Aircraft Pose Recognition Using Locally Linear Embedding

Abstract

Locally linear embedding (LLE) is a prevalent manifold learning method in pattern recognition and machine learning. It preserves the intrinsic structure information of data set and has been widely applied to feature extraction and dimensionality reduction. This paper introduces LLE to aircraft pose recognition. The representative motion poses of an aircraft in the air are analyzed. Unfolding results of aircraft images in different poses show LLE has a natural connection to clustering. Moreover, we employ back propagation neural networks and nearest neighbor algorithms to classify the input samples after dimensionality reduction. Computer simulation testifies the efficiency and accuracy of LLE in aircraft pose recognition.