基于改进混合高斯模型的运动物体检测研究

发布时间：2018-07-04 21:48

本文选题：运动物体检测 + 混合高斯模型　；参考：《计算机应用研究》2017年06期

【摘要】：针对传统高斯建模的初始化问题、参数值的计算依赖于先前所有帧和零散噪点较多等问题,提出了一种改进混合高斯模型的方法,即在初始化每个像素点时采用邻域特性与中值滤波相结合的方法,用来获取更接近实际的初始背景。同时对背景模型的更新提出了改进方法,在原有的背景排序基础上增加定时清零策略,使新加入的像素点能快速匹配。最后对特定区域的学习速率进行重新设定,再结合像素点的空间分布特性,达到消除零散噪点和部分空洞的目的。实验结果表明,与传统的混合高斯模型相比,该算法能准确地检测出运动物体,并对阴影和噪声有一定的抑制作用。
[Abstract]:In order to solve the initialization problem of traditional Gao Si modeling, the calculation of parameters depends on the number of previous frames and scattered noise. A method to improve the hybrid Gao Si model is proposed. That is to say, when initializing each pixel, the method of combining neighborhood characteristic with median filter is used to obtain the original background which is closer to the actual situation. At the same time, an improved method is proposed to update the background model, which adds a timing zero clearing strategy on the basis of the original background ranking, so that the newly added pixels can match quickly. Finally, the learning rate of a specific region is reset, and then the spatial distribution of pixels is combined to eliminate the scattered noise and partial voids. The experimental results show that compared with the traditional hybrid Gao Si model, the proposed algorithm can accurately detect moving objects and can suppress shadows and noises to a certain extent.
【作者单位】：中南大学信息科学与工程学院计算机科学与技术系;
【基金】：国家自然科学基金资助项目(60873188)
【分类号】：TP391.41

【相似文献】