基于高斯混合模型联合CamShift的运动图像检测跟踪方法

A moving image detection and tracking method based on Gauss mixed model combined with CamShift

摘要:

CamShift算法是典型的运动图像跟踪方法,但是单纯CamShift准确率低,容易丢失目标。为提高准确略,本文提出将高斯混合模型运用到CamShift算法中进行目标跟踪操作。首先采用高斯混合模型标示出目标局部区域,并将其作为CamShift的初始搜索窗,提高效率；随后对目标进行跟踪时将CamShift算法的窗中心同差分法计算出的目标区域中心作对比,确定后续帧搜索窗,避免目标跟踪丢失。最后,实验证明了该方法可以对目标进行有效跟踪,且在目标颜色同背景色差异小的情况下依然具有非常高的准确率。

Abstract:

The CamShift algorithm is a typical method of moving image tracking, but the accuracy of CamShift is low and it is easy to lose the target. In order to improve the accuracy, this paper applies the Gauss hybrid model to the target tracking operation in the CamShift algorithm. First the Gauss mixture model marked the local area, and the CamShift as the initial search window, improve efficiency; then track the target when the target area center CamShift algorithm with window center difference method to calculate the comparison, determine the subsequent search window frame, avoid the loss of target tracking. Finally, experiments show that this method can track the target effectively, and still has very high accuracy in the case that the color difference between the target and the background is small.