In this paper, the constrained trajectory optimization problem for quad-rotors is solved by a sequential convex programming-based method. The considered nonconvex problem is approximated by convex subproblems which are solved successively to generate the optimal trajectory efficiently. A reference correction strategy that utilizes the information contained in the previous solutions is proposed. Several solutions to convex subproblems are combined and their weights are optimized in the sense of minimizing the original objective. Such a method provides high-quality reference solutions for the following iterations. Numerical results showed that the proposed reference correction method generated a feasible trajectory that bypasses all the obstacles with faster convergence.