Abstract: Different methods of GPS height fitting have limitations respectively and the single fitting method does not have universality in different survey areas.Aiming at these problems,this paper mainly studies four kinds of GPS height fitting methods:polynomial curve fitting,multisurface function fitting,quadric surface fitting and BP neural network method. Combined with actual engineering projects, height fitting of the study area is carried out,experimental results are analyzed, and the fitting accuracy is evaluated.Based on the analysis and comparison of the results obtained by the four fitting methods, conclusions are as follows:All four methods meet the requirements of the fourthgradeleveling survey, and the fitting residuals are less than 3 cm;the accuracy of BP neural network fitting is higher than others;according to the characteristics and the scope of application in different fitting method,correct use of the fitting method is necessary condition to obtain high precision normal height.This study has corresponding application and reference value for GPS height fitting in different environments.