Využití dálkového průzkumu Země pro klasifikaci vegetace nové divočiny v zázemí Kutné Hory

Dančejová, Daniela

Numerous areas in the Czech landscape have been abandoned by human activity, allowing natural processes to take over. Some of these areas have transformed into new wilderness characterized by diverse vegetation compositions, representing va- rious successional stages. The aim of this work is to conduct a comprehensive and accurate classification of vegetation in the new wilderness area utilizing remote sensing techniques. For this purpose, multispectral UAS data with a 5 cm spatial resolution, hyperspectral aerial data with a 60 cm spatial resolution, and botani- cal data collected at three different dates within the area of interest were used. Based on the collected data and the assessment of species separability, three clas- sification legends were proposed to classify the area of interest using Maximum Likelihood, Random Forest and object-based classifiers. The F1-score was used to assess the classification accuracy of vegetation classes. The results demonstrated the suitability of the object classifier for classifying a highly diverse vegetational area at a very high spatial resolution (achieving the highest overall accuracy of 84.06% across 22 classes). The Random Forest classifier yielded better results for vegetation classification on hyperspectral data with a lower spatial resolution...

guest :: login Digital Repository
		Search		Submit		Help		About