The effective use of road information is probably the most important for improving ground target tracking.A modeling and filtering method for ground moving target tracking under road constraints is proposed in this paper.A mathematic model of road net and a process noise modification method with road constraints are given.Then the method of state and covariance correcting for a target moving from one road to another is presented based on maximum posterior probability.A variable structure of multiple models (VSMM) approach generated by adaptive filtering model of current road is proposed for reducing dynamic error between filtering model and the true path of target.Monte Carlo simulation result shows the good performance of proposed method.