Detection of Road for Landing of Aircraft in an Unfamiliar Environment: A Comparative Study

This paper is a comparative study about detecting straight road from satellite images. There are multiple applications of road detection. Here, only straight road is considered for use as a landing strip for aircraft emergency landing. To fully guarantee the safe landing of aircraft, multiple criterions are required to be addressed, for example, buildings and traffic. However, the focus of this paper is only to detect straight road from aerial image to ensure it is feasible to land an aircraft or not. If it is feasible, only then the detection process can move forward to the analysis of non-road objects. To find such road, three different image processing methods are used which are (Canny, Sobel and Prewitt), Fuzzy C-Means (FCM) clustering method and Markov Random Field (MRF) classification model. Each method is used to segment the roads from non-road objects. Since, edge detectors and segmentation models may have broken segments morphological operations are applied to join these broken segments, namely dilation and erosion. Then, the Hough transform is applied to detect a straight road. The results obtained were compared and was concluded that canny performed better as compared to other methods used in this comparative study. But practically none of them were found effective enough as straight road detectors. In the end, some issues are addressed and few solutions have been proposed for future work on this paper.