Abstract: The underwater dynamic navigation based on sectional observation generates multi-source and heterogeneous data, creating crossed or forked tracks caused by asynchronous time delay and unknown system errors, which brings on difficulties to describe the continuous process in a integrated way and identify the local characteristic points. So the effective fusion method is significant for the complex motion performance evaluation of underwater vehicle. Aiming at the problem, the polynomial constraint fusion method(PCF) and spline function fusion method(SFF) for the track data base on sectional observation are proposed, which reconstructed the entire underwater track by function and solved the problem of discontinuous dynamic parameter sequence and fuzzy data in overlapped section. The numerical simulations show that both PCF and SFF could capture the main characteristics of underwater dynamic motion and obtain a complete and continuous track. By comparing the results of PCF, SFF and general data fusion method(GDF), the track derived by SFF has the similar smoothness and continuity as the one derived by PCF, and its value is closer to the track derived by GDF which relies on weights, and also the Runge phenomenon created by PCF is effectively avoided. The statistical results of RMSE indicate that the differences in position RMSE between SFF and GDF are 0.05m by data I and 0.02m by data II respectively, and that between PCF and GDF are 0.19m by data I and 0.28m data II respectively. The complementary and integration of SFF based on its reconstruction performance and PCF based on its extrapolation performance make it possible to accurately identify the motion characteristic points. Verified by sea trial, the SFF was used to obtain the re-analysis track in the observation section with the time accuracy superior to 1% at the characteristic points, and the PCF was used to obtain the predicted track in the subsequent section with the time accuracy superior to 10%. The proposed method shows application values for the multi-source and heterogeneous data processing and analysis in complex underwater motion, and is also useful for real-time estimation and assistant decision of underwater navigation.