Color Optic Flow: A Computer Vision Approach for Object Detection on UAVs

Implementation of reliable object avoidance is crucial for the survival of autonomous vehicles intended to navigate in densely populated environments. Uninhabited aerial vehicles (UAVs) constitute a type of self guided platform which must rely on different types of sensors in order to navigate autonomously. The exploratory nature of UAVs has made vision systems a desirable option to solve the object avoidance problem by utilizing image data as a rich source of environmental information. The base of object detection and collision avoidance is motion estimation. Computer vision techniques have been developed throughout the years in order to estimate motion from two dimensional images. Most of these algorithms are computationally expensive, thus fail to meet the real time constraints imposed by UAV flight. However, recent developments in optic flow techniques using color imagery are giving vision systems the potential to solve certain navigation challenges on UAV platforms. In this paper a fixed wing UAV platform is described and constraining issues are presented and discussed. Different implementations of color optic flow algorithms are tested on actual flight video footage. Finally, the viability of optic flow as a desirable technique for real time motion estimation is evaluated