为实现在多楼层大面积的室内环境下获得精准定位信息，以芬兰Tampere大学的公开数据集为基础，建立基于Wi-Fi位置指纹算法模型。使用自动编码器模型和神经网络模型结合训练进行主动降维和主成分提取，进而采用分类训练构建改良后的楼层分类模型。使用改良后楼层分类模型楼层分类精确率达到92%，结果优于自动编码器模型和神经网络模型；神经网络模型结合加权KNN算法实现定位精度在4米左右，算法有效性高于传统Wi-Fi指纹室内定位算法。改良后楼层分类模型结合加权KNN算法，可作为大型室内定位最佳模型。
In order to obtain accurate location information in multi-floor and large area indoor environment, based on the open data set of Tampere University in Finland, a fingerprint algorithm model based on Wi-Fi position is established. The automatic encoder model and neural network model are combined with training to extract the active reduction and principal components, and then the improved floor classification model is constructed by classification training. The accuracy rate of floor classification using the improved floor classification model is 92%, which is better than that of automatic encoder model and neural network model. The location accuracy of neural network model combined with weighted KNN algorithm is about 4 meters, and the efficiency of the algorithm is higher than that of the traditional Wi-Fi fingerprint indoor location algorithm. The improved floor classification model combined with weighted KNN algorithm can be used as the best model for large indoor positioning.
Kolodziej, K.W. and Hjelm, J. (2006) Local Positioning Systems: LBS Applications and Services. CRC Press, Boca Raton, 101-158. https://doi.org/10.1201/9780849333491
[4]
Brunato, M. and Battiti, R. (2005) Statistical Learning Theory for Location Fingerprinting in Wireless LANs. Computer Networks, 47, 825-845. https://doi.org/10.1016/j.comnet.2004.09.004
