【作者】 罗鹏；

【导师】 董亚波；

【作者基本信息】 浙江大学 ， 计算机应用技术， 2016， 硕士

【摘要】 随着现代社会科技的发展和互联网技术的发展,室外定位技术也越来越成熟,谷歌地图、百度地图、高德地图的被大面积使用,人类从室外定位技术中获取到的方便也越来越多。受助于成熟的室外定位技术,人们可以很方便地查询到附近的各类商店服务,同时也可以很便捷地获取到各地的最佳路线,甚至公交路线、自驾车路线以及步行路线都一目了然。既然室外定位技术能给人类带来这么多的便利,那么,人们的视线开始投向室内定位技术,如何在室内精度定位物体位置成为了急需解决的问题。本文从大量实验入手,采集足够多的定位数据,通过函数拟合、神经网络和滤波预测的训练处理,提出了通过测量信号强度,训练出物体相对于信号基站的夹角,然后通过三角定位求出物体坐标,最后对位置坐标进行滤波预测的一整套定位方案。整个方法的具体工作如下：一、分别进行室外理想环境下定位实验、室内理想环境下定位实验和室内复杂环境下的定位实验。在室内外的理想环境下,本文使用函数拟合的方法进行定位分析,发现该环境下的角度平局定位精度达到5。,同时定位准确率达到70%。在室内复杂环境下时,函数拟合方法不再适用,于是本文采用了神经网络方法对数据进行训练,将180cm精度上的定位准确率提高到65%。二、为了进一步提高定位精度与定位准确率,本文在神经网络基础上运用滤波预测方法对数据进行修正。主要包括两种方式,其一为坐标线性预测算法,通过线性叠加BP神经网络与基本卡尔曼滤波器,并将神经网络输出向量作为滤波预测中的观测变量,从而实现对输出向量的滤波预测；其二为网络权值滤波算法,使用扩展卡尔曼滤波对神经网络的权值矩阵进行预测修正。两种方法对复杂环境下的定位准确率均有提升,其中坐标线性预测算法表现优秀,180cm精度上的定位准确率提高到70%。三、本文实现了基于双天线蓝牙基站、Android蓝牙信号采集模块与定位呈现模块和后台实现定位算法的服务器端的室内复杂环境定位系统。双天线蓝牙基站是在iBeacon蓝牙基站基础上改进的信号发生器,而Android蓝牙信号采集模块是采集蓝牙信号数据的载体,同时,一个Android移动端其实也代表着一个用户,服务器端则是完成定位数据的关键部分,它将对从Android移动端发来的信号强度数据进行上述分析,计算出定位坐标发送给移动端,而移动端则根据发送回来的位置数据进行实时位置更新。整个系统包含信号数据采集、定位算法实现和定位结果呈现等模块,互相配合,为室内定位研究提供了一种实现方式。更多还原

【Abstract】 With the development of modern science technology and internet technology, the outdoor location technology such as GPS has been grown and mature. People get convenient from outdoor location technology by using Google Map, Baidu Map and Gaode Map. With the proven outdoor location technology, we people can inquire about the nearby shops and service, and people can also get the best path to destination include the bus path, self-driving path and foot path. Now that the outdoor location technology can bring so much profit to people, the indoor location technology should also benefit people in some ways. So, it is important that to find out the indoor position of something.This paper introduce the background and the research status of indoor location technology. And then this paper also introduce the three most important experiment. We collected enough data by this experiments and get the angle between the phone and Bluetooth station by neural network and filter forecasting. Then we can calculate the X-Y coordinates of the phone. The full work is as below:1. Research data was collected through positioning experiments in various conditions, including the outdoor experiment with ideal environment, indoor experiment with ideal environment and indoor experiment with complex environment. In the indoor experiment with ideal environment, the method of function fitting was applied to positioning analysis with the discovery that the accuracy of the angle draw positioning reached 5° and the positioning accuracy reached 70% in this kind of condition. In the indoor experiment with complex environment, the function fitting method was no longer applicable, so the neural network method was applied to train the data, improving the 180cm positioning accuracy to 65%.2.Filtering prediction method was applied to revise the data and to improve the positioning accuracy. Two methods were mainly applied. The first method was the linear superposition of neural network and the basic Kalman filter method. In order to realize the filtering prediction to the output variable, the author took the vector of neural network as the observed variables of filtering prediction. The second method was to extend the neural network of the Kalman filter, using the extended Kalman filter to predicate and revise the weight matrix of the neural network. Both methods can improve the positioning accuracy in the complex environment, especially the linear superposition of neural network and the basic Kalman filter method which can improve the 180cm positioning accuracy to 70%.3. In this paper, we make the system of indoor location which is made of Android client, Python server and iBeacon Bluetooth station come true. The iBeacon Bluetooth station is a double antenna station and the Android client is represent a user and the Python server is the most important part of the system which process the calculating of X-Y coordinates.更多还原