摘要
运动视目标检测是视频信息处理的重要研究课题之一.本文提出了一种基于高斯混合模型邻域信息融合的海面运动目标检测算法.该算法融合了背景差分和背景邻域信息差分,充分利用同一幅图像的像素邻域信息得到运动目标的种子点,认为高斯背景差分图像中包含种子点的连通区域为真实前景目标.实验表明,该方法可以避免背景模型在构建或更新阶段对场景的表征不足或错误而造成的误检,对强光下的海杂波也有良好的抑制作用,且对不同的气候环境有较好的鲁棒性.
Moving object detection is one of the most important research topics in the video information processing. This paper proposes a new method of tracking the moving maritime objects in video sequences based on neighboring information fusion using Gaussian mixture model. It fuses the background difference and neighboring background information difference,makes full use of the information of neighboring pixels to get the moving seed. The experimental results show that, when constructing and updating the background model,this method is able to avoid false detection from inadequate or incorrect background characterization completely. The proposed motion detection also restrains illumination from ocean wave and is robust to various climates.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2007年第5期641-645,共5页
Journal of Xiamen University：Natural Science
基金
国家自然科学基金(60175008)
国家创新研究群体资助项目(60024301)
厦门大学985二期信息创新平台项目资助
关键词
邻域信息融合
高斯混合模型
运动目标检测
neighboring information fusion
Gaussian mixture model
moving object detection
作者简介
通讯作者：chli@xmu．edu．cn
