VRR Aero: Cargo Container Defect Detection & Inspections with Computer Vision

VRR Aero, a global leader in Unit Load Devices (ULDs) for air cargo, is committed to ensuring safety and operational efficiency in aviation logistics. ULDs are essential for securely transporting cargo, from commercial goods to sensitive materials. However, these containers endure constant handling, exposure to harsh conditions and operational wear, leading to potential structural damage.

Ensuring that only airworthy containers are used is a safety-critical task. Traditionally, manual visual inspections are conducted to detect damages such as holes, cracks, and worn straps, but this process is:

• Time-consuming: it requires inspectors to assess each unit individually
• Prone to inconsistencies: as subjective evaluations may vary between inspectors
• Difficult to scale: especially in high-volume logistic environments

To improve inspection accuracy, efficiency, and scalability, VRR was looking for an automated and consistent way of performing damage detection on ULDs and partnered up with Faktion.

Faktion developed an AI-driven damage classification system, embedded in a fully functional smartphone and tablet application for in-field use. The application enhanced container inspection workflows with a seamless, user-friendly interface. The solution leverages computer vision to analyse images of cargo containers, identifying potential damages and assisting inspectors in making informed decisions directly from their mobile devices.

A key challenge in training the AI model was the limited availability of real-world damaged container images. To address this, we created 3D models of containers to generate synthetic images of both undamaged and damaged ULDs. This approach significantly expanded the training dataset, allowing the model to learn from a broader range of damage scenarios and improve detection accuracy

Key features of the solution include:

• AI-powered damage detection: the model identifies specific damage types, including holes in aluminium plating and damaged container straps.
• Synthetic data augmentation: 3D-generated images of containers with synthetic damages enhanced the model’s ability to generalise and detect real-world damage more accurately
• Mobile inspection application: a dedicated smartphone and tablet app allows inspectors to capture images, receive real-time AI suggestions for detected damages and damage types, and verify results in the field
• Automated decision support: the AI system classifies containers as airworthy ('GO') or non-airworthy ('NO-GO'), providing a structured damage overview
• Data-driven process improvement: all images and inspection results are logged for ongoing model refinement and process optimisation

This approach allows VRR Aero to maintain safety and compliance while reducing reliance on manual inspection efforts, all within a purpose-built mobile solution.

We implemented a structured development process, ensuring robust AI performance and seamless integration into VRR Aero’s workflows:

1. Prototype development & testing: an initial Proof-of-Concept (PoC) was built to validate the model’s ability to detect two damage types. The prototype was tested on real-world container images to refine accuracy and usability.
2. Data annotation & model training: the AI model was trained on a labeled dataset of (synthetic) container images, with bounding boxes and polygon masks used to highlight damage locations
3. Smartphone & tablet application: the PoC was further scaled to cover all damage types and the inspection system was developed as a fully functional mobile app, designed to work efficiently in the field, even in harsh conditions.
4. User-centric application design: the app was designed for quick and easy navigation, ensuring usability for ground personnel with minimal training
5. Integration & feedback loop: the AI system was iteratively improved based on operator feedback and additional dataset expansions, ensuring optimal performance in real-world conditions

This phased approach ensured that the AI model was reliable, the mobile application was user-friendly, and the system adapted to real-world inspection challenges.

The inspection system has streamlined the damage detection process at VRR Aero, delivering key benefits:

• Faster inspections: reducing the time required for visual assessments while ensuring thorough evaluations
• Seamless mobile experience: inspectors can capture, analyse, and review damage data directly from their smartphone or tablet
• Consistent damage classification: standardising inspections to minimise human subjectivity
• Scalability for high-volume inspections: allowing for increased throughput as cargo operations grow
• Improved data utilisation: enabling continuous AI model improvements through logged inspection data

While the system currently assists human inspectors, it lays the groundwork for future fully automated cargo inspections. By integrating AI into cargo container inspections, VRR Aero is making its safety and compliance processes more efficient, data-driven, and future-proof, all through a dedicated mobile solution built for operational use.