一种基于内河航行环境图像的特征匹配方法  

A feature matching method based on the images of inland river navigation environment

作　　者：彭纾闵 刘磊[2] PENG Shumin;LIU Lei(College of Public Administration and Humanities,Dalian Maritime University,Dalian 116026,China;Faculty of Education,Liaoning Normal University,Dalian 116026,China)

机构地区：[1]大连海事大学公共管理与人文艺术学院,辽宁大连116026 [2]辽宁师范大学教育学部,辽宁大连116026

出　　处：《中国航海》2025年第1期69-76,共8页Navigation of China

摘　　要：图像特征配准是船舶在内河航行中拼接生成大视场图像的关键步骤。针对传统特征匹配算法在内河航行环境图像匹配时水面特征点稀疏且效率低下的问题,提出一种基于图像超分辨率重建的特征匹配方法。对输入图像利用超分辨率生成对抗网络(SRGAN)进行超分重建,丰富图像细节信息,增加图像特征点数量。使用导向快速、旋转简短算子和增强的高效二值局部图像特征描述符进行特征点检测和描述。基于汉明距离进行粗匹配,使用改进的随机抽样一致算法(RANSAC)进一步剔除粗差、提纯内点,从而获得稳健的匹配效果。选取5组具有低能见度、光照差异、尺度变化、模糊变化和旋转变化的内河航行环境图像进行试验。结果表明:该方法通过图像超分辨率重建提取的特征点数量增加,匹配正确率(R_(CM))和匹配速度均优于对比算法,满足内河航行环境图像高精度和实时匹配的需求。Image feature registration is a critical step for stitching and generating large-field images during the inland navigation of ships.To address the problems of sparse water surface feature points and low efficiency in traditional feature matching algorithms for image registration in inland navigation environments,this paper proposes a feature matching method based on image super-resolution reconstruction.Firstly,the input images are subjected to super-resolution reconstruction using generative adversarial networks to enrich image details and increase the number of image feature points.Secondly,the ORB operator and BEBLID algorithm are employed for feature point detection and description.Then,coarse matching is performed based on Hamming distance.Finally,an improved random sampling consistency algorithm is utilized to further eliminate gross errors and purify inliers,achieving robust matching results.The study conducts experiments using five sets of inland navigation environment images with challenges such as low visibility,varying lighting conditions,scale changes,blur,and rotation.The results demonstrate that the proposed approach,leveraging image super-resolution reconstruction for feature point extraction,achieves an increased number of feature points and outperforms comparative algorithms in terms of matching accuracy and speed.This method meets the requirements of high-precision and real-time image matching in inland navigation environments.

关 键 词：内河航行环境图像 超分辨率生成对抗网络 特征描述符 随机抽样一致性 图像配准 

分 类 号：U6[交通运输工程—船舶与海洋工程]