一种基于卷积神经网络的室内定位方法
An Indoor Positioning Method Based on Convolutional Neural Network
张丽 1董建 1孙长智 1刘成刚1
作者信息
- 1. 亳州学院 电子与信息工程系,安徽 亳州 236800
- 折叠
摘要
针对实际室内定位场景中的无线接入点信号不稳定引起定位精度低的问题,提出一种基于卷积神经网络的室内定位方法,该方法包括离线阶段和在线阶段,其中离线阶段主要完成对无线接入点信号采集,经过预处理后作为卷积神经网络模型的训练数据.在线阶段利用训练好的模型完成粗定位,估计位置所在的区域,最后利用加权k近邻算法计算精确的位置坐标.通过与SVR、KNN算法对比,结果表明,在二维平面回归定位问题中优于其他算法.
Abstract
According to the problem of low positioning accuracy caused by unstable wireless access point sig-nals in actual indoor positioning scenarios,an indoor positioning method based on convolutional neural network is proposed.The algorithm includes an offline stage and an online stage.The offline stage mainly completes the ac-quisition of wireless access point signals,which is used as the training data for the convolutional neural network model.The online stage uses the trained model to estimate the area where the location is located,and finally the weighted k-nearest neighbour algorithm is used to calculate the exact position coordinates.By comparing with SVR and KNN algorithms,the results show that it is superior to other algorithms in the two-dimensional planar regression positioning.
关键词
卷积神经网络/室内定位/加权k近邻算法/位置指纹算法Key words
convolutional neural network/indoor localisation/weighted k-nearest neighbour algorithm/lo-cation fingerprinting algorithm引用本文复制引用
基金项目
2022年安徽省大学生创新创业训练计划项目(S202212926009)
亳州学院质量工程项目(2023XJXM052)
出版年
2024
NETL
NSTL
万方数据