Stochastic modeling and triangulation for an airborne digital three line scanner

Won Jo Jung, Purdue University

Abstract

The 3-OC is one of the newest digital three line scanners on the market. Unlike other three line scanners using a single optical system, the 3-OC uses three different optical systems moving together. Therefore, this thesis aimed to develop a photogrammetric model for the 3-OC. To precisely relate ground space and the corresponding image space, all the exterior orientation (E.O.) parameters of image lines need to be estimated using a bundle block adjustment. The biggest hurdle in this problem is the large number of exterior orientation parameters because one image strip of the 3-OC usually contains tens of thousands of lines. To reduce the number of unknown E.O. parameters, the E.O. parameters of all the three cameras at an instant imaging time were represented by transformed parameters with respect to the gimbal rotation center. As a result, the unknown E.O. parameters were reduced to one third of original number of parameters. However, the number of E.O. parameters is still too big and estimating these E.O. parameters requires enough observations which are practically very difficult to obtain. To resolve this problem, there have been two kinds of approaches. One is reducing the number of unknown parameters and the other is providing fictitious observations using a stochastic model. As the title of this thesis implies, a stochastic trajectory model was implemented in this thesis. The stochastic relationships between two adjacent lines, as described in previous work, were expanded to the stochastic relationships between two adjacent image observations, so that the E.O. parameters of the lines between two adjacent observations can be recovered by interpolation. By providing enough pass points, it was possible to recover all the E.O. parameters accurately. In addition, the number of unknown E.O. parameters was drastically reduced as well. In this thesis, aerial triangulations of the suggested photogrammetric model were performed with self-calibrating some of the system parameters. As a result, the exterior orientation parameters were successfully estimated and the system parameters were calibrated as well. The RMSE of image misclosures on check points was less than 1.2 pixel and the RMSE of ground misclosures at check points was less than 0.6 ft (nominal GSD is 0.5ft).

Degree

Ph.D.

Advisors

Bethel, Purdue University.

Subject Area

Civil engineering|Remote sensing

COinS