基于Walsh-hadamard变换的混合高斯背景模型

Mixture of Gussian Background Model Based on Walsh-hadamard Transformation

混合高斯背景建模技术及其改进算法有效地解决了光照渐变和周期性动态背景条件下的运动目标检测问题。但是,当场景存在光照突变时检测准确率将大大降低。为此,提出一种基于Walsh-hadamard变换的混合高斯背景模型。该模型从颜色、边缘和纹理来描述背景区域的特征,较全面地刻画出了背景的本质属性,同时对前景目标有着非常好的区分力。实验证明:该方法较有效地解决了传统GMM中存在的问题,并为后续视觉分析打下了基础。

The mixture of Gaussian background model technology（GMM） and its improved algorithm have been widely used in the moving object detection and achieved better performance.The algorithm can effectively solve the motion object detection under the condition of the light gradient and Periodic dynamic background.However,when the scene illumination changed,the detection accuracy will be greatly reduced.To overcome the weakness,in the paper,a new modeling method based on Walshhadamard transformation is proposed.The modeling described the features of background region by the color,texture and edge,comprehensively characterized the nature of the background property,and had a very good capacity of distinguishing goals from the background.Experiment results show that our new method solved the above problems effectively,which will be a baseline for high-level vision analysis.