2020年-2021年树种识别深度学习模型实验数据集

针对廊坊地区典型的6类树种分别设计6条无人机调查样带，每条调查样带划分为4个测区，编号为1~4。训练样本、测试样本和验证样本从不同的测区中分别选取。本项目最终得到的树种识别深度学习模型实验数据集包括：（1）无人机拍摄的高分辨率正射遥感影像19,200张，影像尺寸为7952×5304，影像分辨率5cm；（2）树冠样本标记数据（7,200张影像）；（3）树种深度学习模型识别结果数据（4,800张影像）；（4）树种影像解译标注库1套；（5）树冠轮廓矢量集6个；（6）树种识别精度表6个；（7）树种识别混淆矩阵表1个。以上数据集以.jpg、.json、.shp等格式存储，由31,214个数据文件组成，数据量约为298G。

Six UAV survey transects were designed for six typical tree species in the Langfang area. Each transect was divided into four survey plots numbered 1 to 4. Training, test, and validation samples were selected from distinct survey plots respectively.
The experimental dataset for tree species recognition deep learning models obtained in this project includes:
(1) 19,200 high-resolution orthophoto remote sensing images captured by UAVs, with a size of 7952×5304 and a spatial resolution of 5 cm;
(2) Crown sample annotation data (7,200 images);
(3) Tree species recognition result data via deep learning models (4,800 images);
(4) One set of tree species image interpretation and annotation database;
(5) Six crown contour vector datasets;
(6) Six tree species recognition accuracy evaluation tables;
(7) One tree species recognition confusion matrix table.
All the above datasets are stored in formats including .jpg, .json, .shp, etc., comprising 31,214 data files with a total data volume of approximately 298 GB.

搜集汇总

背景与挑战

背景概述

该数据集包含2020年至2021年间在廊坊、邯郸、潍坊、济南等地采集的六种典型树种的无人机调查数据，主要用于树种识别深度学习模型的训练和测试。数据集包括19200张高分辨率正射遥感图像、7200张树冠样本标记数据、4800张树种识别结果数据，以及树冠轮廓矢量集、识别精度表等，总数据量约298GB，存储格式多样，适用于测绘技术和遥感反演等领域的研究。