Anzeige
Erschienen in:
2017 | OriginalPaper | Buchkapitel
Unsupervised and Supervised Activity Analysis of Drone Sensor Data
verfasst von : Roman Neruda, Martin Pilát, Josef Moudřík
Erschienen in: Applied Computer Sciences in Engineering
Abstract
This paper deals with methods for identification of drone activities based on its sensor data. Several unsupervised and supervised approaches are proposed and tested for the task of activity analysis. We demonstrate that sensor data, although quite correlated, are still prone to standard dimensionality reduction techniques that in fact make the problem hard for unsupervised methods. On the other hand, a supervised model based on deep neural network is capable of learning the task from human operator data reformulated as a classification problem.