Assessment of Image-Texture Improvement Applied to Unmanned Aerial Vehicle Imagery for the Identification of Biotic Stress in Espeletia. Case Study: Moorlands of Chingaza (Colombia)

Abstract

Espeletia is one of the most representative endemic species of moorland ecosystems, and is currently being affected by biotic stress. Meanwhile, the analysis of images obtained by means of unmanned aerial vehicle imagery has proved its usefulness in environmental monitoring activities. The present work is aimed at establishing whether image-texture analysis applied to unmanned aerial vehicle imagery from Moorlands of Chingaza (Colombia) allows the identification of biotic stress in Espeletia. To this end, this study makes use of occurrence analysis, gray-level co-occurrence matrix, and Fourier transform. Identification of healthy/unhealthy Espeletia is conducted using maximum likelihood tests and support vector machines. The results are assessed based on overall accuracy, the kappa coefficient and bhattacharyya distance. By combining spectral and image-texture information, it is shown that classification accuracy increases, reaching kappa coefficient values of 0,9824 and overall accuracy values of 99,51%.