TY - JOUR
T1 - Thematic accuracy consequences in cadastre land-cover enrichment from a pixel and from a polygon perspective
AU - Serra, P.
AU - More, G.
AU - Pons, X.
PY - 2009/1/1
Y1 - 2009/1/1
N2 - In this paper, cadastre agricultural cartography was enriched using crop raster maps obtained from remote sensing images. The work demonstrates the implications of applying two new terms: fidelity and purity. Per-pixel classifications and polygon enrichments were compared taking into account: (a) the consequences of using a more or less conservative strategy at the classification stage, using fidelity, and (b) the consequences of using modal thresholds at the enrichment stage when deciding which category each polygon is to be assigned to, using purity. More than 300,000 pixels and 2,800 polygons were used to measure the thematic accuracy of ten agricultural categories by means of confusion matrices. These were computed at pixel, polygon, and area level. Thematic accuracy was calculated in the classical way and without taking into account unclassified pixels as errors, as well as by paying special attention to the consequences for the classified area. The results show that polygon enrichment is a useful methodology, achieving thematic accuracies of 95.6 percent, when optimum parameters are used, while classifying 87.4 percent of the area. © 2009 American Society for Photogrammetry and Remote Sensing.
AB - In this paper, cadastre agricultural cartography was enriched using crop raster maps obtained from remote sensing images. The work demonstrates the implications of applying two new terms: fidelity and purity. Per-pixel classifications and polygon enrichments were compared taking into account: (a) the consequences of using a more or less conservative strategy at the classification stage, using fidelity, and (b) the consequences of using modal thresholds at the enrichment stage when deciding which category each polygon is to be assigned to, using purity. More than 300,000 pixels and 2,800 polygons were used to measure the thematic accuracy of ten agricultural categories by means of confusion matrices. These were computed at pixel, polygon, and area level. Thematic accuracy was calculated in the classical way and without taking into account unclassified pixels as errors, as well as by paying special attention to the consequences for the classified area. The results show that polygon enrichment is a useful methodology, achieving thematic accuracies of 95.6 percent, when optimum parameters are used, while classifying 87.4 percent of the area. © 2009 American Society for Photogrammetry and Remote Sensing.
U2 - 10.14358/PERS.75.12.1441
DO - 10.14358/PERS.75.12.1441
M3 - Article
VL - 75
SP - 1441
EP - 1449
JO - Photogrammetric Engineering and Remote Sensing
JF - Photogrammetric Engineering and Remote Sensing
SN - 0099-1112
IS - 12
ER -