Precision of digital orthoimages: Assessment and application to the occlusion detection problem

This research investigates the problem of automatic assessment of the precision of digital orthoimages. A mathematical model is designed for that purpose. The problem of automatic detection of occlusions in digital images is also addressed as an application of the assessment of the precision of a digital orthoimage. A probabilistic approach--through a Bayesian network--is implemented for that purpose.;The study is divided into six chapters, organized as follows: (1) an introductory chapter that establishes the context of this research, reviews the literature on the production and testing of digital orthoimages, states the general and specific objectives of this work, and shows the gaps to be filled by this research; (2) a study of the error sources in the production process of digital orthoimages; (3) a discussion of the development of a mathematical model for assessing the precision of digital orthoimages based upon the computation of the error components and their propagation through the digital orthorectification process; (4) an application of the assessment of the precision of orthoimages to the problem of automatic occlusion detection in digital images; (5) a description of the tests implemented and their respective results, and (6) a conclusion presenting the relevancy of the study, a summary of the main findings, the potential applications of this research, and suggestions for further work.;The results validated the mathematical model being proposed. Therefore, assessment of the precision of a digital orthoimage can be performed through the mathematical model implemented in this research. In fact, not only can the precision of a digital orthoimage be assessed but the error in every pixel of a digital orthoimage can also be computed. The results also proved that occlusions in digital images can be automatically detected. This use of a probabilistic approach for solving that problem through a Bayesian network has no previous references in the literature.