Abstract: To address target-tracking accuracy under nonlinear equality constraints, this paper proposes a general projection-based constrained filtering method. Within this unified framework, the constrained extended Kalman filter (CEKF), constrained cubature Kalman filter (CCKF), and constrained unscented Kalman filter (CUKF) are comparatively evaluated. By treating the nonlinear constraints as noise-free pseudo-measurements and incorporating them into the filtering process, a projection technique is employed to correct the state estimates so that they satisfy the constraints. Simulation results demonstrate that, under the same constraint model and noise settings, the CUKF outperforms the CEKF and CCKF in terms of position accuracy, numerical stability, and convergence speed. In particular, when the constraints are strongly nonlinear, the CUKF tracks the target trajectory more precisely, maintaining consistently low estimation errors and achieving markedly better performance than the other two methods.