用于无人机输电线检测的双目立体匹配算法

打开文本图片集

中图分类号：V279 文献标志码：A 文章编号：1671-5276（2025）05-0224-04

Abstract：Toddress thechanllngetotheeficient detectionofsmallobjectssuchaspowerlinesinUAVapplications， this study designs a stereo matching algorithm based on Transformer and geometry encoding，which leverages across一 imageatention mechanism tointegrate information fromanotherimage，facilitating therecoveryoffine一depth details. Geometry encoding is employed to transform the conventional cost volume intoa geometry encoding volume，encode geometric and contextual information for precise estimation of power line targets and aid of UAVobstacle avoidance. Experimental results demonstrate thatcompared to state一of-the一art stereo matching algorithms，the proposed algorithmexhibits superior performancein power line depth estimation，capableofaccurately estimating the targets.By enhancing the accuracy ofsmall object detection，this study provides robust support for UAVapplications inreal一world scenarios such as power line inspection，and contributes to theadvancement of UAV technology in engineering domains. Keywords：stereo matching；UAV；Transformer；power line detection； depth estimation

0 引言

术估算输电线的三维深度，帮助无人机规避飞行中的输电线障碍物，确保飞行安全。（剩余7769字）