基于移动终端传感器的室内地理围栏的研究

发布时间：2018-08-23 12:05

【摘要】：地理围栏技术(Geo-fencing)在信息推送、智能家居、考勤签到、儿童安全监控以及智慧医疗等领域有着十分重要的应用。地理围栏的核心是定位技术,GPS占主导的室外定位研究已相对成熟,但是在室内、地下通道等复杂易变环境下,无法通过GPS来提供高精度连续导航需求。大规模普及的移动终端已经成为应用最为广泛的导航终端设备,研究基于移动终端传感器的室内导航技术应用在室内地理围栏领域具有很好的前景。然而基于移动终端传感器的室内地理围栏的挑战有:移动终端传感器信号采样频率低和周围环境或者行人活动对传感器干扰大;持移动终端姿势繁杂、运动自由度高,移动终端方向传感器航向并不等同于行人航向;移动终端功耗和性能有限;室内虚拟地理围栏要求终端具备持续高精度导航能力。首先分析了各种定位技术的优劣势,充分考虑到上述移动终端传感器室内地理围栏的挑战,本文采用行人航迹推算方法,并且把它应用在室内地理围栏模型中,通过分析移动终端传感器信号特征,依据行人运动生理学特性,推算行人持移动终端在各种应用场景下的步频、步长和航向,计算行人在室内平面活动位移,'实时监测行人是否离开虚拟地理围栏区域;其次针对传统行人航迹航向算法初始化校准、误差动态补偿、亚稳定场景下误差模型等弊端,不适用于室内复杂环境中姿势繁杂和运动高度自由的行人移动终端,本文在调研室内平面98%区域布局呈现规则的方角特性后,提出通过小波变换分析移动终端方向传感器信号特征,然后使用神经网络半监督学习模型预测用户在各种使用场景中的航向角方法。实验结果表明基于移动终端传感器的室内地理围栏解决方案可行,当用户在室内虚拟围栏区域活动15～60分钟内时,地理围栏的实时准确率达到93.5%,当延时3秒激活通知用户事件时,地理围栏的准确率达到98%,而且改进后的航向预测精确度达到96.6%。
[Abstract]:Geographic fence technology (Geo-fencing) has important applications in the fields of information push, smart home, attendance check in, child safety monitoring and intelligent medical treatment. The core of the geographical fence is the positioning technology. The research on GPS-dominated outdoor positioning has been relatively mature, but in the complex and changeable environment, such as indoor and underground passage, it is impossible to provide high precision continuous navigation through GPS. Large-scale mobile terminal has become the most widely used navigation terminal equipment. The research of indoor navigation technology based on mobile terminal sensor has a good prospect in the field of indoor geographical fence. However, the challenges of indoor geographic fence based on mobile terminal sensor are: low sampling frequency of mobile terminal sensor signal, large disturbance to sensor by surrounding environment or pedestrian activity, complex posture of holding mobile terminal and high degree of freedom of movement. The heading of mobile terminal direction sensor is not equal to pedestrian heading; the power consumption and performance of mobile terminal are limited; the indoor virtual geographic fence requires the terminal to have continuous high precision navigation capability. Firstly, the advantages and disadvantages of various positioning techniques are analyzed, and the challenges of the indoor geographic fence of the mobile terminal sensor are fully considered. In this paper, the pedestrian track estimation method is adopted, and it is applied to the indoor geographical fence model. According to the physiological characteristics of pedestrian movement, the step frequency, step size and heading of the mobile terminal in various application scenarios are calculated by analyzing the signal characteristics of the sensor in the mobile terminal. Calculating the pedestrian displacement in the indoor plane and monitoring whether the pedestrian leaves the virtual geographic fence in real time; secondly, aiming at the disadvantages of the traditional pedestrian course algorithm initialization and calibration, the error dynamic compensation, the error model under the sub-stable scene, and so on. It is not suitable for pedestrian mobile terminals with complicated posture and high freedom of movement in complex indoor environment. In this paper, wavelet transform is used to analyze the signal features of the mobile terminal direction sensor, and then the neural network semi-supervised learning model is used to predict the course angle of the user in various usage scenarios. The experimental results show that the indoor geographic fence solution based on mobile terminal sensor is feasible. When the user moves in the indoor virtual fence area within 1560 minutes, The real-time accuracy of the geographic fence reaches 93.5. when the delay of 3 seconds activates notifying the user, the accuracy of the geographic fence reaches 98, and the precision of the improved course prediction reaches 96.6.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9

【参考文献】