基于SIFT特征度量的Mean Shift目标跟踪算法

MEAN SHIFT OBJECT TRACKING ALGORITHM BASED ON SIFT DESCRIPTOR

传统的Mean Shift算法,在诸如跟踪目标出现尺度变化、旋转、噪声干扰等复杂情况下,无法得到准确的跟踪结果。提出了一种基于尺度不变特征变换SIFT(Scale Invariant Feature Transform)特征度量的Mean Shift目标跟踪算法,首先根据SIFT算子计算跟踪目标附近的关键点位置和尺度,并获取该尺度空间下关键点邻域的特征向量,然后用跟踪目标区域内的特征向量的模值-方向分布直方图表示该目标,最后使用Mean Shift算法进行跟踪。实验结果表明,该算法在跟踪目标出现尺度变化、旋转、噪声干扰和遮挡等情况下能够准确地跟踪物体,鲁棒性好。

When the intricate conditions,such as scale modification,rotation,noise interference and so on,occur to the tracking object,ordinary object tracking method based on Mean Shift is difficult to get accurate tracking result.This paper proposes a feature description SIFT-based Mean Shift algorithm.It first calculates the position and scale of key points around the tracking object using SIFT descriptor,as well as gets feature vectors of neighbourhood of the key point in the scale space,and then uses the histogram of module value-direction distribution of the feature vectors within the region of tracing object to delegate the moving object,at last it uses Mean Shift algorithm to track the object.Experiments results demonstrate that this algorithm can track the object accurately in conditions of scale modifications,rotation,noise interference and occlusion occurring to the tracking object with good robustness.