DSSU: Dual-Stage Sparse Unmixing for Asynchronous Mixed Signal of Infrared Targets

Kewen Huang, Linhui Li, Hongxia Mao, Bo Yang, Yan Tian

Abstract: The unmixing of infrared mixed targets is the foundation for long-range cluster target recognition and scale estimation. In this article, a dual-stage sparse unmixing (DSSU) method is proposed for asynchronous mixed infrared signals with high correlations. Considering the asynchrony of each source signal in the mixed signals, first, each candidate source signal is expanded in the time domain to obtain the set of different source signals and the same source signals appearing asynchronously that may be contained in the mixed signals. Then, a signal energy-based layered sparse unmixing (ELSU) method in the wavelet domain is proposed for the mixed signals to obtain the rough unmixing results. Finally, based on the rough unmixing outcomes, a genetic algorithm based on dual-population cooperative evolution method (GA-DP) is proposed to further improve the accuracy of unmixing. The experimental results demonstrate that the proposed method has perfect unmixing performance.