Infrared maritime target tracking via correlation filter with adaptive context-awareness and spatial regularization

Abstract Infrared maritime target tracking has been a typical application of infrared tracking techniques, and it remains a challenging task to design trackers with outstanding robustness and efficiency simultaneously. To address this challenge, we propose a real-time infrared maritime target tracker based on adaptive context-aware and spatially-regularized discriminative correlation filters (ACSDCF). The adaptive context-aware strategy is utilized to incorporate more negative samples explicitly and suppress background interference. We design an automatic spatial regularization mechanism to alleviate the inherent drawbacks of correlation filter-based trackers by utilizing response map information. Moreover, an adaptive updating strategy based on response map variation is introduced to cast a tradeoff between tracking speed and accuracy. Extensive experimental results in different infrared maritime target sequences demonstrate that the comprehensive properties of the proposed ACSDCF, with a tracking speed of 40 + frames per second, are superior to state-of-the-art trackers.