Sammanfattning

Urban trees are a key component of the urban environment. In Sweden, ambitious goals have been expressed by authorities regarding the retention and increase of urban tree cover, aiming to mitigate climate change and provide a healthy, livable urban environment in a highly contested space. Tracking urban tree cover through remote sensing serves as an indicator of how past urban planning has succeeded in retaining trees as part of the urban fabric, and historical imagery spanning back decades for such analysis is widely available. This short study examines the viability of automated detection using open-source Deep Learning methods for long-term change detection in urban tree cover, aiming to evaluate past practices in urban planning. Results indicate that preprocessing of old imagery is necessary to enhance the detection and segmentation of urban tree cover, as the currently available training models were found to be severely lacking upon visual inspection.

Nyckelord

urban forestry; urban trees; historical imagery; automated detection; GIS; deep learning

Publicerad i

Landskapsarkitektur, trädgård, växtproduktionsvetenskap: rapportserie
2024, nummer: 2024:5
eISBN: 978-91-8046-930-2
Utgivare: Faculty of Landscape Architecture, Horticulture and Crop Production Science, Swedish University of Agricultural Sciences

SLU författare

• Klobucar, Blaz

  • Institutionen för landskapsarkitektur, planering och förvaltning, Sveriges lantbruksuniversitet
    

UKÄ forskningsämne

Fjärranalysteknik
Landskapsarkitetktur


Publikationens identifierare

DOI: https://doi.org/10.54612/a.7kn4q7vikr

Permanent länk till denna sida (URI)

https://res.slu.se/id/publ/129259