舰船多方位视觉图像特征深度提取系统设计
Design of a multi-view visual feature extraction system for ship imagery using deep learning
王晓橹1
作者信息
摘要
基于视觉图像处理的特征提取和图像识别方案是对雷达和AIS的有效补充,有效识别其他船舶、礁石等可以有效保障舰船航行安全.本文提出一种基于CNN和拉普拉斯金字塔图像融合的视觉图像特征深度提取系统,设计了系统的基本结构,并分析了系统中各模块的基本功能,提出了基于CNN的特征深度提取方案,对船舶和浮标目标特征进行提取,使用拉普拉斯金字塔融合将不同舰船上获取的图像进行深度融合.本文建立的多方位视觉图像特征深度提取系统可以有效获取多种目标特征,并可以有效提高特征表示的丰富性和准确性.
Abstract
Feature extraction and image recognition schemes based on visual image processing are effective comple-ments to radar and AIS,effectively identifying other ship targets,reefs,and so on,which can effectively ensure the safety of ship navigation.This paper proposes a visual image feature deep extraction system based on CNN and Laplacian pyramid image fusion,designs the basic structure of the system,and analyzes the basic functions of each module within the system.A feature deep extraction scheme based on CNN is proposed,and features of ship and buoy targets are extracted.The Lapla-cian pyramid fusion is used to deeply integrate images obtained from different ships.The multi-directional visual image feature deep extraction system established in this paper can effectively acquire various target features and can significantly enhance the richness and accuracy of feature representation.
关键词
特征深度提取/CNN/视觉图像/深度融合Key words
feature deep extraction/CNN/visual imaging/deep fusion引用本文复制引用
出版年
2024
NETL
NSTL
万方数据