Research and Implementation of Indoor Positioning Algorithm for Personnel Based on Deep Learning

A real-time indoor position algorithm based on deep learning theory for many complicated situations is proposed to satisfy the current demands for collection of position information efficiently. Firstly, the video images captured by the camera in real time are input into the network, ZCA (Zero-phase Component Analysis) whitening preprocessing is used to reduce the feature correlation and reduce the network training complexity. Secondly, deep network feature extractor is constructed based on convolution, pooling, multi-layer sparse auto-encoder. Then, the extracted features are classified by the Softmax regression model. Finally, the collected feature is accurately identified by the face recognition module. The algorithm is evaluated on the Indoor Multi-Camera data set, the experimental results are expected to improve the positioning accuracy greatly and implement indoor precise positioning.