基于航迹特征和深度神经网络MobileNet的舰船分类识别方法  

Ship Classification and Recognition Based on Track Features and MobileNet

作　　者：李磊[1] 张静[1] 王哲 LI Lei;ZHANG Jing;WANG Zhe(Information Engineering University, Zhengzhou 450001, China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2021年第6期743-749,共7页Journal of Information Engineering University

基　　金：国家自然科学基金资助项目(61601516)。

摘　　要：舰船类型的及时准确识别对于目标的意图识别、威胁预警和跟踪监视具有重要意义。提出一种基于航迹特征和深度神经网络MobileNet的舰船分类识别方法。首先,提出一种基于RGB色彩空间的航迹特征提取和转换方法,将历史航迹数据中的速度、航向和加速度等特征映射到RGB色彩空间,转换为航迹特征图像数据;其次,提出一种基于深度神经网络MobileNet的迁移训练方法,对深度神经网络MobileNet及其ImageNet预训练权重进行迁移,并使用之前生成的航迹特征图像数据对网络进行训练;最后,得到舰船类型识别模型,实现船舶分类识别。实验结果表明,所提方法切实有效,与现有研究成果相比具有识别准确率高、识别速度快的特点,可有效应用于舰船目标的分类识别。The timely and accurate identification of ship type is of great significance to target intention identification,threat warning and tracking monitoring.A ship classification and recognition method based on track feature and deep neural network MobileNet is proposed in this paper.Firstly,a track feature extraction and conversion method based on RGB color space is proposed,which maps the speed,course and acceleration features from historical track data into RGB color space and converts them into track feature image data.Then,a migration training method based on deep neural network MobileNet is proposed.MobileNet and its ImageNet pre-training weights are transferred,and the network is trained using the previously generated track feature image data.Finally,the ship type recognition model is obtained to realize ship classification recognition.Experimental results show that the proposed method is effective and has the characteristics of higher accuracy and faster recognition speed compared with the existing research results,and can be effectively applied to ship targets classification and recognition.

关 键 词：RGB色彩空间 航迹特征提取 舰船分类识别 迁移学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]