应用哨兵2A多时相遥感影像对树种的识别

森林树种类别的准确识别和专题图制作对于森林经营管理具有重要作用。以哨兵2A多时相遥感影像数据源，通过对单时相影像以及根据基尼系数(Gini系数)和信息熵进行的不同影像组合，利用随机森林算法对大孤家林场的树种类别进行识别和分类制图。结果表明:哨兵2A多时相影像组合能够显著改善树种分类的精度，多时相影像组合总体分度最高为87．45%，Kappa系数为0．825，比单时相遥感影像分类总体精度提升了4%～7%。

Accurate identification of forest tree species and production of thematic maps play a critical role in forest management and administration. Taking Sentinel-2A multi-temporal remote sensing images as the data source, this study conducted tree species identification and classified mapping for the Daguajia Forest Farm using the random forest algorithm, based on single-temporal images and different image combinations derived via Gini coefficient and information entropy. The results demonstrate that Sentinel-2A multi-temporal image combinations can significantly improve the accuracy of tree species classification. The highest overall accuracy of the multi-temporal image combination reaches 87.45%, with a Kappa coefficient of 0.825, which is 4% to 7% higher than the overall classification accuracy of single-temporal remote sensing images.

搜集汇总

背景与挑战

背景概述

该数据集基于哨兵2A多时相遥感影像，应用随机森林算法进行树种识别与分类制图研究，属于'落叶松高效培育技术研究'项目的一部分。研究结果表明，多时相影像组合能显著提升树种分类精度，总体精度最高达87.45%，比单时相影像提升4%至7%。数据集以文档格式提供，聚焦于植物学领域，旨在支持森林经营管理。