一种改进的Camshift目标跟踪算法

Target tracking based on improved Camshift algorithm

针对传统连续自适应场值漂移(Camshift)算法易受同色物体的干扰、抗遮挡性能差等问题,该文提出一种基于运动分割和H-S二维概率分布直方图的改进型Camshift算法。在Camshift算法跟踪之前,结合运动信息,采用三帧差分法从背景中提取出运动目标;在HSV空间根据Bhattacharyya系数,采用一维H分量颜色直方图与H-S二维直方图相结合的方式生成反向投影图;结合Kalman滤波器预测下一帧目标的位置。实验证明,该文算法克服了传统Camshift算法的缺点,增强了抗干扰能力,能够准确实时地跟踪运动目标,鲁棒性好。

In view of that the traditional Camshift algorithm for the semi-automatic target tracking is easy to be interfered with the color of objects and poor in anti-occlusion performance,an improved Camshift algorithm based on the motion segmentation and the 2-D H-S probability distribution histogram is proposed here.Before tracking of Camshift algorithm,the moving target is extracted from background by the three frame difference method combined with the motion information.According to the Bhattacharyya coefficient in HSV space,the 1-D H component color histogram and the 2-D H-S histogram are combined to generate the back projection.The Kalman filter is used to forecast the location of the object in the next frame.The test results show that,the improved algorithm overcomes shortcomings of the traditional Camshift algorithm effectively,has the strengthened anti-jamming ability,can complete accurate and real-time target tracking,and has better robustness.