Efficiency assessments of GEOBIA in land cover analysis, NE Hungary

Varga, Orsolya Gyöngyi [Varga, Orsolya Gyöngyi (Geoinformatika, G...), szerző] Természetföldrajzi és Geoinformatikai Tanszék (DE / TTK / FoldtI); Szabó, Szilárd [Szabó, Szilárd (Földtudomány), szerző] Természetföldrajzi és Geoinformatikai Tanszék (DE / TTK / FoldtI); Túri, Zoltán [Túri, Zoltán (Földrajz), szerző] Természetföldrajzi és Geoinformatikai Tanszék (DE / TTK / FoldtI)

Angol nyelvű Tudományos Szakcikk (Folyóiratcikk)
    Land cover features of landscapes can be analysed thoroughly by different remotely sensed data. In our research we examined the land use/land cover (LULC) changes in three randomly selected areas located in different microregions of the Great Hungarian Plain,NE Hungary. We intended to observe the efficiency of object-based image analysis compared to the pixel-based method by applying them to archive aerial photos and digital true colour orthophotos representing the different landscape types (alluvial plain, terraced and loess plain, sandy region with partly fixed dunes) of the study area. We applied the bottom-up multiresolution segmentation technique in eCognition and classified the evolved segments based on visual interpretation. We used supervised classification methods in order to compare their efficiency with our results. Production of land cover maps based on segmentation and non-automated classification processes and accuracy assessment results confirmed that the object-based approach was more efficient (with a total accuracy value of 97.2%) and even better than the maximum likelihood classification, which had the best performance (total accuracy: 62.8%) of all the supervised pixel-based methods applied.
    Hivatkozás stílusok: IEEEACMAPAChicagoHarvardCSLMásolásNyomtatás
    2020-08-14 19:17