engleski

Quality Assessment of Worldview-3 Stereo Imagery Derived Models Over Millennial Olive Groves

Worldview-3 (WV-3) stereo-extracted very high resolution (VHR) DSMs represent state-of-the-art products in the domain of satellite-based digital surface modeling. The main goal of our research was a quality assessment of WV-3 stereo imagery derived DSMs of olive groves. Quality assessment was based on extraction of olive trees by geographic object-based image analysis (GEOBIA) and point cloud-based assessment of vertical accuracy.The quality of WV-3 extracted DSMs was assessed at the test area (31 870, 6m2)which represents the core of the Olive Gardens of Lun (Pag Island, Croatia). Areal accuracy of olive trees extracted by GEOBIA from WV-3 multispectral the image was evaluated for three different commonly used classification methods (Support Vector Machine - SVM, Random Trees- RT and Maximum Likelihood - ML) using the following indicators CORrectness, COMpleteness and Overall Quality. Vertical accuracy was evaluated through comparison of reference point cloud and WV-3 derived point cloud, using the M3C2 tool within CloudCompare. RT classifier has achieved the highest areal classification accuracy. Classification matching between reference data and WV-3 is 92.4%. The vertical accuracy of individual olive trees from WV-3 stereo imagery deviates from the reference model (STD = 2.57 m). STD values are lower for single, individual olive trees, than for grouped, dense olive tree canopies, as it is further confirmed with STD value for the whole test area. Performed quality assessment has shown that WV-3 stereo imagery can be used for the successful application of GEOBIA and mapping of olive trees. The creation of high- accuracy WV-3 derived models would allow efficient large-scale management and protection of this valuable agricultural resource.

Worldview-3 stereo imagery ; UAV photogrammetry ; VHR DSMs ; Quality assessment ; GEOBIA

nije evidentirano