Scholarlink

Abstract <p>This paper examines the problem of visual navigation for unmanned aerial vehicles in low-light conditions. To ensure the day/night navigation, a method for cross-modal transformation of visible-light satellite imagery into thermal representations is proposed using the ThermalGen flow-based generative model. The resulting heat maps are used in an absolute visual localization algorithm based on matching data from an onboard LWIR camera with a transformed satellite reference image. The algorithm is integrated into a hybrid visual-inertial navigation module that fuses the results of absolute visual localization, visual odometry, and inertial measurements based on an extended Kalman filter. Testing on images transformed into thermal images demonstrated high accuracy in determining absolute coordinates and confirmed the viability of the proposed approach for nighttime navigation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multimodal Visual Navigation of Unmanned Aerial Vehicles Using Satellite Maps and Thermal Imaging Cameras

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • P. A. Lazareva,
  • A. I. Malikov,
  • G. L. Degtyarev,
  • R. T. Mannanov

摘要

Abstract

This paper examines the problem of visual navigation for unmanned aerial vehicles in low-light conditions. To ensure the day/night navigation, a method for cross-modal transformation of visible-light satellite imagery into thermal representations is proposed using the ThermalGen flow-based generative model. The resulting heat maps are used in an absolute visual localization algorithm based on matching data from an onboard LWIR camera with a transformed satellite reference image. The algorithm is integrated into a hybrid visual-inertial navigation module that fuses the results of absolute visual localization, visual odometry, and inertial measurements based on an extended Kalman filter. Testing on images transformed into thermal images demonstrated high accuracy in determining absolute coordinates and confirmed the viability of the proposed approach for nighttime navigation.