It’s hard for Gaussian mixture probability hypothesis density filter (GMPHDF) tacking nonlinear system target in radar networking. For solving this problem, construct Gaussian unscented mixture probability hypothesis density filter (GUMPHDF) method. Predict and update birth, spawn and existing targets with unscented filter(UF), getting Gaussian unscented mixture elements of each target, then mix together in merging and extraction procedure. The simulating verification demonstrates that the method can track all targets in detection zone in artillery radar networking tracking strong clutter environment, with high precision and meeting engineering application requirements.