Unresolved measurement processing with widely separated radars using sparse modeling

This paper proposes sparse modeling for the processing of possibly unresolved measurements across multiple sensors. The sparse vector is created by discretizing the state space into a grid, then representing the received signal as a linear sum of the response due to reflections originating at those grid points. Our sensors measure the observed SNR at the output of the matched filter. These samples of the matched filter will represent a specific volume in Euclidian space corresponding with range bins. Hence, this work considers noncoherent processing, thus the phase information in each range bin is not used during processing. Target cross sections are as aspect-dependent, and widely separated sensors are considered, so the returns from different sensors are considered to be uncorrelated. Using our sparse model, these assumptions lead to several minimization problems that are related by their sparsity patterns. We demonstrate that by using our modeling technique, targets that may be unresolved in each sensor´s measurement space may be resolvable by combining the spacial information from multiple sensors.