Abstract: ‍ ‍This paper systematically reviews the advancements in radar maritime target detection technology， focusing on joint and cascaded processing methods in two key areas： the improvement of the signal-to-clutter-and-noise ratio （SCNR） and the formation of detection statistics. For joint processing， the energy-domain adaptive constant false alarm rate （CFAR） detection technologies used for various combinations of different target models， clutter models， test criteria， as well as different parameter estimation and training sample screening methods were considered. From the perspective of engineering applications， further in-depth research is needed in three areas： the construction of detection statistics in complex situations， the estimation of clutter covariance matrix when the training samples were insufficient， and the improvement of the information utilization rate of data. For cascaded processing， the emphasis was on radar target feature detection methods. A general framework for studying radar target feature detection methods was proposed， and the five steps involved—feature extraction， feature analysis， feature re-expression， feature selection and formation of detection statistics， and solving the detection threshold or decision-making space—were elaborated. Additionally， the formation of detection statistics based on information geometry was described and several main matrix CFAR detectors based on information geometry were listed. Finally， for integrating data-driven and model-driven approaches， three deep learning-based radar target detection processing frameworks were proposed： （1） intelligent identification of observation conditions and selection of model-based maritime target detection algorithms， （2） intelligent processing to replace traditional radar signal processing links， and （3） end-to-end intelligent integrated processing. This paper also presents suggestions， solutions， and practical results addressing key challenges in radar maritime target detection technology.