光子学报, 2015, 44 (9): 0910003, 网络出版: 2015-10-22
基于引导滤波与时空上下文的红外弱小目标跟踪
Infrared Dim-Small Target Tracking Based on Guide Filter and Spatio-Temporal Context Learning
红外与夜视技术 目标跟踪 时空上下文 红外图像处理 弱小目标 图像滤波 贝叶斯分类 傅里叶变换 Infrared and night vision technology Target tracking Spatio-Temporal context Infrared image processing Dim and small target Image filtering Bayesian classification FFT
摘要
由于传统的跟踪算法没有充分利用目标与其局部背景的时空相关性, 使其不能有效地区分背景边缘和红外弱小目标, 从而在跟踪过程中产生偏移现象.针对这一问题, 本文在时空上下文学习跟踪的原理基础上, 分析了跟踪偏移的原因, 并引入图像引导滤波方法, 提出了一种引导滤波结合时空上下文的红外弱小目标跟踪算法.该算法首先采用引导滤波对上下文区域进行处理, 在保留上下文区域云层边缘的同时剔除目标及噪声, 再将其与滤波结果作差.最后利用小目标的“置信图”检测出目标.为了验证该方法的有效性, 采用五组红外小目标序列图像进行实验, 并与经典时空滤波、改进的模板匹配和移动管道滤波等方法作比较.实验结果表明本文提出的方法在主观视觉和客观评价指标方面均优于其它三种经典方法, 且具有更高的目标跟踪精度与较好的实时性.
Abstract
Because traditional target tracking algorithm don′t take full advantages of spatio-temporal relationships between the target and its local background, small target and edges can not be distinguished effectively, thereby the excursion problem occurs. Based on the Spatio-Temporal Context (STC) learning tracking algorithm, excursion problem was analyzed, and a new IR dim target tracking based on Guided Image Filter and STC is proposed. Guided Filter is adopted to preserve edges and eliminate the noise of context areas, and ideal context prior can be calculated after subtracting the filtering result from original image. Then, an ideal confidence map makes an accurate estimation of target. Several group of experimental results demonstrate that the presented method can track the target effectively, compared with several classical methods, such as improved template matching tracking, temporal-spatial fusion filtering algorithm and moving pipeline filtering algorithm. The experimental results show that the proposed algorithm performs well in terms of efficiency and accuracy.
钱琨, 周慧鑫, 秦翰林, 殷世民, 荣生辉, 赵东. 基于引导滤波与时空上下文的红外弱小目标跟踪[J]. 光子学报, 2015, 44(9): 0910003. QIAN Kun, ZHOU Hui-xin, QIN Han-lin, YIN Shi-min, RONG Sheng-hui, ZHAO Dong. Infrared Dim-Small Target Tracking Based on Guide Filter and Spatio-Temporal Context Learning[J]. ACTA PHOTONICA SINICA, 2015, 44(9): 0910003.