Abstract: ‍ ‍Synthetic Aperture Radar （SAR） can be independent of the weather and environmental impact， long time， long distance， and efficient detection of the target area and data acquisition to obtain high-resolution observation data. Concurrently， SAR can be multi-polarization， multi-band observation to obtain a wealth of reliable information. The above characteristics make SAR useful for field of target detection on board a ship. The above characteristics make SAR play a crucial role in the field of ship target detection. However， the measured SAR images exhibit multi-target neighborhood interference， and the traditional constant false alarm rate algorithm has poor detection performance and low efficiency. Aiming at the low efficiency of the traditional super-pixel level detection algorithm， a fast non-iterative clustering algorithm based on Gamma distribution is proposed in the super-pixel segmentation module. The algorithm uses a priority queue to dynamically update the clustering center in real time， which achieves the non-iterative operation of the clustering process and improves the distance metric based on the Gamma distribution for the characteristics of SAR images， which makes the algorithm more suitable for dealing with grey scale images. For multi-target neighborhood interference， this study selects the truncated Gamma distribution model with a good fitting effect on background clutter in the CFAR detection module. Combining the improved super-pixel segmentation algorithm with the detection algorithm based on truncated Gamma distribution， an improved super-pixel fast detection algorithm for SAR image targets is finally designed. The algorithm first applies the improved fast non-iterative clustering hyperpixel segmentation algorithm proposed in this paper for SAR images to divide the SAR image into blocks of hyperpixels that contain similar pixels based on the improved distance metric based on the Gamma distribution， which effectively improves the computational efficiency of the segmentation process. Second， an adaptive approach is employed to determine the truncation depth within the sliding window of the hyperpixels， which effectively eliminates abnormal pixel points. The relevant parameters of the truncated Gamma distribution are estimated based on the screened plausible sea clutter data. Finally， based on the preset false alarm rate， the CFAR detection threshold is adaptively calculated to complete target detection. Following experimental verification， the improved super-pixel segmentation algorithm proposed in this paper guarantees excellent segmentation performance with significantly lower complexity； in addition， the KL scatter and histogram fitting degree of different schemes for real clutter are compared， which proves that the truncated Gamma distribution chosen in this paper retains the real sea clutter samples as much as possible， and presents improved precision and enhanced fitting efficacy in the construction of a statistical model， which is more advantageous for the subsequent detection algorithm； the truncated Gamma distribution chosen in this study retains the real sea clutter samples. Subsequent detection algorithms are more beneficial； in real SAR images with multi-target neighborhood， the improved target super-pixel fast detection algorithm in this study has a high detection rate and execution efficiency.