摘要
针对WiFi在室内定位中信号波动性较大、地磁指纹存在误匹配等问题,提出一种基于BP神经网络的WiFi/地磁融合定位方法。该方法通过Z-score标准化消除WiFi、地磁数据不同量纲级的影响,同时,选取Tanh函数替代Sigmoid函数作为BP神经网络的激活函数,改善深度学习中梯度消失、梯度爆炸等问题。离线阶段,将处理后的WiFi、地磁数据作为输入层对改进的神经网络进行学习训练,在线阶段,将训练好的BP神经网络用于智能手机的定位。实验结果表明,文中提出的定位方式较单一传感器的定位方式整体定位精度提升约为14%。
Aiming at the problems of WiFi signal fluctuation in indoor positioning and the mismatch of geomagnetic fingerprints,this paper proposes a WiFi/geomagnetic fusion positioning method based on BP neural network.This method eliminates the influence of different dimension levels of WiFi and geomagnetic data through Z-score standardization.At the same time,the Tanh function is selected to replace the Sigmoid function as the activation function of the BP neural network to improve the problems of gradient disappearance and gradient explosion in deep learning.In the offline stage,the processed WiFi and geomagnetic data are used as the input layer to learn and train the improved neural network.In the online stage,the trained BP neural network is used for smartphone positioning.The experimental results show that the positioning method proposed in this paper improves the overall positioning accuracy by about 14%compared with the positioning method based on a single sensor,and has better positioning performance.
作者
杨朝永
赵冬青
贾晓雪
张乐添
赖路广
程振豪
YANG Chaoyong;ZHAO Dongqing;JIA Xiaoxue;ZHANG Letian;LAI Luguang;CHENG Zhenhao(Information Engineering University,Zhengzhou 450001,China)
出处
《测绘工程》
2023年第3期14-18,共5页
Engineering of Surveying and Mapping
基金
国家自然科学基金资助项目(41774037,42104033)。