Year: 2025
Type: article
Abstract: The study of the urban environment is undoubtedly the key to moving towards sustainable transformations. However, remotely sensed observations within such domain are complex and challenging, as these areas present many similar spectral characteristics, making image analysis of urban areas a difficult task. Although sensors systems have been recently improved, they are alone still unable to attain a sufficient level of detail to qualitatively and quantitatively analyze targets of interest in an urban image. In this sense, multisource data fusion emerges as a feasible solution for detailed detection and interpretation of elements that compose an urban scene. This work aims to perform data fusion using a hyperspectral image (HSI), an optical RGB ultra-high-resolution image, and Light Detection and Ranging (LiDAR) data for a detailed characterization of an urban environment under the perspective of land cover. Seven datasets will be employed, including the separate RGB, HSI, and LiDAR data as well as their fusion. The latter one is used to demonstrate the potential of integrating information from manifold sensors when compared with the accuracy results of a unique sensor. The algorithm chosen to perform such classifications is Random Forest since it can handle large amounts of data and achieve satisfactory accuracy. The overall accuracy reached by the data fusion set shows to be significantly superior to the ones obtained by the other datasets, demonstrating that the combined use of multisource data refines the classification results, allowing for an accurate and detailed level of classification legend.
Language: English
Subfield: Media Technology
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4410304174