图像辅助的射频指纹定位系统的设计与实现

发布时间：2018-02-11 09:33

本文关键词： 室内定位射频指纹定位算法位置指纹数据库图像辅助　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着移动智能终端和无线网络的快速普及与发展,基于位置的服务受到了广泛的关注。基于WLAN的室内定位技术由于只需测量接收信号强度,无需额外的无线基础设施,使其作为一种室内定位的解决方案而被广泛应用。但是随着室内环境越来越复杂,接收信号强度表现出高度的复杂性与多变性,因而仅利用单一信号强度进行定位,其定位精度严重受到影响。针对这一问题,本文利用图像辅助进行射频(RF)指纹室内定位,并在提高定位精度和降低计算量方面做了详细的研究与分析。首先,基于契比雪夫距离提出了一种改进的K近邻法。研究发现,该定位算法的平均误差在2.10 m,基本可以达到K近邻法的定位精度。由于计算契比雪夫距离只需加减法运算,与K近邻算法相比,其计算复杂度较低。同时,该算法的定位精度高于传统的MaD算法、CBD算法和VaS算法。其次,基于WLAN的射频指纹室内定位技术,提出了均匀图像辅助的射频指纹室内定位系统。研究图像采集装置的摆放位置对定位精度的影响以及计算工作量的变化情况。测试结果表明,图像采集装置位于定位区域正前方时,定位误差为1.61m;位于定位区域正上方时,定位误差为1.26 m,后者比前者的定位精度提高了 22%。两种情况下的平均定位误差分别比仅利用射频指纹定位的提高了约0.55m和0.85 m。由于使用均匀图像辅助,用户只需与其所在区域内的参考指纹进行匹配计算,因而计算工作量也大幅减少。最后,考虑到图像采集装置位于教室前方的普遍性。因此,为提高装置位于教室前方的均匀图像辅助的射频指纹定位算法的精度,提出了基于非均匀指纹库的均匀图像辅助技术。基本思想是建立非均匀指纹数据库,其中的指纹密度随着距离的增加而增加。基于均匀图像辅助的射频指纹定位算法,利用均匀和非均匀的指纹数据库对用户位置进行匹配计算。测试结果表明,当图像采集装置位于教室前方时,在均匀图像辅助下,利用非均匀指纹库的算法精度达到了1.35m,比利用均匀指纹库提高了约0.30 m。当然,计算工作量也会随参考指纹数增加而相应地增加。图像辅助的射频指纹室内定位技术的平均定位精度可达1.26m。与仅利用单一信号的射频指纹定位相比,该技术的平均定位精度提高了 40%,而计算工作量降低了 75%。
[Abstract]:With the rapid popularization and development of mobile intelligent terminals and wireless networks, location-based services have attracted wide attention. Indoor positioning technology based on WLAN only needs to measure the strength of received signals, so there is no need for additional wireless infrastructure. It is widely used as a kind of indoor positioning solution. However, as the indoor environment becomes more and more complex, the received signal intensity shows a high degree of complexity and variability, so it only uses a single signal intensity to locate. The positioning accuracy is seriously affected. Aiming at this problem, this paper makes use of image assistant to carry out RF fingerprint indoor positioning, and makes detailed research and analysis in improving the location accuracy and reducing the calculation amount. Based on Chebyshev distance, an improved K-nearest neighbor method is proposed. It is found that the average error of the algorithm is 2.10 m, which can basically achieve the accuracy of K-nearest neighbor method. Compared with K-nearest neighbor algorithm, its computational complexity is lower. At the same time, the location accuracy of this algorithm is higher than that of traditional MaD algorithm and VaS algorithm. Secondly, RF fingerprint indoor location technology based on WLAN is proposed. A uniform image-assisted radio-frequency fingerprint indoor positioning system is proposed. The influence of the location of the image acquisition device on the positioning accuracy and the change of the calculation workload are studied. The test results show that, When the image acquisition device is in front of the location area, the positioning error is 1.61m; when the image acquisition device is located directly above the location area, The positioning error is 1.26 m, and the accuracy of the latter is 22% higher than that of the former. The average positioning error in both cases is about 0.55 m and 0.85 m higher than that using radio-frequency fingerprint only. Users only need to match the reference fingerprints in their area, so the amount of computing work is greatly reduced. Finally, considering the universality of the image acquisition device in front of the classroom, In order to improve the accuracy of the radio-frequency fingerprint location algorithm in front of the classroom, the uniform image assistant technology based on the non-uniform fingerprint database is proposed, and the basic idea is to establish the non-uniform fingerprint database. The fingerprint density increases with the increase of distance. Based on the radio-frequency fingerprint location algorithm aided by uniform image, the user's location is matched by uniform and non-uniform fingerprint database. The test results show that, When the image acquisition device is in front of the classroom, with the help of uniform image, the accuracy of the algorithm using non-uniform fingerprint database is 1.35 m, which is about 0.30 m higher than that of using uniform fingerprint database. The calculation workload will also increase with the increase of reference fingerprint number. The average positioning accuracy of radio-frequency fingerprint indoor location technique can reach 1.26m.Compared with the radio-frequency fingerprint location using only a single signal, The average positioning accuracy of this technology is increased by 40%, and the calculation workload is reduced by 75%.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN925.93

【相似文献】