Global optimization techniques have been used extensively due to their capability in handling complex engineering problems. Metamodel becomes effective method to enhance global optimization. In this paper, we propose a new global optimization method base on incremental metamodel. At each sampling step, we adopt inherited Latin HyperCube design to sample points step by step, and propose a new incremental metamodel to update the cofficient matrix gradually. Experiments proved that the global optimization method has highest efficiency and can be finding global minimum fastly.