Vision Based Obstacle Detection mechanism of a Fixed Wing UAV

In this paper we have developed a vision based navigation and obstacle detection mechanism for unmanned aerial vehicles (UAVs) which can be used effectively in GPS denied regions as well as in regions where remote controlled UAV navigation is impossible thus making the UAV more versatile and fully autonomous. We used a fixed single onboard video camera on the UAV that extracts images of the environment of a UAV. These images are then processed and detect an obstacle in the path if any. This method is effective in detecting dark as well as light coloured obstacles in the vicinity of the UAV. We developed two algorithms. The first one is to detect the horizon and land in the images extracted from the camera and to detect an obstacle in its path. The second one is specifically to detect a light coloured obstacle in the environment thus making our method more precise. The time taken for processing of the images and generating a result is very small thus this algorithm is also fit to be used in real time applications. These Algorithms are more effective than previously developed in this field because this algorithm does the detection of any obstacle without knowing the size of it beforehand. This algorithm is also capable of detecting light coloured obstacles in the sky which otherwise might be missed by an UAV or even a human pilot sometimes. Thus it makes the navigation more precise.