Analyzing the Potential of Multi-Spectral Street-View Imagery

Monitoring urban environments is vital for developing climate adaptive cities. The current monitoring approaches rely on human inspections, embedded sensors, spectrally rich remote sensing imagery, spectrally limited street-view imagery, or a combination of a subset of these. A large number of urban environment parameters exhibit bio-physical characterstrics which are distinguishable in spectra not present in the limited RGB domain. In this work, we explore the value addition of multi-spectral street-view imagery, and its combination with spectrally rich remote sensing imagery. Our experiments for the complex task of urban tree health condition monitoring for two urban areas in the Netherlands yields promising results with the fusion model utilizing both modalities of multi-spectral street-view and VHR satellite-view imagery performing better than models utilizing each modality alone.