High speed detection of aircraft targets based on proposal oriented FAST and adaptive matching of local invariant features

Abstract

In this paper, a high speed detection method of aircraft targets in remote sensing images is proposed based on proposal oriented FAST and adaptive matching of local invariant features. In order to reduce the search scope, the region of parking apron is extracted by region growing based on OTSU segmentation. Moreover, Binarized Normed Gradient (BING) and Spectral Residual Saliency (SRS) are applied respectively to find useful proposals of potential aircraft targets with minor computing cost. Towards extracted proposals, the algorithm of Features from Accelerated Segment (FAST) is employed to locate key feature points precisely for various sizes of aircraft targets even very small ones. Then local invariant features characterized with well robustness against environment changes are constructed. Finally, the high speed detection algorithm of aircraft targets is implemented through adaptive matching of local invariant features with parameters adjustable accompanied by the size of aircraft targets. Comprehensive experiment results validate the well performance of our method with outstanding superiority in detection speed and accuracy for various sizes of aircraft targets.