Improving stability of aerial videos acquired through vision sensors onboard UAVs for applications in precision agriculture

Certain nonlinear dynamic factors come into play to perturb the stability of UAVs during aerial data acquisition process. For instance, the vibration(s) caused by the motor(s), gust of winds in different directions and additional payload for custom applications are the major reasons for in-flight instability that may lead to alter the pitch, roll and yaw of the UAV thus affecting the quality of aerial data. The perturbations are generally managed autonomously using different mechanical and electrical sensors installed within the UAVs leading to capture smooth aerial data. Nevertheless, additional vision sensors mounted directly on UAVs remain prone to these perturbations and as a result instability is created in the video frames being captured producing a shaky video. In this paper, a novel algorithm is proposed to stabilize the instable videos captured through vision sensors attached directly to the frame of a UAV for applications in precision agriculture. The scope for the proposed approach covers a wide range of applications, e.g., generation of orthomosaic maps, vegetation maps, image registration, etc. The proposed approach eliminates geometric distortions in video frames using estimations and corrections based on affine and projective transformations. The results show significant improvement in video stability.