基于侧窗滤波与时空正则化相关滤波的红外弱小目标跟踪

Infrared dim small target tracking based on side window filtering and spatial-temporal regularized correlation filters

  • 摘要: 当目标远离红外系统,其在成像图像上的尺寸较小且信息量较少,使得小目标的持续精确定位成为一项有挑战性的问题。针对这一问题,在相关滤波跟踪框架上,引入能够区分红外弱小目标边缘信息与杂波噪声的侧窗图像滤波方法,提出了一种弱小目标跟踪算法。具体来说,首先利用时空正则化的相关滤波跟踪模型,对目标位置附近更大范围的背景进行考虑。然后,利用侧窗滤波对当前局部搜索区域进行侧窗滤波处理,达到了保留边缘效果的同时剔除了图像噪声。最后,通过原始图像与滤波后图像作差,降低了背景边缘对目标定位错误的影响,并实现小目标状态估计。为验证本文所提算法性能,采用六组红外真实弱小目标图像序列进行实验,并与核相关滤波、空间正则化的相关滤波,以及时空正则化的相关滤波等经典算法作比较。实验结果表明,所提算法在多组复杂背景的图像序列上,获得了较高的跟踪精度,验证了所提算法能有效应对红外弱小目标跟踪任务中的快速运动、低分辨率和强背景杂波等问题。

     

    Abstract: Due to the less information of distant target, it is always challenging to accurately track the target in the task of infrared dim small target tracking. To improve the accuracy, based on correlation filtering framework, the side window filtering method which can extract the edge features of small infrared target is introduced, and an algorithm of distant target tracking is proposed. Specifically, the side window filtering method is used to process the searching area of the current target, this method could restrain the negative influence of background edge on dim small target location. Next, the correlation filters tracking model is constructed with temporal and spatial regularities to achieve accurate target tracking. To verify the performance of the proposed algorithm, six groups of real infrared dim small target image sequences were used for experiments, and the algorithm is compared with other typical algorithms such as KCF, SRDCF and STRCF. The experimental results show that the algorithm could effectively solve the problems of fast motion, low resolution and strong light background in infrared dim small target tracking tasks, getting higher accuracy with image sequences and complex background.

     

/

返回文章
返回