To overcome the scope limitation of video observation from fixed platforms, a new method of detecting vehicular trajectory from the traffic flow video provided by model aircraft is proposed. First, model aircraft shoots the traffic flow in the air to get the video which is consequently extracted into continuous frames. Second, by using the pinhole camera model and the algorithm of space coordinate transformation, the coordinates of vehicles from the continuous frames are obtained to form the spatial temporal trajectory of vehicles. Finally, three kinds of bias of the proposed method are analyzed. The results show that the relative bias in the standard checkerboard testing is less than 5％, and the accuracy of the data collected by the system under the real road is greater than 90％. The proposed approach provides a large scope and low-cost way for traffic observation, which can meet the demand of survey in transportation engineering.