基于缺损目标的CamShift序列图像跟踪方法

Tracking Method of CamShift Sequence Image Based on Incomplete Object

在标准CamShift算法对缺损目标进行跟踪的过程中,由于算法不能根据目标的变化实时调整跟踪窗口形状,造成跟踪窗口内包含缺损区域的背景图像信息,降低了目标的匹配能力,导致CamShift算法不能适用于缺损目标的跟踪。针对此问题,文中提出一种新的缺损目标CamShift序列测量图像跟踪方法,将目标模型和候选模型之间的相似性作为权值,对核函数的概率范围进行限制,减少数据集内的数据量,实时调整跟踪窗口的形状,从而提高匹配相似度,解决了经纬仪序列测量图像中缺损目标的稳定跟踪问题。实验结果表明,此算法在目标缺损50%的情况下,仍然能够稳定跟踪。

The algorithm can not adjust in real time the shape of the tracking window when the incomplete object is tracked by CamShift algorithm.The background image of the incomplete area is included in the tracking window,and the ability of the object matching is reduced,so CamShift algorithm can not track the incomplete ob ject accurately.A new method of CamShift sequence measurement image is proposed,taking the similarity be tween the object model and the candidate model as the weight,the probability range of the kernel function is lim ited,the data volume of data set is reduced and the shape of the tracking window is adjusted in real time.The method can improve the matching similarity and solve the problem of stable tracking incomplete object in the the odolite measurement image.The experimental results show that the method can track steadily even when the loss of the object reaches 50%。