It is necessary to carry out the reasoning to realize effective judgement of target maneuver recognition based on information of target position and motion state in order to meet the needs of target behavior prediction in the multi-purpose application such as battlefield identification and civilian target surveillance. Aiming at the problem that the original geographical location cannot provide semantic information, this paper adopts fuzzy membership theory to construct road grid model to extract semantic feature of target position, and the possible derivation error of location semantic is avoided by using K Nearest Neighbor(KNN) method. On the basis of positional semantic modeling, the target maneuver action is inferred with Hidden Markov Model(HMM). Combining with the application of airport surface surveillance, the simulation results show that the accuracy of the proposed method based on model of location sematic feature and KNN is improved compared with the general HMM.