In this paper, the target tracking problem of nonlinear systems with state nonlinear inequality constraints with noise uncertainty is studied. Considering the abnormal noise, the constraint information is determined a priori by the known true trajectory, in order to fuse the prior information, the unscented Kalman filter algorithm based on probability density truncated variational inference is proposed. The skew normal distribution is used to model the measurement noise, and Gaussian stratification is carried out under the Bayesian framework. The noise and posterior probability distribution of the system are calculated by the recursive unscented Kalman filtering method of variational reasoning. Considering the influence of nonlinear constraints, the constrained dynamics model is established, and the constrained information and unconstrained dynamics are optimized, and the truncated mean and variance are obtained by the method of probability density stage. Numerical simulation and experimental simulation results show that the proposed filtering algorithm has higher filtering accuracy when the noise is abnormal under nonlinear constraints.