Om prosjektet
Kongsberg's Remote Tower System (RTS), developed in collaboration with Avinor, is a remote control tower system for airports. It includes a 360-degree camera delivering high-resolution images in both visual and infrared (IR) spectrums. It also features a Pan Tilt Platform equipped with PTZ (Pan-Tilt-Zoom) and IR cameras, a Laser Range Finder, and a Signal Light Gun. The project owners are seeking the ability to train customized machine learning models using RTS data. These models can later assist air traffic controllers by providing annotated information on their Head-Up Displays (HUDs). Training a machine learning model necessitates a substantial volume of high-quality data, the generation of which can be a tedious process when performed manually. The ability to delegate this labor-intensive task to a computer capable of autonomously and continuously collecting data is highly beneficial. This approach not only enhances human resource efficiency, but also boosts profitability. The project assignment is to create a data gathering pipeline that automates the task of collecting such training data. The primary sensor data for the assignment will be videos from the 360-degree camera in both visual and IR spectrums, along with Moving Target Indicator (MTI) tracking data. Our goal with this project is to provide a functional data gathering pipeline that will enable the use of video streams from operational RTSs for the storage of large volumes of images with associated data. The data will be processed by a program aligning tracking data with the corresponding moving object of interest in the video. The program will utilize a machine learning model for object labeling, and store the labeled raw data in a database with relevant information, including MTI data. Additionally, the data gathering pipeline will include a quality assurance mechanism to verify the accuracy of object labeling. By establishing a functional pipeline, we can achieve automated storage of vast quantities of labeled images. These images can then be efficiently retrieved to construct a dataset for training a customized machine learning model.
Bilder:
-
The image shows Remote Tower with 360 camera and Pan tilt zoom camera at one of Avinor's airports. We have received this image from Kongsberg.
-
The image shows the operations center at Avinor's Remote Towers Centre in Bodø, where operators control and manage airport activities across Norway using remote control technology. We have received this image from Kongsberg.
Prosjektdeltakere
Shokoufeh Ghadersouri, Daniel Eggereide, Ibragim Yusupov, Ainhize Ituarte
Om oppdragsgiveren
Kongsberg: Established in Norway in 1814, Kongsberg is a distinguished international company headquartered in its namesake city. Specializing in advanced technological solutions, Kongsberg is dedicated to improving safety, efficiency, and security for its diverse clientele. These solutions are meticulously crafted to excel in complex operations and withstand extreme conditions. Avinor: Avinor is a state-owned company that operates 43 airports in Norway. With approximately 2,800 employees, the company is responsible for ensuring that 50 million annual passengers safely reach their destinations. Avinor aims to make aviation more environmentally friendly and sustainable in the long term. The company collaborates closely with environmental organizations and plays a crucial role in the development of sustainable aviation fuels and electric aircraft. Avinor has ambitious plans to achieve its goal of becoming fossil-free by 2030.
Veileder
Selina Demi, Høgskolen i Østfold