Algorithm of Unmanned Aerial Vehicle’s Positioning Using Sequence of Video Stream Frames

Abstract

The paper describes the algorithm of unmanned aerial vehicles (UAVs) positioning using images of the terrain obtained from the camera mounted on board. This algorithm can be used in UAV navigation systems in the conditions of strong interference of global positioning systems (GPS) signal. The algorithm consists of several stages: preprocessing, selection of special points, matching of neighboring frames and determination of displacement. In the first stage, each image undergoes a preliminary processing, which consists of size reduction and conversion to grayscale mode. In the second stage the strongly expressed points with difference module which exceeds a certain threshold are defined on images by using the difference between two smoothing filters. Each reference point is described by its type and coordinates. These points are required to match neighboring frames. In the third step, the control points of neighboring frames are matched by searching for points with the same type in a certain radius. After matching all frames are joined together to form a single image of the traversed route. The current frame undergoes a perspective transformation before joining. To do this, four points which form a quadrilateral with a maximum area are selected from the matched points. The transformation matrix parameters are calculated from the coordinates of these points on the current and previous frames. After that, frames overlap each other and form an image of the traversed route. When processing is complete, the center of mass offset of the singular points is determined and the new coordinates are calculated. In conclusion an example of the generated image of the traversed route is given, and the effectiveness of this algorithm is evaluated. The algorithm was studied on the images obtained with the help of special software, which simulates the video stream from the onboard camera.