Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Using Texture to Classify Forests Separately from Vegetation

David R. Treadwell, Derek Jacoby, Will Parkinson, Bruce Maxwell, Yvonne Coady

TL;DR

The paper tackles the challenge of distinguishing forest from non-forest vegetation in Sentinel-2 satellite imagery. It introduces a static, interpretable algorithm that fuses a texture mask derived from edge-based processing with an NDVI vegetation mask to identify forest regions. Early results on a test site show the method performing competitively with, and sometimes exceeding, a contemporary tree detector, though ground-truth data limitations temper definitive conclusions. If further developed, the approach could offer an accessible, scalable tool for forest monitoring and emergency planning using freely available satellite data.

Abstract

Identifying terrain within satellite image data is a key issue in geographical information sciences, with numerous environmental and safety implications. Many techniques exist to derive classifications from spectral data captured by satellites. However, the ability to reliably classify vegetation remains a challenge. In particular, no precise methods exist for classifying forest vs. non-forest vegetation in high-level satellite images. This paper provides an initial proposal for a static, algorithmic process to identify forest regions in satellite image data through texture features created from detected edges and the NDVI ratio captured by Sentinel-2 satellite images. With strong initial results, this paper also identifies the next steps to improve the accuracy of the classification and verification processes.

Using Texture to Classify Forests Separately from Vegetation

TL;DR

The paper tackles the challenge of distinguishing forest from non-forest vegetation in Sentinel-2 satellite imagery. It introduces a static, interpretable algorithm that fuses a texture mask derived from edge-based processing with an NDVI vegetation mask to identify forest regions. Early results on a test site show the method performing competitively with, and sometimes exceeding, a contemporary tree detector, though ground-truth data limitations temper definitive conclusions. If further developed, the approach could offer an accessible, scalable tool for forest monitoring and emergency planning using freely available satellite data.

Abstract

Identifying terrain within satellite image data is a key issue in geographical information sciences, with numerous environmental and safety implications. Many techniques exist to derive classifications from spectral data captured by satellites. However, the ability to reliably classify vegetation remains a challenge. In particular, no precise methods exist for classifying forest vs. non-forest vegetation in high-level satellite images. This paper provides an initial proposal for a static, algorithmic process to identify forest regions in satellite image data through texture features created from detected edges and the NDVI ratio captured by Sentinel-2 satellite images. With strong initial results, this paper also identifies the next steps to improve the accuracy of the classification and verification processes.
Paper Structure (7 sections, 6 figures)

This paper contains 7 sections, 6 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Satellite imagery and preparation
  4. Texture Detection
  5. Methods
  6. Results
  7. Conclusion and Next Steps

Figures (6)

  • Figure 3: Example of an area with mixed forest and non-forest vegetation, as well as other terrain types such as lake water. The picture shows Keating Lake, Southeast of Pickle Lake in Ontario, Canada.
  • Figure 4: Comparison of the RGB image with the algorithmic classification of pixels as forest, the DetecTree classification of pixels as forest, and the ground-truth image generated through method 1.
  • Figure 5: Comparison of the RGB image with the algorithmic classification of pixels as forest, the DetecTree classification of pixels as forest, and the ground-truth image generated through method 2.
  • Figure 6: Accuracy, precision, recall, and F1 scores for the static algorithm in this paper and DetecTree, for both ground-truth methods.
  • Figure 7: Confusion matrix for both the static algorithm and DetecTree using ground-truth method 1.
  • ...and 1 more figures