skip to main content


Title: Unsupervised individual tree crown detection in high-resolution satellite imagery

Rapidly and accurately detecting individual tree crowns in satellite imagery is a critical need for monitoring and characterizing forest resources. We present a two-stage semiautomated approach for detecting individual tree crowns using high spatial resolution (0.6 m) satellite imagery. First, active contours are used to recognize tree canopy areas in a normalized difference vegetation index image. Given the image areas corresponding to tree canopies, we then identify individual tree crowns as local extrema points in the Laplacian of Gaussian scale-space pyramid. The approach simultaneously detects tree crown centers and estimates tree crown sizes, parameters critical to multiple ecosystem models. As a demonstration, we used a ground validated, 0.6 m resolution QuickBird image of a sparse forest site. The two-stage approach produced a tree count estimate with an accuracy of 78% for a naturally regenerating forest with irregularly spaced trees, a success rate equivalent to or better than existing approaches. In addition, our approach detects tree canopy areas and individual tree crowns in an unsupervised manner and helps identify overlapping crowns. Furthermore, the method also demonstrates significant potential for further improvement.
 [1] ;  [1] ;  [1]
  1. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Publication Date:
Report Number(s):
Journal ID: ISSN 1931-3195
Grant/Contract Number:
Accepted Manuscript
Journal Name:
Jounral of Applied Remote Sensing
Additional Journal Information:
Journal Volume: 10; Journal Issue: 1; Journal ID: ISSN 1931-3195
Research Org:
Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Sponsoring Org:
Country of Publication:
United States
54 ENVIRONMENTAL SCIENCES; 97 MATHEMATICS AND COMPUTING; land cover classification; variational segmentation; active contours; scale-space theory; multi-scale analysis; tree crown detection
OSTI Identifier: