|
Résumé :
|
Research on remote sensing data quality has traditionally focused on thematic accuracy assessment. Data processing involves various steps and different types of data. Both spatial and thematic errors may occur at any level of a process. Thus, the quality of data may be decreased during the manipulation procedure as data are further processed. Recently, research has focused on error propagation, but mainly, within GIS. The overall objective of this research was error propagation using Monte Carlo simulation during rectification of remote sensing data. Remote sensing data included a SPOT XS image level 1B and a single aerial photograph. Taking into consideration three main errors: systematic, human, and image source errors, this research tried to: 1. Investigate relationships between both spatial and thematic accuracies considering the propagation of different types of errors, and 2. Search, whether, a model exists which relates to spatial and thematic accuracies. Error propagation and Monte Carlo simulation were performed during two main processes. The first process, using an affine and a twelve-parameter model, focused on geometric rectification of a SPOT XS level 1B with and without consideration of a DEM. The second process, using indirect differential rectification, was concerned with the generation of an orthophoto from a single aerial photograph. Furthermore, error propagation using Monte Carlo simulation was applied to soil surveys and a land use classification scheme. The results showed that error propagation and Monte Carlo could define a model describing relationships between spatial and thematic accuracies. The results also showed that associations between spatial and thematic accuracies can be described, according to the type of process performed and the error studied, by either a linear model, curvilinear, or two curvilinear models. Some cases in the process of orthophoto generation showed that error propagation and Monte Carlo simulation could not define a model describing relationships between spatial and thematic accuracies. The results showed also that appropriate selection and identification of ground control points are very important factors for a good quality rectification. Although error propagation using Monte Carlo simulation is a valuable tool, simulation still remains time consuming, especially when deriving a new image or a DEM using geostatistical techniques.
|
