Affordable Access

deepdyve-link
Publisher Website

A Spectral Database for the Recognition of Urban Objects in Kaunas City: Performance and Morphometric Issues

Authors
  • GADAL, Sébastien
  • Mozgeris, Gintautas
  • Jonikavicius, Donatas
  • Kamičaitytė, Jūratė
  • Ouerghemmi, Walid
Publication Date
Sep 23, 2020
Identifiers
DOI: 10.5557/e01.2669-1922.2020
OAI: oai:HAL:hal-02951336v1
Source
HAL-Descartes
Keywords
Language
English
License
Unknown
External links

Abstract

The possibilities of use of urban objects spectral library as an important component of city’s knowledge database are extensive. They are used for the recognition, characterisation and identification of the urban objects and spaces in different towns. But their geographic accuracy is dependent on the study’s area, and possibilities of using in another city are often restricted. Use of urban objects spectral libraries is limited because of the sensors, the geographic and temporal variabilities of the object’s spectral signatures, the environmental conditions, or the remote sensing data. The difficulty of standardisation of the existing spectral databases and sensors fostered us to create a specific urban object spectral database for Kaunas. Despite the fact that spectral databases of urban objects have a low level of transposability to another urban territory, the specificity of the urban environment of Kaunas constitutes another main interest: the diversity of urban materials, contemporary and historical urban structures, urban planning peculiarities and spatial complexity of post-Soviet Baltic cities permitted the development of a "broad spectrum" object-oriented spectral database. The created spectral database was tested applying conventional artificial intelligence algorithms used in remote sensing and image processing. The results obtained on the recognition and characterization of urban materials and buildings are substantial. The performance of the recognition of built-up areas is improved with the use of the morphometric urban object database. The performance of the detection of urban vegetation is poor or average according to the species. This is due to the differences in the scalar level of the spectral measurements between the spectral library created and the spatial resolution of the airborne hyperspectral images.

Report this publication

Statistics

Seen <100 times