A regression prediction algorithm based on deep learning was proposed to solve the problem of predicting height through footprint information in criminal investigation． Firstly，the original data was preprocessed to obtain the data set suitable for the deep learning regression model． Secondly，a new network architecture foot to height-net (FtH-Net) containing edge extraction and regression was proposed according to the characteristics of the footprint data． Finally，a prediction model with good performance was achieved by the data set based on the first two steps and regression network training． The experimental results show that，compared with the traditional one，the new method can greatly improve the accuracy of prediction while ensuring the generalization ability of the model，and the accuracy of prediction for people within 2cm height can reach 67%．