基于深度学习的船舶标识号识别方法  

Recognition Method of Ship Identification Number Based on Deep Learning

作　　者：杨旺奇 刘俊[1] 陈振 徐浩浩 谷雨[1] 陈华杰[1] YANG Wangqi;LIU Jun;CHEN Zhen;XU Haohao;GU Yu;CHEN Huajie(Key Laboratory of Fundamental Science on Communication Information Transmission and Fusion Technology,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学通信信息传输与融合技术国防重点学科实验室,杭州310018

出　　处：《指挥信息系统与技术》2022年第1期58-63,共6页Command Information System and Technology

基　　金：浙江省重点研发计划(2019C05005)资助项目。

摘　　要：船舶标识号识别对船舶管理和海洋权益保障具有重要意义。为了提高船舶标识号的识别准确率,提出了基于深度学习的船舶标识号识别方法。该方法采用检测、矫正和识别3个模块对船舶标识号进行定位识别,并在自建船舶标识号数据集上进行了检测准确率和识别准确率验证。试验结果表明,该方法检测准确率为92.1%,识别准确率为90.1%,可有效识别船舶标识号。The identification of ship identification number is of great significance to ship management and maritime rights and interests protection.In order to improve the identification accuracy of ship identification number,the recognition method of ship identification number based on the deep learning is proposed.The method uses three modules of detection,correction and recognition to locate the ship identification number,and verifies the detection accuracy and recognition accuracy on the self-built ship identification number data set.Experimental results show that the detection accuracy of this method is 92.1%,and the recognition accuracy is 90.1%,thus it can effectively identify the ship identification number.

关 键 词：船舶标识号 目标检测 目标识别 图像矫正 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]