基于预测的WMSNs目标跟踪协作处理方法  

Forcasting Based Cooperative Processing Method for Target Tracking in WMSNs

作　　者：孙伟[1] 金心宇[1] 张昱[1] 唐军[1] 

机构地区：[1]浙江大学信息科学与工程学院,杭州310027

出　　处：《传感技术学报》2009年第8期1175-1181,共7页Chinese Journal of Sensors and Actuators

基　　金：浙江省科技计划项目资助(2005CS31001)

摘　　要：提出一种基于预测的无线多媒体传感器网络(WMSNs)目标跟踪协作处理方法。建立全向感知模型,由图像准确定位目标空间坐标位置。组织传感器节点休眠,基于自回归移动平均模型(ARMA)和径向基函数网络(RBFN)进行目标运动轨迹预测,唤醒节点完成协作处理。使用形状上下文的匹配(Shape Matching)算法,避免多节点冗余信息传输。仿真结果表明,基于ARMA-RBFN预测的唤醒机制使能耗降低25倍以上,Shape Matching可减少传输数据量20%。A forcasting based cooperative processing method for target tracking in Wireless Multimedia Sensor Networks is proposed. Through establishing sensor model, target localization is performed by image sensor nodes. Combining autoregressive moving average（ARMA）model and radial basis function networks （RBFNs） to perform motive target position forecasting during target tracking, based on which, cooperative wake-up and decision is realized among multi-nodes. By Shape Matching, redundant image data transmission is avoided. The simulation results have shown that wake-up mechanism based on ARMA-RBFN improves the performance of energy consumption by 25 times. Data transmission scale is reduced by 20 % after using Shape Matching.

关 键 词：无线多媒体传感器网络 目标跟踪 运动轨迹预测 协作处理 

分 类 号：TN919[电子电信—通信与信息系统]