| Abstrakt: |
This paper aims to investigate the possibility of automatic identification of the forest main group species by the use of Airborne LiDAR data (ALS) and multispectral aerial images recorded by unmanned airplanes. The case study is located in South West Romania, in the Mic Mountain area, in the forests belonging to the Production Unit 6 Cuntu, Experimental Base of Carasenbeş. Three research plots with a size of 1 ha (100 x 100 m) each were placed within representative mixed stands of coniferous and deciduous tree species. The tree species within the plots were measured using the FieldMap equipment, the tree heights were recorded by the use of the Vertex IV ultrasound dendrometer, and the diameters were measured with an electronic caliper. Airborne LiDAR data has been processed using specific algorithms, resulting the digital terrain model (DTM), the digital surface model (DSM), and the normalized canopy height model (CHM). Very high spatial resolution images were captured using an eBee RTK device equipped with a CanonS110 RGB camera, in the case of the 2016 flight, or with a CanonS110 NIR camera in the case of the 2017 flight. The Object-Based Image Analysis (OBIA) supervised classification was performed for both RGB and CIR images through several stages, presented as a logical succession of specific processes. The OBIA classification of the RGB image gave better results compared to the OBIA classification of the CIR image, the overall accuracy of the RGB image classification was up to 16.7% higher than in the case of the CIR image. The object-based image analysis gave promising results in terms of automatic identification of the main group trees species, indicating a high level of standardization as well as the ability to be replicated under various conditions (different study areas and / or types of images). [ABSTRACT FROM AUTHOR] |
