An Autonomous Continuous Positioning Method for Low-Altitude UAVs Based on Image Recognition in GNSS-Denied Environments
Taozhe Du, Yongjie Lin
South China University of Technology
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
The article proposes a five-stage autonomous continuous positioning method that integrates inertial navigation and image recognition to address the prob-lem of accumulated positioning errors in unmanned aerial vehicles (UAV) in satellite-denied environments. This method first used inertial navigation (INS) for initial positioning to capture satellite image basemaps, and achieved aerial image preprocessing through heading angle correction, scale registration, and maximum inscribed rectangle cropping. Then, this study combined a U-Net semantic segmentation model to extract building areas, and innovatively adopted a two-stage matching strategy of template matching and SuperPoint & SuperGlue feature point matching to effectively over-come the interference of seasonal and lighting changes on matching. Finally, the height data was integrated into output three-dimensional coordinates. The experiment was based on the UAV-VisLOC dataset, and showed that twostage matching can reduce the average plane positioning error from 26.32 m in single-stage matching to 12.95 m, a decrease of 49.2%. The maximum error decreased from 68.74 m to 18.67 m. The proposed framework significantly improves the robustness and accuracy of positioning in complex environments, providing reliable technical support for urban low-altitude operations, emergency rescue, and other scenarios.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
计算机 / AIRobotics and Sensor-Based Localization
3D Surveying and Cultural Heritage · Aerospace Engineering and Control Systems