Implementing Machine Vision Systems with a Dynamically Typed Language

Authors: 
  • Jan Wedekind - Sheffield Hallam University, United Kingdom
  • Jacques Penders - Sheffield Hallam University, United Kingdom
  • Hussein Abdul-Rahman - Sheffield Hallam University, United Kingdom
  • Martin Howarth - Sheffield Hallam University, United Kingdom
  • Ken Dutton - Sheffield Hallam University, United Kingdom
  • Aiden Lockwood - Sheffield Hallam University, United Kingdom
Time: 
Thursday, July 28, 11.00 am - 12.30 pm
Location: 
George Fox, Lecture Theatre 4

HornetsEye is a machine vision extension using existing free and open source software (FOSS) which facilitates capture, processing, display, and recording of live video. This work is about combining the productivity gains of a modern scripting language and real-time performance in a single system.
The core of the system is implemented in Ruby. In order to achieve high performance, the GNU C compiler is used for just-in-time compilation. The system provides an internal Domain Specific Language (DSL) for manipulating uniform arrays.
Finally the system comes with a comprehensive set of classes for input and output. There is support for DC1394-, V4L-, and V4L2-compatible cameras, ALSA sound devices, image and video files, the Microsoft Kinect, and X.Org display. Furthermore there is Qt4-integration for developing graphical user interfaces.