Optimization of stereo image matching using genetic algorithms | |
| Author | Chen, Lei |
| Call Number | AIT Thesis no. CS-97-10 |
| Subject(s) | Image processing Genetic algorithms |
| Note | A thesis submitted in partial fulfillment of requirements for the degree of Master of Engineering, School of Advanced Technologies |
| Publisher | Asian Institute of Technology |
| Abstract | In computer vision and photogrammetry field, automatically finding the corresponding features in different images is one of the main difficult problem in photogrametry and computer vision. Conespondence is necessary for 2D and 3D measurement of an objection for objection detection, classification and identification. Further more, finding the conesponding feature which can be viewed as a searching problem is NP hard problem. The objective of this study is to apply genetic algorithms (GAs) to the above features matching problem and try to get more accurate matching results compared with other traditional methods. In this study, we developed a (GAs)-based system for the feature based stereo image matching problem. The feature points were detected by the SUSAN corner detector, then the matching solutions were represented by the chromosomes, which in our approach were arrays whose elements were labels and fitness value of matching points. The powerful GAs operators were applied to get a optimal feature points matching solution. The fitness function was designed considering both the photogrammetric and geometric constraints in order to remove the side-effect of only one constraint. A checking program was developed, which . incorporated both intensity and continuity constraints to improve the matching accuracy of GAs results. The system could easily accommodate the user requirements by adjusting the GAs parameters accordingly. Several experiments were conducted on six different pairs of stereo images, which contains different number of feature points and photogrammetric and geometric characteristics. The effects of different GAs parameters were discussed, also the results of GAs and those of Non-GAs were compared. The results showed that our system could be efficiently used in feature based stereo image matching problems. |
| Year | 1997 |
| Type | Thesis |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Yulu, Qi; |
| Examination Committee(s) | Murai, S. ;Sadananda, R.; |
| Scholarship Donor(s) | Asian Development Bank ; |
| Degree | Thesis (M.Eng.) - Asian Institute of Technology, 1997 |