Members
Overall Objectives
Research Program
Application Domains
Highlights of the Year
New Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography
XML PDF e-pub
PDF e-Pub


Section: New Results

Design of UHD Panoramic Video Camera

Participants : Carolina Da Silva Gomes Crispim, Rachid Guerchouche, Daniel Gaffé, François Brémond.

The goal of this work is to investigate the possibilities of designing a new camera-based system for retail. This work was carried in the context of a collaboration between STARS and Neosensys. The system is composed of several high definition cameras placed in a configuration such as it makes it possible to obtain a panoramic vision with 360° of field of view. The work was divided into 2 parts: theoretical part and practical part.

In the theoretical part, the different characteristics of the desired system were studied, such as: number of cameras, resolution of each camera, the different characteristics of the sensors (WDR, HDR, exposure) etc. Depending on these characteristics, data transmission through an IP network was addressed. In addition to the hardware characteristics, the possibility of embedding stitching capabilities was studied. After spending some time understanding the background behind such techniques, an existing implementation of the stitching was adopted. Simulations were then made in order to estimate the characteristics of an FPGA capable of handling 5 cameras with 12MP resolution each. An existing FPGA architecture extensively used in the industry was chosen and a mathematical model was provided in order to estimate the characteristics of such FPGA according to the different parameters of the camera-based system.

In the practical part, an implementation of the two first steps of the stitching algorithm (homograph estimation and warping) was performed on a FPGA using 2 cameras. The problems of code optimisation were addressed in order to achieve a functioning implementation with respect to the memory and computation capabilities of the FPGA.