Aiming at the problem of cooperative collision detection and resolution of multi-UAVs by heading control, this paper proposes a local centralized two-layer optimization method. First, the practical conflict constraints and the potential conflict constraints are regarded as the same type of constraints to ensure that the multi-UAV conflict problem can be solved in great degree. The method of conflict detection based on sampling is designed, and the number of searching feasible regions is reduced by rotating local coordinate system, and two kinds of constraint conditions, including the terminal point constraint and the tangential constraint, are analyzed. Then, the conflict relation of multi-UAV conflict problem is divided by graph theory, and the additional flight distance caused by maneuver is taken as the cost of resolution to design the maneuver cost function. In order to solve the non-linear optimization problem of the designed maneuvering cost function, a two-layer optimization strategy is proposed. The initial feasible solution is firstly searched by Stochastic Parallel Gradient Descent (SPGD), and then the optimal solution is obtained by using Sequential Quadratic Programming (SQP). Finally, Monte Carlo method is used to evaluate the reliability of the algorithm. The simulation results show that this method can satisfy the need of online planning and can achieve 100% conflict resolution under the condition of conflict start distance D_avo=τ×v_i(τ=25 s). This method can reduce the maneuvering consumption on the basis of ensuring the security of multi-UAV conflict resolution.