一种基于水纹识别的视觉定位技术

  • 打印
  • 收藏
收藏成功


打开文本图片集

中图分类号:TP391.41 文献标志码:A 文章编号:1001-3695(2026)01-035-0291-07

doi:10.19734/j. issn.1001-3695.2025.04.0144

Visual positioning technique based on water pattern recognition

Liu Baichuan¹,MingLi²,Lu Peng1†,ShangKejun² (1.ScooloffclellUifseo,e,; matic Control Equipment,Beijing 10oo74,China)

Abstract:Withtherapid development of thedrone industry inrecent years,visual positioning technologyhasbeen widelyapplied inlow-cost ightweight positioning systems.However,when drones operate in weakly texturedareassuchas water surfaces,the sparseandindistinct visual features insuch environments cause high image mismatch rates andsevereposeestimationdrift.Thislimitationrestrictsteapplicationofvisualpositioning techniqueinaquaticenvironments.Toaddressthischallenge,this paperproosedavisualpositioning technque basedonwaterpaterecognitionamedWPRVP.Firstlyitextracted multi-granularitycorelatonfeaturesfromwaterpatters,which establishedlocaland globalmatchingrelationshipsbetween imageframes.Secondly,itusedarecurentnetwork toiterativelyprocessthematching information,whichrefinedwaterpatternmatchingaccuracyprogressivelyfromcoarsetofinelevels.Finally,basedonmulti-viewgeometricconstraints,itperformed match screningand motion modeling,realizing robust pose estimation.Experimental resultsdemonstrate thatcompared with existing image matching and visual positioning techniques,WPRVPachieves applicabilityinaquatic environments and maintains higher acuracy.Itcan provideareliable,low-cost,high-precision navigation sourcefor navigation devices in aquatic environments,effectively assisting drones in executing flight missions over water.

Key Words:weakly textured environment; water pattern recognition; image matching;visual positioning

0 引言

随着科技的不断进步,低空经济逐渐成为全球经济发展的新热点。(剩余19345字)

目录
monitor
客服机器人