Abstract:

In order to solve the problem of tracking failure caused by the inadaptability of frame scale and occlusion of target in traditional Kernelized Correlation Filters（KCF） tracking method, taking circular cone sleeve target tracking in aerial refueling visual navigation as an example, a scale adaptive correlation tracking method based on block detection for circular target is proposed in this paper. The problem of target scale transformation in circular target tracking is solved by using the method of extracting feature points and contour fitting. The occlusion and tracking failure are distinguished by block detection, which eliminates the accumulated tracking error under occlusion and achieves target tracking under partial occlusion. Experimental results show that the proposed scale adaptive correlation tracking method for circular target has high real-time performance, reliability and accuracy, and achieves stable and reliable tracking for cone sleeve in air refueling docking process.

Key words: kernelized correlation filters, target tracking, feature point extraction, block detection, scale adaptation, occlusion judgment

摘要：

为解决传统基于核相关滤波器（KCF）的目标跟踪方法中跟踪框尺度无法自适应及目标被遮挡导致的跟踪失败问题，以空中加油视觉导航中圆形锥套目标的跟踪为例，提出了一种分块检测尺度自适应的圆形目标相关跟踪方法。利用目标外部边缘特征点的提取与轮廓拟合方法解决了圆形目标跟踪所存在的目标尺度变换问题，通过分块检测将跟踪过程中目标遮挡情况与跟踪失败情况区分开来，消除了遮挡情况下跟踪积累误差，实现了部分遮挡情况下目标的跟踪。实验结果表明，提出的圆形目标尺度自适应相关跟踪方法具有较高实时性、可靠性与准确性，实现了空中加油对接过程中对于锥套的稳定、可靠跟踪。

关键词: 核相关滤波器, 目标跟踪, 特征点提取, 分块检测, 尺度自适应, 遮挡判断