A novel distributed architecture for IoT image processing using low-cost devices and open internet standards

Abstract

Industry 4.0 can be defined as the integration of computers and automation to current industrial processes, with addition of smart and autonomous systems leveraged by machine learning techniques. In this scenario, a compact, dependable and fast controller is desired, featuring low energy consumption, easily programming and maintenance, with no mobile parts. Nowadays, computing power in single board computers, e.g. the Raspberry Pi among others, has been increased at a very important rate. In just three generations, Pi computers offer almost a two-fold speed gain, when compared to first models. Its design, an underlying video driver with general capabilities of regular OSes, makes them quite suitable to build image processing systems at very low cost, with no mobile parts and low energy consumption. However, designing such a system for industrial image processing is a tough challenge, since it implies to integrate cameras, image processing libraries, database servers and application software with graphical user interface, in an already resource constrained device. This work presents a new architecture for this kind of systems, by means of open internet standards, using a self-contained, high performance web server to publish a RESTful API and a set of web pages that use latest HTML5 capabilities to manage USB webcams and system data. This proposal also integrates OpenCV as a compiled script on client-side using the new WASM paradigm, with an optimized storage for images using -industry-standard RDBMS and a modular design that can target Windows and Linux as well.