A Novel Initialization Method For Em-Based Isar Scatterer Trajectory Matrix Completion

To improve the performance of expectation maximization (EM) in retrieving the missing data of inverse synthetic aperture radar (ISAR) scatterer trajectory matrix, a novel initialization method is proposed. Firstly, we derive the ellipse motion dynamics of the projected scatterer trajectory. Then, based on estimated ellipse parameters using known data of each scatterer trajectory, we propose the bidirectional Kalman filter to initialize the missing data. Finally, EM algorithm is applied to estimate the missing data and factorization method is performed on the complete trajectory matrix to obtain the three-dimensional geometry of scatterers. Experimental results using simulated data verify the effectiveness of the proposed initialization method.