Bachelor Thesis - Activity Recognition in the Aircraft Turnaround Process Using Video-Based Analysis

Description

Task mining is a subfield of process mining that focuses on capturing, analyzing, and optimizing human-centered tasks at a detailed level. Traditionally, task mining leverages digital interaction data, such as clicks or keystrokes, to provide insights into how tasks are carried out in software. However, similar techniques can be applied in physical contexts through video-based activity recognition, enabling us to analyze real-world actions and interactions where direct system data is unavailable.

One compelling application of such video-based analysis is the aircraft turnaround process, a critical factor in airline profitability, as efficient ground operations allow for reduced downtime and more airtime. This process involves numerous interdependent activities, such as passenger deboarding, baggage handling, and refueling, all of which must be synchronized under strict regulations. While basic data on flight schedules and delays is available publicly, it provides only a coarse-grained view of these operations.

This bachelor thesis project aims to explore the use of video-based activity recognition to detect and analyze specific tasks in the aircraft turnaround process, as captured by live video streams. Many airports offer public live streams that, despite variable camera angles, may still provide enough visibility to detect key activities and resources, such as gangways, baggage carts, and fueling trucks. By applying object detection algorithms, we seek to capture these activities and incorporate them into a refined process model. This model will include confidence levels to reflect the uncertainty of detection, distinguishing between real-world events, detected events, and interpolated activities.

Prerequisites

The thesis will involve a combination of process mining and computer vision techniques. Candidates should have a basic understanding of process modeling and computer vision, and experience with machine learning frameworks (such as TensorFlow or PyTorch) are beneficial. Familiarity with video processing and process mining is a plus.

Pointers

• e.g., Schmidt, 2017: A review of aircraft turnaround operations and simulations, https://doi.org/10.1016/j.paerosci.2017.05.002

More Information and Application

For more information send an e-mail to István Koren. Please include your CV and transcript of records when applying for this thesis project.