Particle Filter Based on Harris Hawks Optimization Algorithm for Underwater Visual Tracking

Due to the complexity of underwater environment, tracking targets via traditional particle filters is a challenging task. To resolve problem that accuracy filter low due sample impoverishment caused by resampling, in this paper, new algorithm using Harris-hawks-optimized (HHOPF) proposed. At same time, target feature construction and scale transformation addressed, corrected background-weighted histogram method introduced into recognition, combined realize scaling during tracking. In addition, enhance computational speed tracking, paper constructs nonlinear escape energy Harris hawks order balance exploration exploitation processes. Based on proposed HHOPF tracker, we performed detection evaluation Underwater Object Tracking (UOT100) vision database. The compared with evolution-based algorithms filters, as well recent tracker-based correlation some other state-of-the-art methods. By comparing results test data sets, it determined presented improves overlap 11% algorithms. experiments demonstrate visual provides better results.