City Research Online

Abstract

Estimation of the size, shape, and location of an object(s) using images acquired from multiple view points is required in many close range measurement and computer vision tasks. The first step in this process is the extraction of features in the images. In the second step, the features in images that correspond to the same 3-D points are identified. This identification process is termed the establishment of correspondences.

In this thesis a geometric technique, based on image rectification principles, for the establishment of multiple view point correspondences is developed. The technique uses parallel epipolar lines in rectified image space hence, it is termed the PEL algorithm. This technique can establish faster image point correspondences than the usual epipolar line or 3-D space intersection techniques. This is due to avoiding the non-linear search. An initialisation process, which is not required by the epipolar line or the 3-D space intersection methods, is required for the PEL algorithm to pre-estimate rectification and search band window parameters. Hence, the speed performance of the technique is better when the number of points is larger than a certain value which is depending upon the software implementation and the speed of the computer used. The technique has been rigorously tested with both simulated and real-image data under various situations.

Another aspect of the work conducted for this thesis contributed to the development of a real-time multiple camera 3-D measurement system. In conventional frame grabber based multiple camera systems sequential image data acquisition is a time consuming task. In order to achieve frame rate functionality, dedicated hardware based on a Digital Signal Processor for real-time video signal processing and data communications were developed. The dedicated hardware plus a CCD camera is termed an intelligent camera. In a multiple camera system each intelligent camera processes images simultaneously and the processed data are transferred to a central computing unit for further processing. This system provided the context for the development of the PEL algorithm. The real-time performances of the system was analysed and the PEL algorithm was tested using the data acquired from multiple intelligent camera system.

Publication Type:	Thesis (Doctoral)
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments:	School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses School of Science & Technology > Engineering

Export

Downloads