双层PSO-ELM融合室内定位算法  

Double-Layer PSO-ELM Fusion Indoor Positioning Algorithm

作　　者：徐岩 李宁宁 Xu Yan;Li Ningning(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学电气自动化与信息工程学院,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2021年第1期61-68,共8页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(61372145)。

摘　　要：随着基于位置服务需求的增长,室内定位成为国内外学者研究的重点领域.研究发现采用多传感器信息融合方法可以提高定位准确度,目前人们普遍认为利用多传感器的互补特性,结合各融合算法提升导航系统的整体精度是室内定位领域未来的发展趋势.本文提出一种基于双层粒子群极限学习机(PSO-ELM)神经网络的融合视觉和惯性信息的室内定位算法.第1层粒子群极限学习机(PSO-ELM)引入图像模糊判断来解决采集图像模糊时视觉定位算法误差大的问题,并计算出全局最优仿射变换矩阵作为粒子群极限学习机(PSO-ELM)的输入.同时,提出了一种基于视觉静态反馈和惯性特性的漂移校正方法来有效控制惯性导航系统(INS)的误差累积.第2层粒子群极限学习机(PSO-ELM)神经网络用于融合第1层粒子群极限学习机(PSO-ELM)获得的视觉定位结果和漂移校正后获得的惯性定位结果.将本算法所得融合后的定位结果分别与改进后的惯性定位结果和视觉定位结果进行比较,实验结果表明融合后的效果要优于单一算法的实验效果,定位精度和稳定性均得到提升.同时通过对比实验证明了本算法在存在外界干扰时也能保持良好的定位精度,具有较强的鲁棒性.With the growing demand for location-based services,indoor positioning has become a priority for researchers worldwide.This study shows that the fusion approach of multi-sensor information can improve the positioning accuracy.At present,it is widely assumed that future development trend in the indoor positioning field involves using the complementary characteristics of multi-sensors and merging of the fusion algorithms to improve the overall accuracy of the navigation system.This paper proposes an indoor positioning algorithm based on a double-layer PSOELM neural network to fuse vision and inertial information.The first layer of PSO-ELM introduces image fuzzy judgment to solve the problem of large error in the visual positioning algorithm when the captured image is blurred;the global optimal affine transformation matrix is calculated as the PSO-ELM input.At the same time,a drift correction method based on visual static feedback and inertial characteristics is proposed to effectively control error accumulation in the inertial navigation system.The second layer PSO-ELM neural network is used to combine the results of the visual positioning obtained by the first layer PSO-ELM and the results of inertial positioning obtained after the drift correction.Experimental results show that the proposed method performs better than the improved inertial positioning and visual positioning methods,enhancing both positioning accuracy and stability.The comparison experiments show that even in the presence of external interference,the proposed algorithm can maintain good positioning accuracy and robustness.

关 键 词：室内定位 数据融合 视觉导航系统 惯性导航系统 粒子群优化极限学习机 

分 类 号：TN916.72[电子电信—通信与信息系统]