基于自适应核窗宽的红外目标跟踪算法

Infrared target tracking algorithm based on adaptive bandwidth of Mean Shift

针对传统均值漂移算法无法对对比度低、尺度变化的红外目标进行有效跟踪的问题,提出一种改进MeanShift算法.首先融合灰度和纹理两方面的信息,并分别定义背景灰度和纹理加权系数,实现了目标的准确定位;然后,提出一种基于背景和前景目标相似度的核窗宽选取算法,自动选取窗口缩放比例,得到与目标尺度一致的跟踪窗口.实验结果表明,所提出的算法能够实现对红外目标的跟踪,并且对尺度变化的目标具有较好的适应性.

For the problem that the target,tracking algorithm based on Mean Shift may be lost when the infrared target has low SNR or owns a dynamic change in scale, an improved Mean Shift tracking algorithm is proposed. Firstly, the features of grayand texture are fused to enhance the target information. Then Coefficients based on the gray and texture histograms of the background pixels around the target are computed and incorporated into the computation of gray and texture histograms. After the accurate localization of the infrared target obtained, an objective function based on the similarity of background and target is proposed. Finally, the adaptive bandwidth is obtained through making the objective function minimum. Experimental resultsshow that the proposed algorithm is effectiv and robust and can be adapfed to the target change in scale.