一种基于脉冲耦合神经网络的运动目标提取方法  

A method for the extraction of moving objects based on pulse-coupled neural network

作　　者：叶谷平[1] 王伟东[1] 

机构地区：[1]湛江师范学院信息与教育技术中心,湛江524048

出　　处：《四川大学学报（自然科学版）》2014年第1期69-75,共7页Journal of Sichuan University(Natural Science Edition)

基　　金：广东省自然科学基金项目(S2012010009759);湛江市科技攻关计划项目(2012C3103022)

摘　　要：运动目标提取是视频分析和理解的基础,在视频监控应用中具有重要的意义.本文在帧间差分法运动检测的基础上,研究了运动目标的提取方法.针对现有方法存在的对帧间差结果噪声敏感和低效的问题,引入了一种脉冲耦合神经网络PCNN(Pulse Coupled Neural Network)方法.通过一种特别的映射方法,该PCNN网络模型能有效地抑制噪声,并能高效地提取运动目标.实验结果证明,该PCNN方法比现有一些方法具有更高的精度和更快的速度.The moving object extraction is the basis of video analysis and understanding, and it is signifi- cant for video surveillance applications. This paper mainly focus on the research of moving target extrac- tion based on the frame differencing. Because some existing methods are sensitive to the noise and ineffi-eient, a pulse coupled neural network （Pulse Coupled Neural Network, PCNN） method is introduced in this paper. By using a special mapping scheme, the PCNN method is more efficient and can effectively suppress the noise. Some experimental results show that the PCNN method has a higher accuracy and faster speed than some other methods.

关 键 词：运动目标提取 帧差法 脉冲耦合神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]