Metodologia para Avaliação de Rodovias Extraídas Computacionalmente em Imagens Digitais

JOSÉ LEONARDO MAIA, ALUIR PORFÍRIO DAL POZ

Resumo


The evaluation of cartographic features (semi-) automatically extracted from digital images is of great importance in the context of the validation of computational algorithms for feature extraction. The general procedure to evaluate the geometrical quality of cartographic features is based on the comparison between the entities extracted via computational algorithms and the corresponding ones extracted manually, the latter being referred to as reference features. Such comparison between the two groups of features is performed in three steps, i.e.: (1) calculation of corresponding points between the extracted features and the reference features; (2) estimation of quality parameters (completeness, correctness, quality, mean error (RM) and root mean squared (RMSE)) based on the corresponding points; and (3) analysis involving the quality parameters obtained in step 2. The evaluation methodology was tested in many situations, involving different image data as well as extraction methodologies with different characteristics. The results show that the methodology enables a very detailed evaluation of the results regarding the extraction methods.


Palavras-chave


cartography; digital images; road; road extraction; road network evaluation.

Texto completo:

PDF/A


DOI: https://doi.org/10.22456/1807-9806.19564

Apontamentos

  • Não há apontamentos.







Instituto de Geociências, Universidade Federal do Rio Grande do Sul
Av. Bento Gonçalves, 9500 - Prédio 43113
91.540-000, Porto Alegre, RS, Brasil
E-mail: pesqgeoc@ufrgs.br