一种动态背景下目标检测与提取算法研究

Research on target detection and extraction algorithm in dynamic background

针对在动态背景下,Vi Be算法存在检测鬼影现象和鬼影消除时间长、算法适应性差以及存在变化较为频繁的背景下前景检测存在误判前景干扰噪声像素点多的问题,提出了一种以Vi Be为基本算法框架,引入了由分散系数建立的自适应半径阈值、借鉴帧差分法建立的鬼影模型和二次判断,以及目标提取后的优化算法。该算法能够提高算法对动态背景的自适应性,能够加速鬼影的消除,较好地减少鬼影对背景模型的干扰,抑制了动态幅度较大的水面区域,消除了整个场景中出现的噪声像素点,降低了误判率。实验结果证明所给出的算法能够有效的提取前景目标,抗干扰性鲁棒性好。

In the dynamic background,the Vi Be algorithm has a series of disadvantages,such as having ghosts while detecting,the long time needed to eliminate ghosts and poor adaptability.While in the frequently changed background,the algorithm has different problems,like misjudgment in the forground detection,and too many noise pixels caused by foreground interference.To solve these problems,the paper proposes an optimized algorithm,which is based on the basic framework of Vi Be and introduces the adaptive radius threshold establised by dispersion coefficient,the ghost model,the secondary judgment and target extraction.The algorithm can improve its adaptabilty to dynamic backgrouds,accelerate the elimination of shaows and ghosts,better reduce the interference of ghosts on the background model,suppress the water surface area with large dynamic amplitude,and reduce the rate of misjudgment by eliminating the noise that appears in the entire scene Pixels.The experimental results prove that the algorithm presented in this paper can effectively extract foreground targets and has good anti-interference robustness.