Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Dissertation
Method for Attaining Object-Side Light Fields and Metric Depth Maps from a Commercial Plenoptic Camera
Doctor of Philosophy (Ph.D.), Drexel University
Feb 2018
DOI:
https://doi.org/10.17918/D8Z084
Abstract
Determining object distances and positions is a common computer vision task. This has obvious use in tasks such as autonomous navigation, robotic object retrieval, and surface profiling. Traditional methods for obtaining scene depth information may require active illumination, multiple cameras, or multiple exposures as hardware parameters are incremented. A plenoptic camera can be used to obtain scene depths and has the advantages of being passive and only requiring a single exposure with a single device. Traditional photographic cameras integrate all light arriving at a single detector element or film location, discarding information about the incidence angle of any individual contribution. In a plenoptic camera, an array of microlenses is placed in front of the image sensor, such that light arriving at a given microlens from different directions will arrive at different detector elements. With this, a signal can be attributed to all rays arriving at the microlens array, for which the assemblage is referred to as a "light field." Common light field operations include the creation multiple digital images from one exposure to simulate the effect of changing the camera's focal length ("digital refocus") and the determination of relative distances to scene objects. Existing light field methods are concerned with the light field within the camera (the "image-side light field"), as this is sufficient for most tasks. In contrast, this thesis describes a method for characterization of the light field external to a plenoptic camera (the "object-side light field") by performing experiments that determine the mapping from object-space rays to image-side light field coordinates. Use of an object-side light field allows for the determination of distances to scene objects in absolute terms, rather than the relative depths commonly obtained with image-side light fields. The object-side light field is obtained by observing the appearance of an occluding edge in the image-side light field as it is translated across the plane of the camera's front lens. Perspective images are created for simulated pinhole apertures at the front lens plane and used to view a patterned target, completing the characterization. In this thesis, descriptions of the characterized object-side light field are provided, offering an intuitive understanding of the structure of the light collected by the plenoptic camera. Demonstrations of common light field operations are provided to confirm that there is no loss of functionality through use of an object-side light field rather than an image-side light field. Simulations confirm that the distance to objects near to the camera can be determined to within 1~mm, with accuracy decreasing as the distance to an object increases.
Metrics
19 File views/ downloads
14 Record Views
Details
- Title
- Method for Attaining Object-Side Light Fields and Metric Depth Maps from a Commercial Plenoptic Camera
- Creators
- Weston Aenchbacher - DU
- Contributors
- Timothy P. Kurzweg (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- xx, 223 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Electrical (and Computer) Engineering [Historical]; Drexel University
- Other Identifier
- 8175; 991014632617604721