基于置信传播的视频运动目标检测
Video moving object detection based on belief propagation
郭春生 1张大状1
作者信息
- 1. 杭州电子科技大学通信工程学院,浙江杭州310018
- 折叠
摘要
视频运动目标的检测是目标分类和行为理解等后续处理的基础.本文提出一种复杂背景条件下基于置信传播(BP)的视频运动目标检测算法.该算法依据视频背景的空间和时间相关性,分别定义了视频背景的空间和时间能量函数,完成视频运动目标的优化检测.利用置信传播算法在相邻像素之间的信息交换来实现能量最小化,不同背景条件下的实验结果验证了该算法的有效性和准确性.
Abstract
Video Moving Object Detection is the foundation of the object classification and behavior understanding follow-up processing.In this paper,an algorithm based on the belief propagation (BP) is proposed to detect the video moving object in complex backgrounds.According to the correlation of spatial and temporal of video background,the spatial and temporal energy function are defined respectively,then it minimize the energy function,accomplish the optimal detection of video moving objects.The minimization of energy is achieved by information exchange between adjacent pixels by means of BP algorithm.The availability and accuracy of the algorithm are verified by experimental results in under the different backgrounds.
关键词
运动目标检测/置信传播/信息交换/背景更新Key words
moving object detection/belief propagation/information exchange/background update引用本文复制引用
基金项目
浙江省自然科学基金(LY12F01003)
出版年
2013
NETL
NSTL
维普
万方数据