In this paper, a vision based method is proposed for fixed-wing unmanned aerial vehicles' emergency landing field auto-selecting and security analyzing. Firstly, a strip structure detector SSD is designed to detect candidate landing fields in aerial images such as runways and roads. The SSD can also give the candidate strip's precise direction. Secondly, the relationship of image position, direction, the real length of a strip on the ground and its length in image is deduced. Based on this relationship, a region can be determined around a candidate point in the image for security analyzing. In the end, we present an emergency landing field auto-selecting algorithm. Experiments with real aerial images show the validity of SSD, simulations verify the correctness of the deduced conclusion and the feasibility of the presented algorithm.