A study of photometry and image formation for application in localization and mapping
Wong, Xue Iuan
MetadataShow full item record
This dissertation studies the image formation process and physically based light reflection model. Purpose of this study is to expand the concept commonly used in computer graphic community toward develop a set of surface estimation algorithm and relative navigation sensor interested by aeronautical and astronautical application. Simplified camera model has been extensively applied toward developing image based relative pose sensing algorithm who are widely used in navigation and simultaneous localization and mapping purpose. Beside of providing data of measurement, navigation sensors are also expecting to provide uncertainty measurement who describe the accuracy of output data. Since most of the image space algorithms are commonly rely on various of image processing technique such as feature extract and tracking, who are generally difficult in providing uncertainty measurement in real time, one of the major objective of this research is to develop a computational efficient method to estimate confidential level of sensor outputs. Base on the study of camera projection model and light reflection model, this dissertation develop set of three algorithms. Digital VISNAV who provide real time six degree of freedom relative pose measurement from a set of structured light beacons and a camera. Photometric stereo in motion exploit the possibility of applying photometric stereo algorithm toward estimate surface geometry model of space object. At last, a newly developed algorithm base on both camera projection model and surface reflectance model named as PhotoNAV is providing a new method for dense image base simultaneous localization and mapping application on textureless surface. Each of these proposed algorithm are providing both sensing output along with estimated uncertainty measurement of the output. Performance of these algorithms are evaluated by both simulation and experiment. Under certain constraint when real measurements are not available, the experiment is carry out by supplying synthetic image measurement generated from a ray tracer base scene simulator with physically base reflectance model.