A new approach of generating orthoimage through polynomial model | |
| Author | Taravudh Tipdecho |
| Call Number | AIT DISS. no. SR-03-04 |
| Subject(s) | Polynomials Transformations (Mathematics) |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Technical Science. |
| Publisher | Asian Institute of Technology |
| Abstract | A new approach relying on "Global Transformation" followed by "Local Transfonnation" to generate orthoimages is presented. These transformations are proposed to perfo1m in both manual and automatic processes, which aim to avoid the complexity of extracting sensor parameters. First a process to simplify polynomial fu nction was established. This is because solving for the twenty constants of the original polynomial formula, defined by three parameters (x, y, z), became unduly difficult; a simplified form of a polynomial function with only eight constants is proposed to project pixels from the source image onto the digital surface model after which their object space coordinates are determined. Global Transformation as a mainly manual process was initiated. To produce an orthoimage with the proposed polynomial function, the process begins by employing all control points and tie points in the polynomial function. All points calculated and chosen are then used as a transformation model. Height and density considerations were two conditions involved in the transformation. The result of Global Transformation is an orthoimage with an accuracy of 1.5 pixels. However, several problems occurred within the process of Global Transformation, namely, failure to define a large numbers of tie points, heavy manual tasks, and impaired accuracy of output. These problems led to difficulties for large-sized study areas. Therefore, Local Transformation was established with the objective of automatic processing, including automatic positioning of additional tie points, point template and block processing. From these results, the problem of area size was solved, with an automatic increase in volume of tie points. Moreover, the accuracy of output also increased to 1 pixel. Following Local Transformation procedures, weakness of point allocation and incompleteness of orthoimage mandated fmther refinement of the model. Hence, consideration of similarity based on edge relation within a 5x5 grid and color interpolation were undertaken. A 5 x 5 grid window was employed to find similarity. The result was accuracy repmting at less than 1 pixel. In sum, the Global and Local Transformations are prom1smg procedures for generating orthoimages. With high accuracy of output and less manual processing, these procedures are acceptable for orthoimage production. |
| Year | 2003 |
| Type | Dissertation |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Space Technology Application and Research (SR) |
| Chairperson(s) | Chen, Xiaoyong |
| Examination Committee(s) | Yokoyama, Ryuzo;Huynh Ngoc Phien |
| Scholarship Donor(s) | National Electronics and Computer Technology Center (NECTEC), Thailand |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2003 |