Urban Areas Surveys
Unlocking The Potential Of Urban Surveys: The Fusion Of Multispectral And LiDAR Data
Urban landscapes are complex, requiring nuanced approaches for accurate mapping. Multispectral and airborne LiDAR data each have their merits, but their fusion can compensate for individual weaknesses. This work explores the integration of these datasets to achieve enhanced land cover mapping in large urban areas.
Feature Types And Classification: The Numbers Game
The study delves into LiDAR-derived features like height, intensity, and multiple-returns. While height features alone yielded an 83.17% accuracy, adding intensity and multiple-return data increased it to 87.69%. Even higher accuracy (94.59%) was achieved by integrating SPOT-5 and LiDAR data using all available input variables.
Feature Importance: NDSM Takes The Cake
Among the variables, the normalized digital surface model (nDSM) emerged as most beneficial for classifying land cover. The importance of features varied significantly depending on the land cover class, suggesting that a one-size-fits-all approach is inadequate.
Classification Uncertainty: A Nuanced Picture
While feature combinations generally reduced classification uncertainty, it’s not a silver bullet. The study found significant differences in uncertainty values between land cover classes, with notably low uncertainties for water features.