Abstract: The problem of target tracking has been widely investigated in air surveillance systems. The hidden target state estimation is generally resolved in dynamic state space under assumptions of Gaussian noise and model linearity on motion and observation processes. A successful state estimation should be performed based on useful information extracted from motion observations, such as the target turn angle. In this paper, we present a generalised turn-prediction technique based on Hough transform to derive turn angle formed by two adjacent segments, and use this angle to predict target position. A H? filter is adapted to estimate and update turn angle after reception of a new measurement. The derived filter is combined with a straight motion filter in a hybrid scheme for manoeuvring target tracking. Simulation results have demonstrated the improvement of tracking performances with the proposed algorithm, without computation complexity growing.