图像分析与传感器融合的室内导航定位系统设计及实践

Design and practice of indoor navigation and positioning system based on image analysis and sensor fusion

面对当前公共建筑愈发复杂庞大,居民对陌生环境下室内定位导航的需求日益高涨。传统基于无线网络传感器的方法在网络系统未知场所不适用,惯性传感器定位导航方法需要先验环境支持。因此,提出了一种基于双目图像分析与传感器融合的室内导航定位方法。通过角点检测、视觉残差计算、高斯球方向检测实现图像处理,使用紧耦合策略将视觉数据与惯导数据融合,采用边缘化先验法消除无关标量。基于相关原理设计室内导航定位系统,经过在工作室内测试,该系统能够良好建立起虚拟地图并提供相对准确的导航和定位服务,误差可达分米级。为陌生环境下的室内导航提供了一种新方法。

In the face of the increasingly complex and huge public buildings,residents'demand for indoor positioning and navigation in unfamiliar environments is increasing.Traditional wireless network-based sensor methods are not applicable in unknown places of network systems,and inertial sensor positioning and navigation methods need a priori environment support.Therefore,an indoor navigation and positioning method based on binocular image analysis and sensor fusion was proposed.Image processing was realized through corner detection,visual residual calculation and Gaussian sphere direction detection.Tight coupling strategy was applied to fuse visual data with inertial navigation data,and edge prior method was explored to eliminate irrelevant scalar.The indoor navigation and positioning system was designed based on the relevant principles.After testing in the working room,the system can build a virtual map well and provide relatively accurate navigation and positioning services,with an error of up to decimeter level.It provides a new method for indoor navigation in strange environment.