基于改进的混合高斯模型的红外运动目标检测

Infrared Moving Object Detection Based on Improved Gaussian Mixture Model

传统混合高斯模型中背景容易留下运动"虚影",同时在噪声或目标区域对比度低时会导致提取目标区域时出现断裂和空洞的现象,针对这些问题在混合高斯方法中赋予图像中运动和静止区域不同的背景更新速率,并充分利用混合高斯模型中的背景和前景信息,将背景减除的结果与高斯建模中的前景图像按照一定比例融合获得目标图像。实验结果表明:改进后的混合高斯模型运动目标检测方法,能够克服传统高斯模型目标检测中存在的问题,从复杂的背景中较完整的提取出运动目标,且具有一定的抗噪能力。

Traditional Gaussian mixture modeling is likely to cause motion artifact, while the noise or regions with low contrast will bring gaps or holes when extracting the targets. In view of above problems, this paper assigns different background updating rate to the motion and static regions, and make the most of the background and foreground information, and then the final results are fusion with Gaussian background subtraction and the foreground image from Gaussian mixture modeling with a certain proportion. The experimental results demonstrate that improved Gaussian mixture modeling for infrared moving targets detection can overcome problems existing in traditional algorithm. It could extract the moving object completely from complex backgrounds, and also has a good anti-noise capability.