Drogue Detection for Autonomous Aerial Refueling Based on Adaboost and Convolutional Neural Networks

• Published in: Neural Information Processing Vol. 11304; pp. 437 - 443
• Main Authors: Guo, Yanjie, Deng, Yimin, Duan, Haibin
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Adaboost; Autonomous aerial refueling; CNN; Sample imbalance
• ISBN: 9783030042110; 3030042111
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-04212-7_38

Autonomous aerial refueling (AAR) is an important capability for the future development of unmanned aerial vehicles (UAVs). A robust and accurate algorithm of detecting the drogue is crucial to such a capability. In this paper, we present an innovative algorithm based on the adaptive boosting algorithm and convolutional neural networks (CNN) classifier with improved focal loss (IFL). The IFL function addresses the sample imbalance during the training stage of the CNN classifier. The pytorch deep learning framework with the graphics processing units (GPUs) is used to implement the system. Real scenario images that contain drogue carried by UAVs are for training and testing. The results show that the algorithm not only accelerates the speed but also improves the accuracy.