Cork oak woodland land-cover types classification: a comparison between UAV sensed imagery and field survey

ABSTRACT

This work assesses the use of aerial imagery for the vegetation cover characterization in cork oak woodlands. The study was conducted in a cork oak woodland in central Portugal during the summer of 2017. Two supervised classification methods, pixel-based and object-based image analysis (OBIA), were tested using a high spatial resolution image mosaic. Images were captured by an unmanned aerial vehicle (UAV) equipped with a red, green, blue (RGB) camera. Four different vegetation covers were distinguished: cork oak, shrubs, grass and other (bare soil and tree shadow). Results have been compared with field data obtained by the point-intercept (PI) method. Data comparison reveals the reliability of aerial imagery classification methods in cork oak woodlands. Results show that cork oak was accurately classified at a level of 82.7% with pixel-based method and 79.5% with OBIA . 96.7% of shrubs were identified by OBIA, whereas there was an overestimation of 21.7% with pixel approach. Grass presents an overestimation of 22.7% with OBIA and 12.0% with pixel-based method. Limitations rise from using only spectral information in the visible range. Thus, further research with the use of additional bands (vegetation indices or height information) could result in better land-cover type classification.

Acknowledgments

Authors acknowledge the support of the project MEDSPEC (Monitoring gross primary productivity in Mediterranean oak woodlands through remote sensing and biophysical modelling, PTDC/AAG-MAA/3699/2014), and the research activities of the Forest Research Centre (UID/AGR/00239/2019) both funded by Fundação para a Ciência e a Tecnologia I.P. (FCT), Portugal, and PORBIOTA (Portuguese E-Infrastructure for Information and Research on Biodiversity) (POCI-01-0145-FEDER-022127), supported by Operational Thematic Program for Competitiveness and Internationalization (POCI), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (FEDER). FH was funded by the Erasmus+ France program (Erasmus+ traineeship 2017 Mobility Scholarship).

Declaration of interest statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Erasmus+ France program; Fundação para a Ciência e a Tecnologia I.P. (FCT), Portugal - PTDC/AAG-MAA/3699/2014; Fundação para a Ciência e a Tecnologia I.P. (FCT), Portugal - ID/AGR/00239/2019; Operational Thematic Program for Competitiveness and Internationalization (POCI), Portugal -  POCI-01-0145-FEDER-022127.