Aircraft recognition in remote sensing images based on saliency and invariant moments
Jun Fu, Lilue Fan, Zhiguo Yang · 2016
The traditional aircraft target recognition algorithms in remote sensing images are mostly based on specific theories. In the case of less interference, traditional algorithms can work well. But however, there are a large number of interfering factors in the remote sensing images actually, such as background, noise and so on. At this time, traditional algorithms fail because of low recognition accuracy and large time spent. Aiming at the shortcomings of traditional methods, this research has proposed a new kind of aircraft target recognition algorithm based on saliency images and invariant moments. The algorithm uses Itti algorithm to extract salient targets after pretreatment, then uses the 8 neighborhood searching method to find connected regions in binary images for determining the numbers and location of the candidate targets. Finally, identify the candidate targets by using the combined moments based on affine invariant moments and Pseudo-Zernike moments. The experiment results show that this algorithm has high detection accuracy, less time spent, low rate of false alarm.