基于三维直方图的改进Camshift目标跟踪算法

经典的连续自适应均值漂移算法Camshift通过HSV空间的色调Hue分量建立一维直方图,在有光照变化及有相似颜色目标或背景的干扰下,跟踪效果不好。提出一种融合HSV空间中色调、饱和度以及反应物体形状信息的边缘梯度的三维直方图特征,并基于背景模型自适应调整特征直方图三种分量的权重值,提高了算法的跟踪准确度。通过与传统Camshift跟踪实验比较,提出的改进算法在光照变化及相似颜色目标/背景干扰下具有更好的鲁棒性,同样也满足跟踪系统的实时性要求。

Improved Camshift objects tracking algorithm based on three-dimensional histogram

The classcial Camshift （continuously adaptive mean-shift） algorithm builds a one-dimensional histogram only with Hue component from HSV color space,which may lead to the failure of tracking when interferes by illumination variation and similar color object or background. To solve this problem,an improved algorithm based on a three dimensional histogram is proposed,which is built with hue and saturation components from HSV space and edge gradient from object’s shape informa-tion. The object tracking accuracy of the algorithm under background interference was improved on the basis of the weighted value of these three components of background model adaptive adjustment histogram. Compared with the traditional Camshift al-gorithm,experimental results indicate that tracking failure incurred by illumination variations and interference from similar color object or background can be alleviated in the proposed algorithm. The improved algorithm can meet the applicability require-ments of real-time tracking systems.