The aim of this module is to acquire a basic understanding of Embedded Systems for Computer Vision and Robotic applications and to enable you to understand how to develop innovative and imaginative ways to exploit computer vision and embedded systems. Embedded computer vision is a niche application sub-field within embedded systems that requires a good set of skills and knowledge of embedded systems and computer vision. Demand for embedded systems in science and industry is increasing together with the increasing demand on automation, quality management, safety and efficiency.
First, we will introduce the fundamental concepts of Digital signal processing (DSP). DSP is a pervasive tool in the modern world, though much of its use is embedded within specialist software and hardware. Most modern instrumentation systems will employ DSP algorithms to analyze sensor readings, and in some cases (e.g. flight control systems) automatically initiate appropriate responses. Mobile phone technology is singularly dependent on the ability of DSP algorithms to extract meaningful information from broadcast signals. DSP algorithms underpin the revolution in the availability of digital video and audio recordings. Next, we will introduce the fundamental concepts of Real-time Programming and of performance engineering. The aim of Real-time Programming and Robust Feedback-based Control (where uncertainty in the dynamics and environment must be considered during the design process of algorithms) is to optimize the performance of algorithms such as object tracking in video surveillance, motion capture in sport activities, assisting living, automotive industry, aerospace, etc.
Several examples of algorithms will be introduced, such as: Visual Odometry, SLAM, RGBD-sensors, Robot Operating Systems.