Ship Identification Based on Improved SSD
Cheng Hu, Zhilong Zhu, Ze Yu · 2022
Aiming at the low accuracy of ship recognition at sea and the fact that traditional target detection algorithms are greatly affected by the environment, this paper proposes a marine ship recognition method based on an improved SSD (Single shot multibox detector). On the basis of the original SSD algorithm, the original backbone network VGG16 was first replaced with resnet50, and the convolution in the last 5 layers of the Resnet50 Conv4_x structure was replaced by a hole convolution, followed by the removal of Conv5_x, followed by the subsequent prediction feature layer. A multiscale receptive field module was introduced to improve the feature extraction capabilities of the model. Finally, the high-level semantic information is enhanced by introducing a CBAM attention mechanism to improve the detection ability of the model. The experimental results show that the map of the algorithm is 94.5%, which is 17.7% higher than the original SSD algorithm, and the improved SSD algorithm has a better recognition effect of ships at sea.