| 注册
首页|期刊导航|测绘科学技术学报|长焦大倾角面阵摆扫无人机影像的位姿优化与目标定位

长焦大倾角面阵摆扫无人机影像的位姿优化与目标定位

叶勤 靳雨佳 罗俊奇 林怡

测绘科学技术学报2025,Vol.41Issue(2):154-162,9.
测绘科学技术学报2025,Vol.41Issue(2):154-162,9.DOI:10.3969/j.issn.1673-6338.2025.02.007

长焦大倾角面阵摆扫无人机影像的位姿优化与目标定位

Image Pose Optimization and Object Localization via Long-focal Large Oblique-view Whiskbroom UAV Imagery

叶勤 1靳雨佳 2罗俊奇 2林怡1

作者信息

  • 1. 同济大学 测绘与地理信息学院,上海 200092||同济大学 遥感应用研究中心,上海 200092
  • 2. 同济大学 测绘与地理信息学院,上海 200092
  • 折叠

摘要

Abstract

Capturing high resolution images from UAV-mounted LF-LO-W camera has become an ideal option to acquire information from unreachable objects in danger,avoiding hovering directly above the areas.Different from normal UAV captured images,it is challenging to perform matching,pose optimization and object localization on these images due to extremely large distortion,small FOV for single array,low overlap,excessively sharp intersec-tion angle of LF-LO-W array sweeping images.A new method of image pose optimization and object localization for LF-LO-W array sweeping UAV imagery is proposed in this paper.Firstly,stable SDS features and histogram de-scriptors resistant to obvious changes of viewpoints and scales are designed for object localization based on image retrieval.Secondly,an image matching method is proposed by using improved SuperPoint network to extract fea-tures considering the uniformity of the distribution of the image matching point pairs.Then,a pose optimization method adapted to the images with little overlap is designed by introducing a local loopback optimization strategy.Finally,a novel image intersection-based localization method incorporating elevation range constraints of the sur-veyed area is proposed.Experiments on two test sites in Weinan,Shaanxi and Shanghai show the following results.The object localization results in LF-LO-W images based on image retrieval are reliable and the localization accura-cy is improved by about 10%compared to that of SOTA methods.The object localization error based on image pair intersection reduces by about 50%compared to those of traditional intersection-based object localization methods.

关键词

长焦大倾角影像/影像检索/位姿优化/交会定位/高度范围约束

Key words

long-focal large oblique-view image(LF-LO-W)/image retrieval/pose optimization/intersection-based positioning/elevation range constraint

分类

测绘与仪器

引用本文复制引用

叶勤,靳雨佳,罗俊奇,林怡..长焦大倾角面阵摆扫无人机影像的位姿优化与目标定位[J].测绘科学技术学报,2025,41(2):154-162,9.

基金项目

上海市自然科学基金面上项目(22ZR1465700). (22ZR1465700)

测绘科学技术学报

1673-6338

访问量0
|
下载量0
段落导航相关论文