A Miniaturized Insect Eye Inspired Multi-camera Real-time Panoramic Imaging System

During the last 50 years miniaturization became a key element in human history, since it opens doors for manufacturing new devices that enhances the quality of human life. Camera and imaging systems are following this miniaturization trend as well. Meanwhile, in the imaging domain, usage of multiple aperture camera systems are gaining significance in every aspects of daily life such as entertainment, surveillance, and medical imaging field. Many works focus on multiple camera panoramic and wide field of view imaging systems in industry and academy. Insect vision is the magnificent example of natural multi-aperture wide angle of view imaging systems. There are many attempts to mimic the insect vision capabilities. Current multi-camera systems that are utilizing off-the-shelf components are big in scale and the miniaturization limits are not explored. On the other hand, the multi-aperture systems fabricated using micro-machining techniques cannot meet high resolution requirements due to the micro-machining precision and optical limitations. This thesis discloses a set of methods to enable development of miniaturized, multiple camera, large angle of view imaging systems. A second target is to explore the smart vision capabilities of the proposed imaging system such as automatic detection of objects of interest. The main methodology is combining the real-time image processing techniques with off-the-shelf miniature cameras. The presented work includes the methods for combining many miniature cameras mechanically to have a compact vision system similar to the insect eyes. Moreover, image processing techniques for creating high quality panoramic images and extracting useful information from the multiple camera images are applied. Furthermore, digital hardware system design methodologies are implemented for real-time panoramic video generation from the multiple camera video streams. FPGA implementation of the methods are performed and tested and a migration of the system from FPGA to ASIC design is achieved. In the scope of the thesis work, the proposed methods are implemented and tested by constructing a 5 mm radius hemispherical compound eye, which is capable of imaging a 180°x180° field of view at 18 mm radial distance. An FPGA implementation of the image processing system is performed, which is capable of generating 25 fps panoramic video with 1080x1080 pixel resolution at a 120 MHz processing clock frequency. When compared to the insect eye mimicking systems in literature, the system proposed features more than 1000x resolution increase within the same or even smaller physical dimensions. What is more, by utilizing fiber-optic technology, a built-in illumination capability is added to the compound eye. This is the first time that a compound eye with built-in illumination idea is reported. With this work, the current limits of off-the-shelf component based methods in terms of physical dimension and resolution are explored for multiple aperture, miniature insect eye mimicking vision systems. The system is tested inside a human colon model for endoscopic applications like colonoscopy where there is a need for large field of view high definition imagery. The possible applications are not limited to medical domain and due to its miniature size and high quality video capabilities, the proposed methods and the system built can be utilized in search and rescue systems and robotic applications.

Leblebici, Yusuf
Lausanne, EPFL
Other identifiers:
urn: urn:nbn:ch:bel-epfl-thesis6844-4

 Record created 2015-11-17, last modified 2018-03-17

Download fulltext

Rate this document:

Rate this document:
(Not yet reviewed)