Find in Library

Aiming at the problems encountered by traditional object tracking, such as confusion between object and background and rapid change of target appearance, a fast and robust object tracking algorithm (FRHT) is proposed based on the 2D spatial information and rich 1D spectral information included in hyperspectral video. Firstly, based on the characteristics of hyperspectral data, spectral attention mechanism is introduced into the traditional spatial attention mechanism, and an adaptive update learning tracker under the framework of correlation filtering is designed. Secondly, hyperspectral moving object features are designed manually to speed up the operation of the tracker. Finally, a moving object anomaly detection mechanism is proposed to improve the robustness of the tracker. The simulation results show that the speed and accuracy of the tracker FRHT are significantly better than the traditional tracking algorithms such as KCF, SAMF and CSR-DCF on hyperspectral data sets, and the accuracy is improved by more than 2%. After introducing the moving object detection mechanism, the robustness of the algorithm is improved.