Object Detection and Tracking for OTHR
Detecting ships and low flying targets such as cruise missiles are of great importance in military surveillance and reconnaissance. Over-the-Horizon radar (OTHR) can detect targets from several thousand kilometers away and hence can provide very early warning to operators about the presence of targets. It is well known that there are several challenges in long range surveillance and reconnaissance. First, the natural or manmade clutter can be non-Gaussian, non-stationary, and unpredictable. Second, the distance between radar and a target can be thousands of kilometers, which make the radar returns very weak. Third, in some cases, the target motions may be maneuvering (turning, accelerating or decelerating), making the Doppler frequencies non-stationary. As a result, conventional Fourier transform based target detection techniques are ineffective in dealing with time-varying Doppler frequencies.
Some target tracking results by using Kalman filter and interacting multiple model (IMM) filter are shown in the following figures. Due to the presence of target maneuvering (acceleration), Kalman filter did not perform as well as the IMM filter.
 |
 |
| Fig. 1 Velocity estimation by Kalman filter with slower convergence. | Fig. 2 Acceleration estimation by Kalman filter with slower convergence. |
 |
 |
| Fig. 3 Velocity estimation by the IMM filter with faster convergence. | Fig. 4 Acceleration estimation by the IMM filter with faster convergence. |