Dataset from paper "Canopy palm cover across the Brazilian Amazon forests mapped with airborne LiDAR data and deep learning"

<strong>Data and code from the paper:</strong> Dalagnol, R., Wagner, F. H., Emilio, T., Streher, A. S., Galvão, L. S., Ometto, J. P. H. B., &amp; Aragão, L. E. O. C. (2022). Canopy palm cover across the Brazilian Amazon forests mapped with airborne LiDAR data and deep learning. Remote Sensing in Ecology and Conservation, 1–14. https://doi.org/10.1002/rse2.264 <strong>Link:</strong> https://doi.org/10.1002/rse2.264 <strong>This repository contains:</strong> <strong>1) model_train.R:</strong> This is the code to run the U-Net model in R language. <strong>2) input.rar:</strong> Dataset of lidar canopy height model (CHM) images and masks (labels) patches of canopy palms obtained from four sites in the Brazilian Amazon. The images/masks have 128 x 128 pixels, where each pixel represents 0.5 m in the terrain. The dataset contains 2,269 images and masks, with close to 7,000 palms manually labelled. <strong>3) unet_weights_best.h5:</strong> These are the best weights for the U-Net architecture achieved in the paper. <strong>4) palm_stats.RData:</strong> Data frame with the lat/lon coordinates and palm metrics extracted for the 610 lidar sites in the Brazilian Amazon. (i) n_total is the number of palms, (ii) n_ha is the density of palms per hectare, (iii) crown_ metrics are based on the area of palm segments (in square meters), (iv) cover_total is the total area occupied by palms in the forest canopy (in square meters), (v) cover_rel is the relative cover of palms in the forest canopy (in percentage), (vi) height_ metrics are based on the height of palm segments (in meters), (vii) palm_height_dif_mean is the mean difference between palm height and local canopy height, and (viii) palm_height_dif_pvalue is the p-value assessing the statistical difference between the palm and canopy heights where 0 means no difference and -1/+1 means a negative/positive difference. If you need anything else, please contact the corresponding author: Ricardo Dalagnol ([e-mail]). <strong>If you use these data, please cite the paper:</strong> Dalagnol, R., Wagner, F. H., Emilio, T., Streher, A. S., Galvão, L. S., Ometto, J. P. H. B., &amp; Aragão, L. E. O. C. (2022). Canopy palm cover across the Brazilian Amazon forests mapped with airborne LiDAR data and deep learning. Remote Sensing in Ecology and Conservation, 1–14. https://doi.org/10.1002/rse2.264本仓库包含以下论文的数据与代码：Dalagnol, R., Wagner, F. H., Emilio, T., Streher, A. S., Galvão, L. S., Ometto, J. P. H. B., & Aragão, L. E. O. C. (2022). 《利用机载激光雷达（LiDAR）数据与深度学习绘制巴西亚马逊森林冠层棕榈覆盖度》，发表于《Remote Sensing in Ecology and Conservation》，页码1–14，DOI：https://doi.org/10.1002/rse2.264。链接：https://doi.org/10.1002/rse2.264 本仓库包含以下内容： 1. **model_train.R**：用于在R语言环境中运行U-Net模型的代码文件。 2. **input.rar**：该数据集包含从巴西亚马逊地区4个采样点获取的激光雷达冠层高度模型（CHM）图像以及冠层棕榈的掩膜（标签）图像切块。所有图像与掩膜的尺寸均为128×128像素，每个像素对应地面0.5米的空间范围。本数据集共包含2269组图像与掩膜，其中经人工标注的棕榈个体总计近7000棵。 3. **unet_weights_best.h5**：本论文中训练得到的U-Net架构最优模型权重文件。 4. **palm_stats.RData**：包含巴西亚马逊地区610个激光雷达采样点的经纬度坐标以及棕榈相关统计指标的数据框。各指标含义如下： (i) n_total：研究区内棕榈个体总数； (ii) n_ha：每公顷范围内的棕榈密度； (iii) crown_metrics：基于棕榈冠段面积（单位：平方米）计算的冠层相关指标； (iv) cover_total：森林冠层中棕榈占据的总覆盖面积（单位：平方米）； (v) cover_rel：森林冠层中棕榈的相对覆盖度（百分比形式）； (vi) height_metrics：基于棕榈冠段高度（单位：米）计算的高度相关指标； (vii) palm_height_dif_mean：单棵棕榈高度与对应局部冠层平均高度的平均差值； (viii) palm_height_dif_pvalue：用于检验棕榈高度与冠层高度间统计学差异的p值，其中0代表无统计学差异，-1与+1分别代表负向与正向差异。 若使用本数据集，请引用以下论文：Dalagnol, R., Wagner, F. H., Emilio, T., Streher, A. S., Galvão, L. S., Ometto, J. P. H. B., & Aragão, L. E. O. C. (2022). 《利用机载激光雷达（LiDAR）数据与深度学习绘制巴西亚马逊森林冠层棕榈覆盖度》，《Remote Sensing in Ecology and Conservation》，1–14. https://doi.org/10.1002/rse2.264。如有其他需求，请联系通讯作者里卡多·达拉尼奥尔（Ricardo Dalagnol），邮箱：[e-mail]。