Design and implementation of a vision-based hovering and feature tracking algorithm for a quadrotor

[摘要] This paper demonstrates an approach to the vision-based control of the unmanned quadrotors for hover and object tracking. The algorithms used the Speed Up Robust Features (SURF) algorithm to detect objects. The pose of the object in the image was then calculated in order to pass the pose information to the flight controller. Finally, the flight controller steered the quadrotor to approach the object based on the calculated pose data. The above processes was run using standard onboard resources found in the 3DR Solo quadrotor in an embedded computing environment. The obtained results showed that the algorithm behaved well during its missions, tracking and hovering, although there were significant latencies due to low CPU performance of the onboard image processing system.

[发布机构] Division of Information Technology and Engineering, School of Engineering, University of South Australia, Australia^1;Joint and Operations Analysis Division, Defence Science and Technology Group, Australia^2

[学科分类] 航空航天科学

[关键词] Design and implementations;Embedded computing;Feature tracking algorithm;Flight controllers;On-board resources;Onboard image processing;Pose information;Vision based control