基于虚拟信息库的粒子滤波指纹库定位算法

FINGERPRINT DATABASE POSITIONING ALGORITHM USING PARTICLE FILTER BASED ON VIRTUAL INFORMATION DATABASE

指纹库定位算法的关键在于根据不同参考节点的接受信号强度指示RSSI(Receive signal strength indication)建立有效指纹信息数据库。传统的方法是在定位区域内标定多个信息采样点,而大量样本数据的采集会导致算法离线训练阶段工作量增大。Zigbee传感器网络平台下,综合考虑了目标自身对信号的干扰以及节点数对函数逼近能力的影响,利用信号强度的非线性特性,提出了一种基于多项式分区插值的虚拟指纹库生成方法;同时使用粒子滤波对预估计的结果进行处理,以解决RSSI不规则分布问题。实验结果表明该方式可以快速、简捷地生成细粒度的定位信息数据库,提高了定位精度。

The key of fingerprint database positioning algorithm is to establish effective fingerprint information database according to RSSI of different reference nodes.Traditional method is to calibrate multiple information sampling points within the positioning area,but the collec-tion of a large number of sample data will lead to the algorithm increasing its workload at off-line training stage.On Zigbee sensor networks platform,considering comprehensively the interference of the target itself on signals and the influence of node number on the function approxi-mation ability,and making use of the nonlinear feature of signal intensity,we propose a polynomial partition interpolation-based virtual finger-print database generation method;meanwhile we employ particle filter to process the pre-estimated results to solve the problem of irregular RS-SI distribution.Experimental result shows that this method can quickly generate in simple way the fine-grained localisation information data-base,and improves the positioning accuracy as well.