Aimed at solving the unmanned aerial vehicle (UAV) autonomous collision avoidance problem in the noncooperation environment, a 3D UAV collision avoidance control algorithm is designed. Based on analyzing the spatial relationship of UAV and intruders, a collision measure rule is proved， and the minimum interval and residual conflict resolution time are defined to measure conflict emergency between UAV and intruders. According to the collision measure rule, a UAV collision avoidance decision model is established based on the nonlinear model predictive control (NMPC) algorithm. In order to solve the collision avoidance model effectively, a cut-off search method is designed. The simulations show that, the proposed algorithm can implement UAV collision avoidance in noncooperation environment, has fast performance and less computation, and can reduce the UAV collision danger effectively.