Abstract: [Objectives]It is vital to detect and track ships during coastal monitoring and ship navigation over long distances in complex circumstances, which are sometimes difficult to spot immediately due to their small size and unclear features since they can be readily confused with shorelines, noises, and rocks. To address this issue, a novel ship detection method called ShipDet is proposed, which significantly improves the performance through the design of a dedicated backbone network, improved feature extraction process, and constrained microscopic detection heads. [Methods]At the very beginning, this method constructs a feature fusion and extraction network that is highly sensitive to small objects by integrating the Swin Transformer module (STR) and the classic CSPDarknet53 network. The method enhances the correlation between small target features and the environment, establishing associations between ships and waterways, ships and other ships, and ships and coastline, suppressing irrelevant information. Subsequently, considering the uneven distribution and minor scale variations of ship targets in the dataset, two detection layers are retained to reduce model parameters and further enhance model performance. Moreover, the method employs the SCYLLA-IoU (SIoU) loss function to constrain the detection head, reducing the regression freedom and improving detection accuracy and robustness. [Results]To validate the proposed method, a dataset called 2023ships has been established, consisting of up to 9000 samples covering various scenarios such as inland rivers, coastal areas, daytime, nighttime, and foggy weather. During testing, the proposed method outperformed all existing algorithms in ship detection, with an mAP of 92.9%, a precision rate of 2.1%, and a parameter size of 35.4M. [Conclusions] This method will greatly benefit from maritime monitoring and intelligent navigation.