Detecting aircraft through GAVIOTA
Regardless the COVID 19 pandemic, GAVIOTA project activities have been proceeding as planned. Among others, it must be highlighted the design of Detection Node’s algorithms which allow to detect aircrafts in the airspace monitored from visual and audible information locally obtained through cameras and microphones.
To conduct such design, it was selected the use of Deep Learning and Convolutional Neuronal Networks (CNNs) as the most appropriate strategy. Such approach implies the training of the algorithm based on a pool of data which shall be used as reference to compare its predictions. In this case, the network assesses pictures and audio files, and establish whether this data has any pattern which could outcome in the aircraft detection. By having a big amount of good quality data, and by configuring the network properly, this can learn from its mistakes, improving its success percentage.
In order to test and validate the algorithm, it was used different data records gathered by a Detection Node deployed during two demonstration flights conducted in September 2019 in ATLAS Flight Center, and Beas de Segura aerodromes. Through this data, its has been possible to conduct a different set of strategies in laboratory, using different types of networks with different architecture and classification algorithms, different banks of aircraft data for reference.
As main outcome, it was confirmed that aircraft detections obtained meet the requirements initially established. However, it was also detected the need of performing additional test by using data from other scenarios
The project is being carried out by a consortium formed by the aeronautical engineering companies Pildo Labs Galaica, Pildo Labs, and the Advanced Center for Aerospace Technologies (FADA-CATEC).
GAViOTA project is framed into the call ‘Retos de Colaboración 2017’ of the Spanish state program for research, development and innovation oriented to society challenges, within the framework of the State Plan for Scientific and Technical Research and Innovation, 2013-2016.
The project is funded by the ‘Ministerio de Ciencia, Innovación y Universidades – Agencia Estatal de Investigación – Proyecto RTC-2017-6515-4