This paper presents an efficient approach to real-time trajectory planing for an autonomous aerial vehicle with terrain avoidance constraints. The trajectory optimization problem is formulated as receding horizon mixed-integer linear programming (MILP). The paper considers the case that only the terrain environment within the detection radius around the vehicle is known and modeled as a triangulated irregular network (TIN). The terrain avoidance constraint is represented in MILP form by introducing binary variables. The trajectory obtained is optimal in term of fuel, time, altitude, etc. The algorithm is applied to a small-scale helicopter with an example scenario and the simulation result shows it is applicable to online computation for real-time implementation through the use of powerful MILP solver of CPLEX.