基于CamShift的运动目标跟踪算法研究

发布时间：2019-05-22 16:44

【摘要】：随着计算机软件和硬件的快速发展,基于计算机视觉的智能视频监控技术日益重要。其中,运动目标跟踪技术是智能视频监控最重要的技术之一,结合了众多领域内的高级技术,被广泛地应用于医疗诊断、智能监控、自动驾驶、机器人和基于视频的人机交互等领域。Cam Shift算法利用目标区域在HSV颜色空间下的H分量作为目标特征,对目标形变敏感度较低,又由于该算法的复杂度低、计算量小、实时性好等优点,被广泛用于运动目标跟踪领域。本文基于Cam Shift算法主要进行了两方面的改进,一方面将Cam Shift算法与运动目标检测算法结合。为了减少因加入运动目标检测算法而增加的计算量,采用仅包含运动目标在内的小区域图像进行运算,这种方法大大提高了整体算法的运算速度。另一方面,针对Cam Shift算法在运动目标运动过程中接近与其颜色相近的固定遮挡物时会出现明显偏移的问题,提出运用单高斯背景模型对运动目标进行背景建模的方法。通过实验证明,该方法能够有效地解决了原Cam Shift算法在接近颜色相近固定遮挡物时跟踪出现偏移的问题,在一定程度上增加了原算法的准确性。
[Abstract]:With the rapid development of computer software and hardware, intelligent video surveillance technology based on computer vision is becoming more and more important. Among them, moving target tracking technology is one of the most important technologies of intelligent video surveillance, which combines advanced technology in many fields and is widely used in medical diagnosis, intelligent monitoring and autopilot. Cam Shift algorithm makes use of the H component of the target region in the HSV color space as the target feature, and has low sensitivity to the target deformation, and because of the low complexity of the algorithm and the small amount of computation, the cam algorithm makes use of the H component of the target region in the HSV color space as the target feature. It is widely used in the field of moving target tracking because of its good real-time performance. In this paper, two improvements are made based on Cam Shift algorithm, on the one hand, Cam Shift algorithm is combined with moving target detection algorithm. In order to reduce the computational complexity caused by adding the moving target detection algorithm, the small area image, which only includes the moving target, is used to calculate, which greatly improves the operation speed of the whole algorithm. On the other hand, in order to solve the problem that Cam Shift algorithm will appear obvious offset when it is close to the fixed mask whose color is similar to the moving target in the process of moving target motion, a method of background modeling of moving target using single Gaussian background model is proposed. The experimental results show that this method can effectively solve the problem of tracking migration when the original Cam Shift algorithm is close to the fixed mask with similar color, and increases the accuracy of the original algorithm to a certain extent.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【共引文献】