Scholarlink

Purpose <p>Accurate tree height estimation is essential for orchard phenotyping and management, yet the precision, accuracy, and reference reliability of manual and emerging remote sensing methods remain insufficiently evaluated under repeated field measurements. This study systematically compared five tree height estimation approaches in a citrus orchard to determine their repeatability, identify the most suitable practical reference method, and evaluate their relative accuracy.</p> Methods <p>Five tree height estimation approaches were evaluated: manual pole measurement (MP), airborne LiDAR (AL), airborne SLAM LiDAR (ASL), ground mobile SLAM LiDAR (GMSL), and an AI-driven photogrammetry platform (Agroview). Repeated measurements across four dates were used to assess precision and consistency, and cross-validation between MP and GMSL was performed to identify the most suitable reference method before accuracy evaluation of all approaches.</p> Results <p>GMSL showed the highest repeatability, with the lowest coefficient of variation (0.44%) and the highest intraclass correlation coefficient (ICC = 0.999), whereas Agroview showed the lowest stability (CV = 11.07%, ICC = 0.854). Cross-validation against tape-and-level benchmarks indicated that MP and GMSL had similarly small mean errors, but GMSL provided higher repeatability and more consistent tree-top identification, supporting its use as the practical reference method. Relative to GMSL, MP and ASL showed the closest agreement, with mean biases of − 0.079&#xa0;m and − 0.097&#xa0;m, respectively, whereas AL showed the largest underestimation (–0.250&#xa0;m) and Agroview showed greater variability and slight overestimation (+ 0.104&#xa0;m).</p> Conclusions <p>Overall, GMSL was the most reliable method for reference measurement, while ASL provided the best balance between airborne efficiency and accuracy. AL enabled rapid large-area acquisition but systematically underestimated tree height, whereas Agroview offered the fastest workflow but with lower repeatability and agreement than LiDAR-based methods. These findings clarify trade-offs among manual, LiDAR, and photogrammetric approaches and support more rigorous tree height estimation in cultivated orchards.</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

Precision and accuracy of tree height estimation in citrus orchards: a systematic investigation of manual, airborne LiDAR, SLAM LiDAR, AI-driven photogrammetry

  • 期刊论文

  • 发表日期：

    2026年5月16日

查看全文
  • Wenhao Liu,
  • Yiannis Ampatzidis,
  • Benjamin Wilkinson

摘要

Purpose

Accurate tree height estimation is essential for orchard phenotyping and management, yet the precision, accuracy, and reference reliability of manual and emerging remote sensing methods remain insufficiently evaluated under repeated field measurements. This study systematically compared five tree height estimation approaches in a citrus orchard to determine their repeatability, identify the most suitable practical reference method, and evaluate their relative accuracy.

Methods

Five tree height estimation approaches were evaluated: manual pole measurement (MP), airborne LiDAR (AL), airborne SLAM LiDAR (ASL), ground mobile SLAM LiDAR (GMSL), and an AI-driven photogrammetry platform (Agroview). Repeated measurements across four dates were used to assess precision and consistency, and cross-validation between MP and GMSL was performed to identify the most suitable reference method before accuracy evaluation of all approaches.

Results

GMSL showed the highest repeatability, with the lowest coefficient of variation (0.44%) and the highest intraclass correlation coefficient (ICC = 0.999), whereas Agroview showed the lowest stability (CV = 11.07%, ICC = 0.854). Cross-validation against tape-and-level benchmarks indicated that MP and GMSL had similarly small mean errors, but GMSL provided higher repeatability and more consistent tree-top identification, supporting its use as the practical reference method. Relative to GMSL, MP and ASL showed the closest agreement, with mean biases of − 0.079 m and − 0.097 m, respectively, whereas AL showed the largest underestimation (–0.250 m) and Agroview showed greater variability and slight overestimation (+ 0.104 m).

Conclusions

Overall, GMSL was the most reliable method for reference measurement, while ASL provided the best balance between airborne efficiency and accuracy. AL enabled rapid large-area acquisition but systematically underestimated tree height, whereas Agroview offered the fastest workflow but with lower repeatability and agreement than LiDAR-based methods. These findings clarify trade-offs among manual, LiDAR, and photogrammetric approaches and support more rigorous tree height estimation in cultivated orchards.