Abstract: For most domain decomposition-based parallel algorithms, a single computational grid cannot ensure real-time exchange of solution information between not directly neighboring subdomains, so the algorithms exhibit slow convergence rate. The two-grid method is an effective method to optimize the convergence rate of domain decomposition-based parallel algorithms. Focusing on the elastic contact problem, it involves strong nonlinearities, such as unknown contact region, unknown contact state, inequality contact constraint, complicated frictional constitutive law, etc. If these contact nonlinearities are not well assumed to become linear, it will be very difficult to construct global approximate solution on coarse mesh for two-grid method; on the contrary, if the global approximate solution is constructed under assumed contact configuration, then the iteration will be easy to fall into assumed contact configuration and does not converge when iteration level is not increased. This work constructs a global approximate solution on coarse mesh which does not rely on assumptions about contact configuration, then the two-grid method is adopted to optimize the convergence rate of domain decomposition-based parallel contact analysis algorithm. Firstly, the global domain is decomposed into non-overlapping subdomains along contact boundaries and division boundaries, the alternating direction method of multipliers is adopted to establish the one-grid based parallel contact analysis procedure. Next, similar to the multiscale strategy in LATIN method, the global problem which satisfies equilibrium between subdomains is constructed and restricted onto coarse mesh by geometric multigrid method, the approximate solution on coarse mesh is computed to correct the solution on computational mesh. Through comparison with dual mortar contact analysis algorithm, the effectiveness of the parallel contact analysis algorithm based on domain decomposition and two-grid method is verified for solving large-scale contact problems.