Keeping track of ever growing numbers of baggage across large infrastructure facilities can be costly and error prone.
Enhance and later replace bar code based systems by marker-less computer vision approaches.
Catalysts adapted state of the art machine learning approaches from other identification problem domains (e.g. face recognition) to the task of baggage identification.
Tracking baggage using conventional means (like bar codes etc.) is both costly and error prone, for example, bar codes can be lost. Furthermore, a continuous tracking of baggage within large scale installations requires a large number of tracking sensors – which are both costly in acquisition and maintenance.
By augmenting existing tracking solution and camera based systems the overall tracking accuracy is improved and new use cases can be covered.
Camera systems mounted at key locations of the baggage transport logistic process (e.g. at check in counters, before and after security screenings, at aircraft loading bays, …) can both detect and track baggage and thus ensure a complete audit trail.
This can be used in a variety of use cases:
Instead of developing a large centralized system Catalysts divided the solution into two main parts:
Using this architecture images are processed in near real time. This means that the video doesn’t need to be streamed over the network, what contributes to low network load. Also, central processing server is not needed (no single point of failure).
Catalysts created a multi stage real time image processing pipeline consisting of the following steps:
1. Object detection: detection, localization and classification of baggage on the video stream.
2. Instance segmentation: which parts of the image belong to which object exactly?
3. Object Fingerprinting: finding a numerical and representation of the characteristic of the detected baggage.
A server receives object fingerprints from the image processing services and stores them in a database. By comparing these fingerprints – a quick and easy operation – the server can easily compare current detection of baggage items to previous detection and thus build a complete audit trail.
Emerging Technologies, Computer Vision
Der voll-integrierte Arbeitsplatz für ein ergonomisches Zusammenspiel aller Justizanwendungen.
Creating customized profiles of special sections and tubes online – even without CAD-knowledge! Read more
We apply Machine Learning methods to train models of a certain domain and then use these models for predicting.
An intuitive and unobtrusive method for visualizing and interacting with large multidimensional scan data using state of the art augmented reality technology.
Artificial Intelligence for production process automation
An intelligent and scalable system for managing vehicles and predicting their usage for optimal route and resource planning.
Satellite Earth Observation to highlight hot spots of damaged forest areas.
