In the earthquake rescue, the unmanned aerial vehicle (UAV)equipped with radar life detector can be used for trapped people searching. In order to shorten the time of UAV detection, we propose a coupling method of Dijkstra's algorithm and simulated annealing (SA)algorithm to optimize the search path. Concisely, the mathematical model is further abstracted as the Traveling Salesman Problem (TSP)and the shortest loop can be obtained by SA algorithm. We implement this algorithm to a real earthquake event in Jiuzhaigou (Sichuan Province, China). Moreover, the digital elevation model (DEM)data is extracted from Google Earth to demonstrate the real geo-environment. Setting six critical areas as our life detection targets, UAV 3D path simulation is conducted with MATLAB. The results indicate that our method can achieve the function of obstacle avoidance and speed up the search process obviously.