Vision Based Autonomous Runway Identification and Position Estimation for UAV Landing

Vision based autonomous landing has turned out to be a deep-seated technology of UAV (Un-manned Aerial Vehicle) navigation, guidance and control system in the present scenario of image processing techniques. This study focuses on the two main UAV landing problems reliant on the airborne front-looking cam: how the runway can be recognized in image frames taken by the nose camera (a novel algorithm has been presented which includes Otsu's method, Sobel operator for edge detection, spectral residual saliency map (SRS), binary gradient mask and dilation); Second problem is how to align the UAV according to the runway position for critical landing phase (pixel approximation technique has been employed to estimate the position of the runway left, right, sharp left, sharp right, top or bottom for the support of landing UAV). The results obtained in present experimental method rigorously validated the effectiveness and strength of the proposed method. Moreover, accuracy of the implied runway recognition model is found to be 94% and the processing time is 0.32s which outperforms the other reported models. It is strongly believed that these results would open new vistas for the upcoming studies in the field.