跟踪(教育)
计算机视觉
计算机科学
人工智能
红外线的
平面(几何)
遥感
物理
窗口(计算)
光学
雷达跟踪器
跟踪系统
噪音(视频)
出处
期刊:
日期:2026-05-22
卷期号:: 1-6
标识
DOI:10.1109/cvippr69607.2026.11604545
摘要
Infrared small target tracking on mobile platforms is highly susceptible to correlation surface distortion, leading to degraded localization accuracy and unstable tracking performance. To address this issue, this paper proposes a tracking algorithm based on frequency-domain plane fitting. Built upon the phase correlation framework, the proposed method enables high-precision inter-frame displacement estimation by reformulating sub-pixel localization as a linear phase modeling problem. Specifically, an initial coarse localization is first performed at the integer-pixel level using phase correlation to obtain a reliable displacement estimate and compensate for large motion. Subsequently, sub-pixel displacement is estimated by fitting a two-dimensional plane to the compensated phase surface, effectively exploiting the linear shift property of phase information in the frequency domain. Unlike conventional approaches that rely on the shape of the spatial-domain correlation peak, the proposed method directly estimates motion parameters from phase gradients, thereby improving robustness under distortion and enabling accurate localization within extremely small observation windows. Experimental results on real infrared image sequences demonstrate that the proposed method outperforms traditional phase correlation and discriminative correlation filter (DCF)-based methods in terms of sub-pixel localization accuracy and tracking smoothness Furthermore, compared to deep learning-based trackers that demand significant computational resources, the proposed algorithm maintains superior real-time efficiency, making it better suited for deployment on resource-constrained mobile platforms.
科研通智能强力驱动
Strongly Powered by AbleSci AI