摘要
提出了一种复杂场景视频序列中运动目标精确检测及提取的改进算法,该算法首先采用混合高斯模型(简称GMM)对背景及前景建模快速地实现前景运动区域提取,然后结合目标帧间相关性和随机噪声帧间无关的特点采用时间滤波(Tem-poral Filter)法和数学形态学进行后处理。实验结果表明本文所采用的改进算法能准确的提取运动目标滤除动态噪声,提高了检测鲁棒性,对复杂干扰场景下的实时运动目标检测得到了较令人满意的效果。
An improved algorithm is proposed to detect and extract motion targets accurately in complex video sequences. First,it adopts the method of Gaussian Mixture Model （GMM） to extract the moving regions of prospects quickly. Then,it combines the features of the target frame relevance with random noise unrelated and uses the method of temporal filter and mathematical morphology to make an extra treatment. The experimental results show that the new method proposed in this paper can perfectly detect dynamic objects and filter dynamic noise,improving the robustness of testing; It has been more satisfactory results for the real-time motion detecting under the interference of complex scenes.
出处
《传感技术学报》
CAS
CSCD
北大核心
2009年第8期1146-1149,共4页
Chinese Journal of Sensors and Actuators
关键词
运动目标检测
混合高斯模型
时间滤波
数学形态学
Moving object detection, Gaussian mixture model, temporal filter, mathematical morphology
