A method for finding the transition structures (TS) based on constrained optimization techniques is proposed. The algorithm can be considered as a step-by-step walking uphill process along the minimum energy path, followed by a refining procedure of TS parameters in the saddle point vicinity. By accounting for the constraint conditions in a straightforward manner, it is possible to use efficient quasi-Newton algorithms at every geometry reoptimization step and to manage the moving direction in a reaction valley. This approach may be suitable in the framework of density functional theory. Tests including a potential energy surface model, HNC-->HCN and N2H2 trans-cis isomerizations and LiBH4 rearrangements are given. The possible future development of the approach is discussed.