Data dividing based approach for target detection with limited secondary data

For target detection in unknown noise, sufficient secondary data are needed to form a nonsingular estimate of the noise covariance matrix (NCM). However the secondary data size is usually small in practice Aiming to deal with the cases of limited secondary data, a new detection strategy involving data dividing is proposed for existing detectors in this paper. Firstly the primary/secondary data vectors are divided by row into several groups, ensuring that the data model of each group meets the requirements of the existing detectors. Then the data of each group can be individually used for target detection. The final detection result is synthesized by those of all groups. In the simulation section, the polarization-space-time generalized likelihood ratio (PST-GLR) detector is selected to demonstrate the effectiveness of the detection strategy in the case of limited secondary data.