【摘要】 针对船舶名称识别存在准确率低，所需样本量大的问题，提出一种基于注意力孪生网络的船舶名称识别方法。首先，采用YOLO网络进行数据预处理，降低海上背景干扰；其次，使用VGG16主干网络提取船舶名称图像特征，并引入注意力机制模块增强特征的表征能力；最后，基于孪生网络分类器计算图像对的相似度描述识别船舶名称。实验结果表明，该方法的识别准确率为90.75%,高于原始算法与其他船舶名称识别方法的准确率。更多还原

【Abstract】 Based on the problems of low accuracy, large number of samples in ship name recognition, a ship name recognition method based on attention siamese network is proposed. Firstly, the YOLO network is used for data preprocessing to reduce the background interference at sea. Secondly, the VGG16 backbone network is used to extract the ship name image features, and the attention mechanism module is introduced to enhance the representation ability of the features. Finally, the similarity description of image pairs is calculated based on the siamese network classifier to identify the ship name. Experiment results show that the recognition accuracy of this method is 90.75 %, which is higher than the accuracy of the original algorithm and other ship name recognition methods.更多还原