We propose a method based on a flight situation diagram and an improved artificial potential field algorithm for obstacle avoidance of coaxial rotor unmanned aerial vehicle (CR-UAV) flying in unknown and dangerous environments. First, a flight situation diagram is used to model obstacle information that considers the constraint conditions of flight control of CR-UAVs. By using this obstacle information, the CR-UAV can effectively avoid obstacles, avoid falling into a local minimum, and enhance its control and obstacle avoidance ability. Second, the CR-UAV adopts the unknown parameter adaptive control, which is based on radial basis function neural network (RBFNN) approximation, to approximate estimation and real-time compensation of disturbances for obstacle avoidance. Attitude tracking is realized using the method of state error feedback, which can facilitate attitude stability in obstacle avoidance flight. Finally, we carry out some simulation experiments. The results of our simulation experiments show that the CR-UAV has a good ability for attitude stability in obstacle avoidance flight.