An Edge-Boxes-Based Intruder Detection Algorithm for UAV Sense and Avoid System
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
With the great development of unmanned aircraft system (UAS) over the last decade, sense and avoid (SAA) system has been a crucial technology for integrating unmanned aircraft vehicle (UAV) into national airspace with reliable and safe operations. This paper mainly focuses on intruder detection for SAA system. A robust algorithm based on the combination of edge-boxes and spatial pyramid matching using sparse coding (sc-SPM) is presented. The algorithm is composed of three stages. First, edge?boxes method is adopted to obtain a large number of proposals; Second, the optimization program is presented to obtain intruder area-of-interest (ROI) regions; Third, sc-SPM is employed for feature representation of ROI regions and support vector machines (SVM) is adopted to detect the intruder. The algorithm is evaluated under different weather conditions. The recall reaches to 0.95 in dawn and sunny weather and 0.9 in cloudy weather. The experimental results indicate that the intruder detection algorithm is effective and robust with various weather under complex background.
ZHANG Zhouyu, CAO Yunfeng, ZHONG Peiyi, DING Meng, HU Yunqiang. An Edge-Boxes-Based Intruder Detection Algorithm for UAV Sense and Avoid System[J]. Transactions of Nanjing University of Aeronautics & Astronautics,2019,36(2):253-263