【摘要】 目标跟踪是计算机视觉领域的重要任务之一,工业自动化控制、自动驾驶、安保监控、军事目标侦察等民用和军事领域中广泛存在对视频中的运动目标进行追踪的需求。传统的目标跟踪方法Camshift算法仅基于单一色调直方图进行追踪,在目标被遮挡或目标色调与背景相似时易受到干扰,跟踪效果不佳。针对上述问题,提出改进的Camshift与Kalman滤波联合跟踪算法。首先对目标建立色调与饱和度的二维直方图模板,替代单一信息直方图,提高算法对抗背景干扰的能力;其次将BP神经网络引入Kalman滤波器中,降低噪声与模型变化对预测结果的影响;最后使用Camshift算法和Kalman滤波联合对目标进行跟踪,改善运动目标被遮挡情况下的追踪精度。实验结果表明,改进后的联合跟踪算法有效帧率提升约20%,且单帧图像处理时间小于35 ms,满足实时追踪需求。更多还原

【Abstract】 Target tracking is one of the important tasks in the field of computer vision. The demand for tracking moving targets in video is widely existing in civil and military fields such as industrial automation control,automatic driving,security monitoring,and military target reconnaissance. The traditional target tracking method,Camshift,is based on the hue histogram only. It is susceptible when the target is occluded or the target Hue is similar to the background,and the tracking effect is not good.Aiming at the above problems,an improved joint tracking method of Camshift and Kalman filtering is proposed. Firstly,the two-dimensional histogram template of Hue and Saturation is established for the target,instead of the single information histogram,the ability of the algorithm to resist background interference is improved. Secondly,the BP neural network is introduced into the Kalman filter to reduce the influence of noise and model changes on the prediction results. Finally,the target is tracked by using Camshift algorithm and Kalman filter to improve the tracking accuracy of the occlusion target. The experimental results show that the improved joint tracking algorithm improves the effective frame rate by about 20%,and the single frame image processing time is less than 30 ms,which satisfies the real-time tracking requirements.更多还原