High efficiency milling is one of the main goals of CNC machining. Although roughing operation has low requirement for machining accuracy, the milling parameter selection usually need to consider the cutting force, milling stability, spindle torque and power. Besides, the tool wear in difficult-to-cut material machining is another very prominent problem. Therefore, how to achieve the high efficient milling parameters under the multi-constraint condition has been a severe challenge in the field of CNC machining. In this paper, a milling parameter optimization method for efficient rough machining is proposed by combining the off-line optimization and real-time monitoring. Firstly, a mathematical model of the optimization problem is established, which takes the machining efficiency as the objective, the spindle speed, radial and axial depth of cuts as the variables, and considers the multi-constraint, such as the basic parameter feasible region, cutting force, stability, spindle torque and power. Secondly, the quantitative relationship between the spindle three-phase current and milling force is established to realize the real-time estimation of cutting force coefficients. Furthermore, a numerical optimization method based on the random vector search is proposed, and the overall optimization procedure is given. Finally, a series of experimental verifications on the titanium alloy Ti5Al5Mo5VCrFe machining has been carried out, and the case study results show that the proposed parameter optimization method can greatly improve the machining efficiency, which proves the validity of the method.