Estimation Of Target Altitude In HFSWR Based On EKF

High Frequency Surface Wave Radar (HFSWR) use the characters of low propagation loss and stable propagation of vertical polarized high frequency electromagnetic wave diffracting over the sea, to achieve over-the-horizon detection of ship target on the sea and, low altitude and super low altitude target flying near the sea to compensate the blind area of microwave radar. Because it is difficult to make narrow beam in elevation, high frequency radars (both surface-wave and sky-wave) do not have altitude measurement ability and lack target altitude information as a result. Depending on existent estimation methods of target altitude in surface-wave radars, this paper presents research on estimation of target altitude in HFSWR based on Extended Kalman Filtering (EKF). And, improved Howland algorithm and HPEKF algorithm for altitude/RCS initializing are proposed in this paper.Firstly, starting with the HFSWR radar equation, the basic principle of target altitude estimation in HFSWR is explained by theoretically analyzing each parameter in the radar equation. And, existent altitude estimation methods, such as least squares estimator, Howland algorithm, fuzzy altitude/RCS estimator and so on, are introduced in principle and the shortages of existent methods are indicated by analysis.Secondly, aiming at the problems of low precision and poor practicability in Howland algorithm by imprecise state model, improved Howland algorithm is proposed. This method models the RCS fluctuation adopting AR model on the basis of Howland algorithm and updates AR model coefficient by RCS fluctuation self-adaptively to more exactly describe the dynamics characteristic of RCS fluctuation. In this way, the accuracy of system state model is guaranteed to make the model and observation value match each other. In the first, the improved Howland algorithm under the condition of known AR model coefficient is given. In the second, the estimation algorithm of AR model coefficient under the condition of known RCS fluctuation is given. Then, by fusing these two algorithms derived under ideal conditions the improved Howland algorithm under the condition of unknown AR model coefficient is given. At the same time, aiming at the limit of selecting initial value by experience in Howland algorithm, fuzzy selecting algorithm of initial value in fuzzy altitude/RCS estimator is adopted to provide the initial value for improved Howland algorithm. By simulation, relative error of altitude estimation of improved Howland algorithm is about 10%, reducing more than 30% compared to Howland algorithm. In the result with trial data of airplane target in XXXX radar station, relative error of altitude estimation of improved Howland algorithm is about 40%, reducing more than 40% compared to Howland algorithm. Also, obsolete error achieves 300m, reducing more than 200m compared to Howland algorithm.In the end, aiming at the problem of requiring large prior knowledge in fuzzy selecting algorithm of initial value, a method to gain initial altitude/RCS value without prior knowledge is proposed, that is, HPEKF algorithm. This method adopts several independent Extended Kalman Filters for parallel filtering and each filter is assigned a different initial altitude/RCS value. The weight of each filter is calculated recursively by Bayes Law and by accumulating of observation, the weight of the filter whose initial altitude value is close to the real altitude value dominates gradually. The output altitude estimation approaches real target altitude continuously and in the end of initializing period, the output altitude/RCS filtering value is chosen as required initial value. By simulation, HPEKF algorithm can provide initial altitude/RCS value with small relative error under some conditions.