Wireless positioning method based on deep learning

The invention provides a wireless positioning method based on deep learning. The wireless positioning method comprises the following steps: acquiring channel state information through a channel simulator and preprocessing the channel state information; constructing a CSI image based on the extracted amplitude and phase information; in an offline stage, inputting the CSI image constructed at the reference point into a convolutional neural network for learning to obtain a classification model; in the online stage, a CSI image constructed at a test point is input into a convolutional neural network model obtained through training, and after a rough positioning result of a target is obtained, a final positioning result is obtained in combination with a probability-based coordinate calculation method. According to the method, the positioning problem is converted into the image classification problem based on the CSI image and the convolutional neural network, and the target positioning precision can be improved. 本