There are still some issues existing in the traditional Artificial Potential Field (APF), such as the poor adaptability to the complex environment, easily getting into local standstill and the unsmooth path. In order to solve these problems, an improved artificial potential field method was proposed. Firstly, the connectivity of threats was analyzed by the proposed algorithm, and the optimum feasible solution domain was got by the geometric topology. Secondly, a pre-planning of track points was carried out within the feasible solution domain. The pre-planning was based on the threats' global distribution information, and made up for the deficiencies of falling into local minimum and failing to find a feasible path. Finally, the gravitational function of artificial potential field method was improved, and a sufficient smooth flight path was obtained by several iterations and curvature checking. The simulation results show that the improved algorithm can meet the path planning requirements of unmanned aerial vehicles. The proposed algorithm is simple and feasible with strong searching and adaptability.