运动图像快速跟踪技术研究

Research for the Technology of Motion Image Fast Tracking

针对运动图像目标检测需人工干预,跟踪核窗固定、目标易丢失等问题,提出了融合背景差分、帧间差分和灰度阈值技术的变背景帧间差分法,并结合灰度质心定位和自适应核窗宽改进了MeanShift跟踪算法。该方法能够在复杂环境下检测出各种运动目标,并进行实时跟踪,当目标发生尺度、旋转、无规律大位移变化时都能够快速准确地检出并跟踪。大量实验仿真表明,本算法检出率高,迭代次数少,实时性强,具有很好的适应性和鲁棒性。

according to some problems,such as requiring the manual intervention for motion image target detection,tracking the core window to be fixed,the goal easy to lose and so on,the paper proposes a method that changes the background frame difference based on background difference,frame difference and threshold value,and combines with the grayscale centroid location and auto-adapted core window width,then improved the MeanShift tracking algorithm.This method can detecte all variety of moving targets in a complex environment,and carry on real-time tracking,even when the target makes big changes in the scale, rotation,and disorder large displacement situation,it also can carry on the fast and accurate detection and tracking.A large number of simulation experiments show that this algorithm has the high detection rate,fewer iteration numbers, strong real-time,good adaptability and robustness.