LARS Tech Report Number
This thesis contains the results of an investigation performed in order to define causes of geometric defects within digital multispectral scanner (MSS) data arrays, to analyze the nature of the resulting geometric errors, and to investigate restitution methods to correct or reduce such geometric errors.
The thesis includes a review of digital multispectral scanning systems, including the recording and digital form for computer aided analysis. Causes of geometric errors within the data arrays are presented, and previous investigations of geometric aspects of remote sensing systems are reviewed.
The introduction of geometric transformation relationships for scanned data, from which collinearity equations for MSS may be derived, serves as the basis of parametric methods of analysis and restitution of MSS digital data arrays. The linearization of these collinearity equations is presented, including consideration of the functional assumptions made in order to model the stochastic changes in the exterior orientation of the sensor down the flight line.
Parametric algorithm for analysis and restitution based upon the above analytical treatment which were considered are: the direct use of MSS collinearity equations, and the use of piecewise polynomials based upon the linearized collinearity equations. In addition, nonparametric algorithms for restitution are introduced, as an alternative to the parametric algorithms.
A proposed system for the geometric analysis and restitution of MSS digital data arrays is introduced. This procedure was used to test the methods of analysis and restitution, utilizing actual MSS data arrays from two aircraft flights. The results of these tests indicate that the collinearity equations can yield acceptable results when utilized for the analysis and restitution of such arrays. The investigation indicates that the piecewise polynomial algorithms are in general inferior to other methods of restitution. The nonparametric algorithms show great promise for the restitution of these arrays since the resulting accuracy of restitution is comparable to or better than that using collinearity, and they possess definite advantages in computational efficiency. The arithmetic mean algorithm appears to perhaps represent a particularly efficient nonparametric algorithm for restitution of MSS digital data, at least for the arrays tested.
Date of this Version