Recent blog entries for wedesoft

I've released a Microsoft Windows version of the Ruby real-time computer vision library HornetsEye. Now HornetsEye not only runs under GNU/Linux but under Microsoft Windows as well! Currently Hornetseye for Windows is pretty basic since it does not have Video-I/O and cannot yet display images without the help of Qt4. If you are still interested you can already download a big package with installers for Ruby, HornetsEye, Qt4, and a lot more. See installation instructions for more information.

I have tried to create a software tool which makes implementing computer vision algorithms easy. You can even process images using the interactive Ruby shell. Under GNU/Linux it is already possible to write more sophisticated software such as a Webcam viewer on top of HornetsEye.

Enjoy!

I've dropped developing the graphical programming approach with Qt-designer. Instead I've started developing the Ruby real-time computer vision library HornetsEye. The approach of writing an extension to Ruby has so far gone quite well and it looks like I'm going to stay with this approach.

Version 0.2 of the Mimas computer vision plugin for Qt-designer was released. It is still relying on an unstable release of the real-time computer vision library Mimas. The new version allows to interactively use Gnuplot within Qt-designer. Here is a screenshot of the Gnuplot-plugin for Qt-designer.

Graphical programming in Computer Vision

I've created a first prototype of a plugin for Qt-Designer to graphically compose computer vision software. You can find some screenshots and demonstration videos on the MMVLWiki already. The software is based on the Qt/X11 graphical user interface and on the Mimas C++ real-time computer vision library and therefore will be released as open-source software (planned May 2006).

Computer Vision for Microscopes

The EU MiCRoN project has finished and the Computer Vision Software was delivered. To solve the vision task in a microscopic environment, Geometric Hashing was used as recognition algorithm and the Bounded Hough Transform for tracking in subsequent images.

The software is available under the LGPL here.

My private robot project still hasn't advanced lately.

However here is another demonstration of the work, which I'm doing at University.

My private robot-project hasn't advanced lately.
However there are news in context with the MiCRoN-project.
M. Boissenin (a PhD student) and me have achieved results, which may be a small breakthrough in the specialised field of computer-vision for microscopes.

For more information, see the results-page.

Unfortunately there are conflicting requirements on the kernel-version (by ndiswrapper for running Wireless-card and by the RTAI realtime operating system). I'll leave this problem for later.

I bought the PCAN interface, which is a buffered interface between a CAN-bus and a USB-port. Peak-systems are also offering an obfuscated-source driver (see driver-page).

The next step is a difficult one for me: I'll have to connect microcontrollers to a CAN-bus (and to sensors and actors of course).

So far I tried an Infineon-C167, which was at hand. I gave up, because there is no properly documented gcc-backend.

I'll have to try Atmel AVR next. It seems to have a strong developer-base. So I feel safer to go in this direction.

I've so far set up a Mini-ITX computer, which is (appart from an external power-supply at the moment) already fully self-sufficient. One can communicate with the Mini-ITX PC by wireless network. Here's the detailed documentation: documentation

Using CANBus and micro-controllers in the future, it should be possible to build a robot with Onboard-PC for less than 1000 Euros!

CANBus is used in professional applications (cars, industrial robots). F.e. see Kurt2 project.