一种运动恢复结构和航位推算结合的室内行人视觉定位方法

发布时间：2018-07-03 17:56

本文选题：室内定位 + 手机传感器　；参考：《地球信息科学学报》2017年06期

【摘要】：商业和工业领域中,室内行人、车辆、机器人的位置信息正逐渐成为人们关注的热点,并随之产生了大量的室内定位技术和方法,如使用无线信号、地磁、超宽带和超声波等方式进行室内定位。然而,目前的这些室内定位方法大多需要额外辅助设备的支撑,增加了定位成本和硬件开销。视觉定位作为一种目前较为流行的定位方式,具有实施成本低、不依赖任何外界辅助设备等优势。其中,构建带有位置标签的图像数据库是视觉定位方法的关键环节,而传统的构建图像数据库方法人力开销大、时耗长。因此,本文提出一种运动恢复结构(SFM)和航位推算结合的视觉定位方法,能够快速构建图像位置数据库、大大降低人力开销。该方法主要包括2个阶段:离线阶段和在线阶段。离线阶段主要实现图像序列位置的自动标注,通过采集行走路线上的手机内置传感器信息和视频信息,提出一种多约束图像匹配方法用于视频图像的连续匹配,将匹配结果用于SFM方法,可以得到相邻图像间的运动角度,使用行人航位推算(PDR)方法标注图像序列的轨迹坐标。在线阶段使用提出的图像匹配方法计算查询图像与数据库影像间的匹配点数量,将匹配点最多的K个数据库影像位置坐标加权平均作为查询图像的定位结果。最后,分别在2种典型的室内环境下进行实验,结果表明本文方法在离线阶段位置标注的平均误差为0.58 m,在线阶段图像匹配定位的误差范围在0.2~1.4 m。
[Abstract]:In the commercial and industrial fields, the position information of indoor pedestrian, vehicle and robot is gradually becoming the focus of attention, and a large number of indoor positioning techniques and methods have been generated, such as the use of wireless signals, geomagnetic, Ultra-wideband and ultrasonic wave and other methods for indoor positioning. However, most of these indoor positioning methods need the support of additional auxiliary equipment, which increases the location cost and hardware cost. As a popular positioning method, visual positioning has the advantages of low implementation cost and no dependence on any external auxiliary equipment. Among them, the construction of image database with location label is the key link of the visual positioning method, while the traditional method of building image database has a large human cost and time consuming. Therefore, this paper proposes a visual location method combining motion recovery structure (SFM) with dead-reckoning, which can quickly construct image location database and greatly reduce manpower cost. The method mainly includes two stages: offline and online. In the off-line stage, the automatic tagging of image sequence position is realized. By collecting sensor information and video information from mobile phone, a multi-constraint image matching method is proposed for continuous video image matching. Applying the matching results to SFM method, the motion angle between adjacent images can be obtained, and the track coordinates of image sequences can be marked by using the method of pedestrian carrier reckoning (PDR). In the online stage, the number of matching points between the query image and the database image is calculated by using the proposed image matching method, and the weighted average of the location of K database images with the most matching points is taken as the location result of the query image. Finally, experiments are carried out in two typical indoor environments. The results show that the average error of the method is 0.58 m in off-line phase and 0.21.4 m in online stage.
【作者单位】：深圳大学土木工程学院深圳市空间信息智能感知与服务重点实验室;河南财经政法大学资环与环境学院;武汉大学测绘遥感信息工程国家重点实验室;中山大学地理科学与规划学院综合地理信息研究中心;
【基金】：国家自然科学基金项目(41301511、41371377、41371420、41501424) 国家重点研发计划项目(2016YFB0502203) 深圳市科技计划项目(JCYJ20140418095735587) 深圳大学科研启动基金资助项目(2016064)
【分类号】：TP391.41

【相似文献】