Target group distribution pattern analysis with bagged convolutional neural networks for UAV distribution pattern identification

Abstract

Purpose The purpose of this study is to address the limitations of existing target group distribution pattern analysis methods and identify subtle distribution differences within and between the groups with no pre-specified distribution features. Classical work generally concentrates on either the group distribution tendency or shape as a whole and simply ignores the subtle distribution differences within the group. Other work is constrained to pre-defined spatial distribution features. Design/methodology/approach This study proposes a novel algorithm for target group distribution pattern analysis. This study first transforms the group distribution data with uncertain measurements into a distributional image. Upon that, a bagged convolutional neural network model is constructed to discriminate the delicate group distribution patterns. Findings Experimental results indicate that our method is robust to target missing and location variance and scalable with dataset size. Our method has outperformed the benchmark machine learning methods significantly in pattern identification accuracy. Originality/value Our method is applicable for complex unmanned aerial vehicle distribution pattern identification.