Owing to the measured modal data is far less than the under identified parameters and the damage identification equation always have large errors, it leads to the application in complex structure is always limited in the traditional damage detection based on model updating. In this paper, the drawback is solved by dividing structural degrees of freedom into measured and unmeasured parts. Meanwhile, the unmeasured parts of the mode shapes of the damaged structure are characterized as a function of the structural parameter and measured parts of the mode shape. Considering damage as a reduction in element stiffness, then using the calculated modal data of an intact structure and the measured modal data of a damaged structure to establish the damage detect equation. The dual constrained objective function succeeds to recognize the changes in structural parameters, utilizing the trust region optimization method. The proposed method are verified by two numerical simulation cases and a damage detection experiment, the results show that it has a good ability to detect any damage of structures using incomplete measured modal data, and the algorithm also has good robustness.