Robust Aircraft Recognition using High Resolution Reconstructed Images

Background: Aircraft Recognition with high resolution space borne images is a challenging task.It is different from other natural object recognition. The ground image of aircraft is complex and disturbing. The target image has to be detected in a cluttered background. Methods: The proposed technique takes into account multiple features such as circular Frequency filter (CF-filter), fuselage symmetry and angle between fuselage axis and wing for candidate selection. Normalized Tree cut algorithm isused for Segmentation of image done and aircraft alignment done using Histogram Oriented Gradient (HOG). Automatic jigsaw puzzle solution is implemented for aircraft reconstruction of pieces. Findings: In the candidate selection stage because of cluttered background the aircraft is not identified. Here the multiple identification techniques such ascircular frequency filter (CF-filter), fuselage symmetry and angle between fuselage axis and wing are considered. Applications: The above said techniques were dealt earlier as independent features. Because of proposed multiple feature consideration accuracy in candidate identification is enhanced.