Automatic detection of powerlines in UAV remote sensed images

Powerline detection is one of the important applications of Uninhabited Aerial Vehicle (UAV ) based remote sensing. In this paper, powerlines are detected from UAV remote sensed images. The images are acquired from a Quad rotor UAV fitted with a GoPro® camera. In the proposed method pixel intensity-based clustering is performed followed by morphological operations. K-means clustering is applied for clustering. The number of clusters to be used in k-means clustering is automatically generated using Davies-Bouldin (DB) index. Further, the clustered data is processed to improvise the extraction using mathematical morphological operations. Performance of powerline extraction is analysed using confusion matrix method. In the observed results of powerline extraction using DB index, evaluation features derived from confusion matrix is close to one, indicating good classification.