Abstract: A multi-scale ship detection algorithm for SAR image processing is developed based on CCFPN（Cross-Connected-Feature Pyramid Networks） to improve the ship detection against a complex background. The cross-connected pyramid network is used to enhance the transmission of target’s deep feature and shallow feature. The multi-channel dilated convolution is used to improve the shallow feature extraction. The channel merge is performed to enrich the information contained in the feature map. The algorithm is verified with publicly available data. Experiments show that an AP（Average Precision） of 95.62 % is achieved with overall performance improvement.