基于WiFi位置指纹室内定位技术的研究

发布时间：2018-03-21 03:56

本文选题：WiFi　切入点：位置指纹定位　出处：《南京信息工程大学》2017年硕士论文　论文类型：学位论文

【摘要】：本文在定位的离线建立指纹库阶段,提出了基于正态检验的指纹库优化算法。该算法首先判断测量信号样本是否符合正态分布;其次,对符合正态分布的信号样本用正态分布函数估计其概率密度,反之则用核函数;最后选取大概率信号的均值录入指纹库中。实验表明,该方法准确的估计了室内环境中WiFi信号采集样本的分布情况,有效剔除了信号样本中的奇异值,建立了高精度的指纹库,从而提高系统的定位精度,与常用的均值模型法以及正态模型法相比较,平均定位误差分别减低了 15.62%,8.46%。本文在实时定位阶段提出了混合定位算法,即在传统的K加权近邻算法(WKNN)中引入PSO算法,此算法主要分为两个步骤,第一步:通过WKNN算法得到估算位置;第二步:用PSO算法对估算位置进行优化。实验结果显示,定位精度较K加权近邻算法提高了 19.89%,定位误差小于2m的概率为75%,大于WKNN算法的66.67%。因此本文所述混合定位算法的估算结果更加接近实际位置,同时也具有更高的稳定性。本文在研究基于WiFi的位置指纹定位算法的同时,设计开发了“指纹数据采集器”,该应用程序是基于Android在Eclipse+ADT+Android SDK环境用Java语言进行开发的,具有指纹数据采集与存储功能。
[Abstract]:In this paper, a fingerprint database optimization algorithm based on normal test is proposed in order to determine whether the measured signal samples conform to normal distribution or not. The normal distribution function is used to estimate the probability density of the signal samples in accordance with the normal distribution, and the kernel function is used to estimate the probability density. Finally, the mean value of the large probability signal is selected to be input into the fingerprint database. The experimental results show that, This method accurately estimates the distribution of WiFi signal collection samples in indoor environment, effectively removes the singular values from the signal samples, and establishes a high precision fingerprint database, thus improving the positioning accuracy of the system. Compared with the normal model method and the mean value model method, the average positioning error is reduced by 15.62% and 8.46 respectively. In this paper, a hybrid localization algorithm is proposed in the real time localization stage, that is, the PSO algorithm is introduced into the traditional K-weighted nearest neighbor algorithm (WKNN). The algorithm is divided into two steps: the first step is to get the estimated position by WKNN algorithm, and the second step is to optimize the estimated position by using the PSO algorithm. The experimental results show that, The accuracy of location is 19.89 higher than that of K-weighted nearest neighbor algorithm. The probability of location error less than 2m is 75, larger than that of WKNN algorithm 66.67.Therefore, the estimated result of the hybrid location algorithm described in this paper is closer to the actual location. At the same time, it also has higher stability. In this paper, the location fingerprint location algorithm based on WiFi is studied, and a fingerprint data collector is designed and developed. The application program is developed with Java language in Eclipse ADT Android SDK environment based on Android. Fingerprint data acquisition and storage function.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN92

【参考文献】