieve this, the project introduces a disruptive solution based on three main components: (i) camera/AI-based end-to-end continuous reidentification of luggage, (ii) AI-powered recognition of contraband in the scanning image of the luggage, and (iii) a digital twin for BAG-INTEL system visualisation and performance optimisation for the operational context of an airport.
The overall aim of BAG-INTEL is to provide robust AI-based data utilisation and decision support tools to enhance customs operations, increasing the effectiveness and efficiency of incoming baggage controls at inland border airports, without increasing the number of human resources needed for this operation.
This aim addresses the challenge of maintaining effective and efficient customs controls of incoming passenger baggage amid the substantial growth in air traveller numbers at inland border airports, despite limited human customs resources. To achieve this, the project introduces a disruptive solution based on three main components: (i) camera/AI-based end-to-end continuous reidentification of luggage, (ii) AI-powered recognition of contraband in the scanning image of the luggage, and (iii) a digital twin for BAG-INTEL system visualisation and performance optimisation for the operational context of an airport.
Sure! Here’s the reduced version, keeping all the key information while cutting about 250 characters:
---
**The BAG-INTEL knowledge base and knowledge building**
The first version of the BAG-INTEL knowledge base was built and populated with a domain ontology and knowledge on risk indicators acquired from domain experts among end-user partners.
**AI-based analysis of external data and knowledge**
A model for deriving risk levels from external data and indicator knowledge has been developed. A first AI solution is undergoing further development and testing.
**Luggage risk assessment, and identification and reidentification of luggage to be inspected**
The first version of algorithms for aggregating risk indicators from sensors and assessing reidentification risk has been developed. A preliminary mock-up of the Customs Intrusive Inspection Experience Feedback Database has been built for integration into BAG-INTEL.
**AI-based detection and recognition of contraband in scanning images**
The project’s X-ray/CT scanner collected scanning data from drugs provided by a customs partner. The AI algorithm for recognising drugs in luggage is being trained with the data.
**Cameras for AI-vision-based end-to-end continuous reidentification of luggage**
Cameras were set up in Billund Airport to collect images from flight arrivals for AI training. Optimal camera positions and angles have been determined at two other use case airports for upcoming installation.
**Visualisation for real-time decision support**
A first version of the real-time customs decision support system and baggage pipeline has been developed.
**Digital twin and visualisation for customs management decision support**
A digital twin of Thessaloniki Airport’s non-Schengen arrival area and an initial decision support system were developed, enabling customs to visualise traveler and baggage flows. A second model for a large airport is under development.
**Machine learning for ongoing improvement of decision making**
A machine learning system is being developed to use customs officers’ feedback from inspected bags to continuously refine BAG-INTEL’s knowledge during operations.
**Hierarchical, Multi-Cloud Architecture for the IoT-Edge-Cloud Continuum**
A novel multi-cloud architecture has been designed to serve BAG-INTEL’s needs, enabling use of IoT imaging devices while protecting critical data with an airport-located Cloud. It pushes complex, time-critical computation to the edge and IoT, ensuring security and privacy.
---
**The BAG-INTEL knowledge base and knowledge building**
The first version of the BAG-INTEL knowledge base was built and populated with a domain ontology and knowledge on risk indicators acquired from domain experts among end-user partners.
**AI-based analysis of external data and knowledge**
A model for deriving risk levels from external data and indicator knowledge has been developed. A first AI solution is undergoing further development and testing.
**Luggage risk assessment, and identification and reidentification of luggage to be inspected**
The first version of algorithms for aggregating risk indicators from sensors and assessing reidentification risk has been developed. A preliminary mock-up of the Customs Intrusive Inspection Experience Feedback Database has been built for integration into BAG-INTEL.
**AI-based detection and recognition of contraband in scanning images**
The project’s X-ray/CT scanner collected scanning data from drugs provided by a customs partner. The AI algorithm for recognising drugs in luggage is being trained with the data.
**Cameras for AI-vision-based end-to-end continuous reidentification of luggage**
Cameras were set up in Billund Airport to collect images from flight arrivals for AI training. Optimal camera positions and angles have been determined at two other use case airports for upcoming installation.
**Visualisation for real-time decision support**
A first version of the real-time customs decision support system and baggage pipeline has been developed.
**Digital twin and visualisation for customs management decision support**
A digital twin of Thessaloniki Airport’s non-Schengen arrival area and an initial decision support system were developed, enabling customs to visualise traveler and baggage flows. A second model for a large airport is under development.
**Machine learning for ongoing improvement of decision making**
A machine learning system is being developed to use customs officers’ feedback from inspected bags to continuously refine BAG-INTEL’s knowledge during operations.
**Hierarchical, Multi-Cloud Architecture for the IoT-Edge-Cloud Continuum**
A novel multi-cloud architecture has been designed to serve BAG-INTEL’s needs, enabling use of IoT imaging devices while protecting critical data with an airport-located Cloud. It pushes complex, time-critical computation to the edge and IoT, ensuring security and privacy.
BAG-INTEL research papers, which have already been published or are currently under review include:
Processing heterogeneous data sources for risk management by knowledge graph databases
This research has developed new efficient data structures and querying methods for managing knowledge for risk assessment.
Risk-based Customs’ Intrusive Inspection Decision Making
This researched and developed a method for customs risk assessment through utilisation of customs experience information acquired from the end-user partners. The algorithm applies a mix of probabilistic statistics (deductive approach) and Analytic Hierarchical Process (AHP) classification theory (inductive approach).
Hierarchical Multi-Cloud IoT-Edge-Cloud Architecture for Enhanced Airport Security and Operations
This work developed a scalable and secure solution for the various challenges of airport environments through the implementation of a hierarchical, multi-cloud architecture. The outcome represents a significant advancement in the integration of IoT, edge computing, and cloud technologies for airport security operations.
Fuzzy Risk Level Estimation Framework for Flight Contraband
This paper introduces a fuzzy risk assessment mechanism for airport routes, leveraging expert knowledge and incorporating a graph-based representation of multiple information sources and past experiences. The research provides a valuable foundation for developing new similarity measures and a fuzzy logic-based approach to route risk assessment in a security area that have not previously been extensively researched.
Key needs to ensure further uptake and success
For most of the aforementioned research and ongoing studies, further investigations– involving tests with realistic data and within the integrated BAG-INTEL system–will be conducted during the second reporting period and will include demonstrations at use case airports.
Additionally, the necessary steps to bring the BAG-INTEL solution to market will be analysed and addressed during the second half of the project.
Processing heterogeneous data sources for risk management by knowledge graph databases
This research has developed new efficient data structures and querying methods for managing knowledge for risk assessment.
Risk-based Customs’ Intrusive Inspection Decision Making
This researched and developed a method for customs risk assessment through utilisation of customs experience information acquired from the end-user partners. The algorithm applies a mix of probabilistic statistics (deductive approach) and Analytic Hierarchical Process (AHP) classification theory (inductive approach).
Hierarchical Multi-Cloud IoT-Edge-Cloud Architecture for Enhanced Airport Security and Operations
This work developed a scalable and secure solution for the various challenges of airport environments through the implementation of a hierarchical, multi-cloud architecture. The outcome represents a significant advancement in the integration of IoT, edge computing, and cloud technologies for airport security operations.
Fuzzy Risk Level Estimation Framework for Flight Contraband
This paper introduces a fuzzy risk assessment mechanism for airport routes, leveraging expert knowledge and incorporating a graph-based representation of multiple information sources and past experiences. The research provides a valuable foundation for developing new similarity measures and a fuzzy logic-based approach to route risk assessment in a security area that have not previously been extensively researched.
Key needs to ensure further uptake and success
For most of the aforementioned research and ongoing studies, further investigations– involving tests with realistic data and within the integrated BAG-INTEL system–will be conducted during the second reporting period and will include demonstrations at use case airports.
Additionally, the necessary steps to bring the BAG-INTEL solution to market will be analysed and addressed during the second half of the project.