Software engineers at MVTec Software (Munich, Germany) have released details of how the latest version of the company's Halcon 11 software package can be used to recognize objects based on their characteristic features such as size, color or texture.
The so-called sample-based identification technique makes it possible to identify objects without the need for them to carry bar codes or data codes.
In one specific test developed to demonstrate the power of the technique, the software engineers at MVTec used the software to classify four different classes of apples -- Braeburn, Gala, Jona Gold and Pink Lady -- all of which appear remarkably similar to the naked eye.
After the engineers trained the Halcon 11 software using just one apple from each class, the system was able to identify an unknown apple with an accuracy of 74.3 per cent. But because the recognition rate of the system increases with the number of training samples, when the system was trained with ten apples from each class, the recognition rate shot up to 89.5 per cent.
The Halcon 11 software -- which now supports the Mac OS X operating system -- will be on show at the company’s booth at the upcoming VISION 2012 show in Stuttgart in November.
More information on the software can be found here.
Related articles on the applications of MVTec's Halcon software you might also be interested in reading.
1. Linescan camera system scores a hole in one
Engineers at Strategic Automation have developed a system called the Sentry-1 which is capable of inspecting golf balls at speeds of up to 30 balls/min.
2. Vision system inspects x-ray dosimeter badges
Engineers at Helmholtz-Zentrum have developed a machine-vision system to automatically inspect the films in dosimeter badges.
3. Vision system sorts tomato seedlings
A vision-based inspection system is capable of sorting tomato seedlings at a rate of 18,000/hr.
-- Dave Wilson, Senior Editor, Vision Systems Design