A Preliminarily Study on Embedded Platforms for Computer Vision Applications

Today, silicon chips have become more and more powerful despite reduction in size. Various general processors, digital signal processors (DSPs) and graphic processing units (GPUs) have been embedded in countless electronic gadgets such as Personal Digital Assistants (PDAs), mobile hand-phone, digital cameras etc. With the high processing power of these embedded platforms and ever increasing size of both volatile and non-volatile storage, it opens up a great opportunity to integrate computer vision (CV) applications to these relatively low-cost and stand-alone devices. Applying CV applications inside electronic gadgets can be the next trends in industrial or commercial market. These devices can do specific smart and intelligent job, such as object or text recognition for manufacturing, surveillance and security, as well as entertainment. We found that there is still yet to have a standard embedded software or hardware architecture in embedded systems (ES) for CV. An open architecture is needed for CV researchers to develop CV application without worrying the underlying embedded platform or CV library dependency in Embedded System (ES). This paper provides a preliminary study of embedded computer vision (ECV) and tries to develop a possible framework for ECV. We also suggest a standard/open embedded platform for all the CV applications, which share the same embedded system characteristics and CV library.