Optimal Input Features for Tree Species Classification in Central Europe Based on Multi-Temporal Sentinel-2 Data

Remote Sensing, Journal Year: 2019, Volume and Issue: 11(22), P. 2599 - 2599

Published: Nov. 6, 2019

Detailed knowledge about tree species composition is of great importance for forest management. The two identical European Space Agency (ESA) Sentinel-2 (S2) satellites provide data with unprecedented spectral, spatial and temporal resolution. Here, we investigated the potential benefits using high resolution classification five coniferous seven broadleaved in a diverse Central Forest. To run classification, 18 cloud-free S2 acquisitions were analyzed two-step approach. available scenes first used to stratify study area into six broad land-cover classes. Subsequently, additional models created separately strata. permit deeper analytical insight multi-temporal datasets identification, developed taking account all 262,143 possible permutations scenes. Each model was fine-tuned stepwise recursive feature reduction. use vegetation indices improved performances by around 5 percentage points. Individual mono-temporal accuracies range from 48.1% (January 2017) 78.6% (June 2017). Compared best results, analysis approach improves out-of-bag overall accuracy 72.9% 85.7% 83.8% 95.3% species, respectively. Remarkably, combination six–seven achieves quality equally as based on data; images April until August proved most important. classes Beech Larch attain highest user’s 96.3% 95.9%, important spectral variables distinguish between are located Red (coniferous) short wave infrared (SWIR) bands (broadleaved), Overall, highlights species-level classifications forests.

Language: Английский

---

Explainable identification and mapping of trees using UAV RGB image and deep learning

Scientific Reports, Journal Year: 2021, Volume and Issue: 11(1)

Published: Jan. 13, 2021

The identification and mapping of trees via remotely sensed data for application in forest management is an active area research. Previously proposed methods using airborne hyperspectral sensors can identify tree species with high accuracy but are costly thus unsuitable small-scale managers. In this work, we constructed a machine vision system Red-Green-Blue (RGB) image taken by unmanned aerial vehicle (UAV) convolutional neural network (CNN). system, first calculated the slope from three-dimensional model obtained UAV, segmented UAV RGB photograph into several crown objects automatically colour information model, lastly applied object-based CNN classification each image. This succeeded classifying seven classes, including more than 90% accuracy. guided gradient-weighted class activation (Guided Grad-CAM) showed that classified according to their shapes leaf contrasts, which enhances potential individual similar colours cost-effective manner-a useful feature management.

Language: Английский

---

The assessment of landslide susceptibility mapping using random forest and decision tree methods in the Three Gorges Reservoir area, China

Environmental Earth Sciences, Journal Year: 2017, Volume and Issue: 76(11)

Published: June 1, 2017

Language: Английский

---

Understanding Forest Health with Remote Sensing -Part I—A Review of Spectral Traits, Processes and Remote-Sensing Characteristics

Remote Sensing, Journal Year: 2016, Volume and Issue: 8(12), P. 1029 - 1029

Published: Dec. 18, 2016

Anthropogenic stress and disturbance of forest ecosystems (FES) has been increasing at all scales from local to global. In rapidly changing environments, in-situ terrestrial FES monitoring approaches have made tremendous progress but they are intensive often integrate subjective indicators for health (FH). Remote sensing (RS) bridges the gaps these limitations, by FH on different spatio-temporal scales, in a cost-effective, rapid, repetitive objective manner. this paper, we provide an overview definitions FH, discussing drivers, processes, adaptation mechanisms plants, how can observe with RS. We introduce concept spectral traits (ST) trait variations (STV) context discuss prospects, limitations constraints. Stress, disturbances resource cause changes taxonomic, structural functional diversity; examples ST/STV approach be used characteristics. show that RS based assessments using is competent, affordable, technique monitoring. Even though possibilities observing taxonomic diversity animal species limited RS, taxonomy tree recorded even its accuracy subject certain proved successful impacts diversity. particular, it proven very suitable recording short-term dynamics which cannot cost-effectively methods. This paper gives approach, whereas second series concentrates monitoring, sensors techniques measuring FH.

Language: Английский

---

Performance of vegetation indices from Landsat time series in deforestation monitoring

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2016, Volume and Issue: 52, P. 318 - 327

Published: July 16, 2016

Language: Английский

---