基于自适应混合高斯模型的运动目标检测

随着智能视频监控应用领域的不断拓展,为了在各种复杂场景下都能快速稳定地检测出运动目标,众多研究者提出了很多解决方法。鉴于背景减除法自身在算法理论的复杂度小,且对不同场景适应性强的优点,成为运动目标检测研究领域的基础方法。而在背景减除法的各种背景建模方法中,混合高斯模型方法以其适应性强、灵活高效的优点成为该领域的经典方法。详细介绍了背景更新机制中常用的混合高斯模型,并据此研究了基于自适应更新率的高斯背景建模方法,经过对比试验,论证了改进方法在建模速度上的大幅度提升。

Moving Target Detection Based on Adaptive Gaussian Mixture Model

With the development of intelligent video surveillance,many solutions were proposed to detect moving targets from complex scenarios quickly and stably.In view of its advantages on small theory complexity and strong adaptability to different scenarios,background subtraction method became the basic method for motion detection.In the background modeling methods of background subtraction,Gaussian mixture model is the classical method in the field for its adaptability,flexibility,efficiency.This paper present Gaussian mixture model in background updates mechanism,then researched a new Gaussian background modeling method based on adaptive update rate.After a lot of comparative tests between the typical algorithm and the new method,it can be demonstrate the new method greatly enhances the speed of modeling.