Abstract: Detection of radar targets in non-Gaussian sea clutter background is the focus of research. Diversification of clutter distributions can′t ensure the constant false alarm ratio (CFAR) of detectors with respect to the clutter power mean, so the K-distribution, he generalized Pareto distribution and the compound-Gaussian distribution with inverse-Gaussian texture are normalized. Then three optimal coherent detectors corresponding to the three clutter distribution models are obtained. Theoretical analysis and experimental results prove that three adaptive detectors have the CFAR properties with respect to the clutter power mean, the target Doppler frequency and the speckle covariance matrix, respectively. The simulation results also show that the estimation methods of speckle covariance matrix and the number of secondary data affect the probability of detection and false alarm.