RAYFOS Ltd is a UK-based company founded in 2012 in the space of optical engineering and real-time processing systems. The core staff expertise is based on more than a decade of R&D in optics and metrology system design and the corresponding data analysis associated with large data processing operations, in particular large arrays of detectors such as video cameras. Rayfos is developing OEM solutions based on optical sensors and systems, advanced real time data acquisition and signal processing and high speed, scalable computation architectures. The company designs and produces integrated system-level solutions based on state-of-the-art sensor technology, enhanced with advanced data acquisition and novel signal processing algorithms.
While the model of operations today is Business to Business (B2B), Rayfos has initiated a pilot Business to Customer (B2C) model in 2014 with direct sales of software solutions around robust data processing tools developed by the company. The company plans to grow its B2C sector in addition to serving industrial partners and INNODERM can play a central role in growing this activity. Rayfos will engage in software, graphical user interface, data inversion, quantification and data handling operations associated with the demanding image reconstruction requirements proposed. Achieving fast computational performance in processing of large volume data, while still using a highly flexible off-the shelve Operating System is the key approach.
There are three areas of engagement as follows: 1. Efficient data transfer and storage and implementation of a no-missing frames data transfer scheme. Rayfos has a long experience in handling sustainable high-speed data streams, using smart memory managers and multi-threaded system design. 2. Fast signal processing operations implemented in a Graphics Processing Unit (GPU) using OpenCL optimized kernels to offer fast filtering operations necessary as a pre-processing step before inversion and 3. Fast matrix inversion operations associated with image reconstruction. This is a very challenging computational task given the amount of information available to the RSOM/RSOM2 system. Rayfos has significant expertise with inversion of large data matrices and will implement an ultra-fast inversion scheme based on Graphics Processing Unit (GPU) processing and utilizing parallel processing on precalculated matrices. The number of OpenCL devices used can be expanded by using multiple GPUs or even a mix of multi core CPU and GPU(s) in a single workstation if this is required.
Principal Investigator: George Georgakarakos