摘要
传统舰船运动图像特征提取算法,受到相邻图像像素点特征相似度影响,无法准确提取回转状态下的单帧图像特征,导致部分单帧提取应用场景无法实现。因此,为了解决单帧特征提取问题,提出CNN在单帧舰船回转运动图像特征提取中的应用。利用CNN网络加权系数共享性,首先对单帧运动图像进行单帧像素边缘参量CNN特征确认,获得运动图像回转单帧边缘特征信息后,对其进行CNN特征模型建立,最后根据特征模型,完成单帧像素特征的提取计算。通过实际数据参量的调试表明,提出方法能到准确完成回转图像的单帧特征提取,且特征提取准确度优于传统提取方法,满足实际应用要求。
The traditional ship motion image feature extraction algorithm,affected by the feature similarity of adjacent image pixels,cannot accurately extract the single-frame image features in the rotation state,resulting in part of the singleframe extraction application scenarios cannot be realized.Therefore,to solve the problem of single frame feature extraction,the application of CNN in single frame ship rotary motion image feature extraction is proposed.Using the weighted coefficient sharing of CNN network,the single frame pixel edge parameter CNN feature confirms the single frame motion image,and obtains the CNN frame edge feature information.The debugging of the actual data parameter shows that the method can accurately complete the single frame feature extraction of the rotary image,and the feature extraction accuracy is better than the traditional extraction method and meet the actual application requirements.
作者
郭庆华
GUO Qing-hua(Zhengzhou University of Industria Technolog,Zhengzhou 451150,China)
出处
《舰船科学技术》
北大核心
2021年第22期190-192,共3页
Ship Science and Technology
基金
河南省社科联调研课题(SKL-2019-2714)
关键词
CNN
单帧
回转
特征
提取
CNN
single frame
echo
feature
extraction
