Individualización de copas en encinares (Quercus ilex L.) mediante el empleo de técnicas de segmentación de imágenes y clasificación por objetos

Abstract

When characterising a forest cover, it is of utmost importance to know its density, which can be done through the crown individualisation of each tree or foot. The present article shows the work flow developed for crown individualisation in holm oaks by means of segmentation and classification by objects techniques (OBIA), analysing its effectiveness in four forest formations characterised by high/low density and presence/absence of shrub stratum. For this, we combined the products of “Plan Nacional de Ortofotografía Aérea” relating to NIR orthophotography and first coverage lidar data. The study was conducted in four areas between 23.20 and 50.09ha within the map sheet Calañas-Huelva H50-0959, each representing a formation. As input in the segmentation process, we employed the four spectrum bands of the orthophotography and a digital vegetation model (DVM) obtained by the cloud lidar point. The segmentation flow was iteratively run through the eCognition Developer software. We performed successive segmentations and object classifications, in such a way that the objects which met a series of minimum requirements of individualisation got out of the flow and remained characterized as individual objects. The results of the proposed methodology showed a high capacity for crown individualisation, between 81.23% and 96.86% of success rate depending upon the type of formation studied, enabling the individualisation in holm oaks with crown tangent and adjacent to shrub.