It’s been quite a while since my last webcam tracking post but some magic has happened behind the curtains and two projects emerged during the last weeks which I don’t want to leave unmentioned.

The guys at Neue Digitale did a pretty sweet job with their Audi Quattro Urban Curving Online Special which offers a game that lets you steer your way down steep roads by tilting your head left and right with the help of my tracking code.

Mariusz Kreft and Matthias Gomille from argonautenG2 used my code for an awesome banner prototype as an entry for the Young Lions Award in Cannes. Let’s hope they rock the contest.

Here’s their Home Sweet Home Cinema website:

I’ve got a small update for my tracking code in the pipe which both of these projects are using. It’s basically a little cleaner than the current public version and I will probably release it in a few weeks. I’m also looking into Haar tracking algortihms which allow face recognition, but I couldn’t decide wether I want to implement a C++ Eninge (maybe OpenCV?) via alchemy or try it in AS3. I’ll look into that when I have some spare time after work at Big Spaceship, so stay tuned.

I just uploaded a little demo video of my Flash webcam motion tracking project for all you guys out there without a webam, just to give you an idea of how it works.

Lay back and watch the demo video or set up your webcam and play a session of Webcam-Pong yourself.

Due to numerous requests I have decided to publish the source code of my little tracking project. Unfortunately I haven’t had the time to complete the documentation and feature set yet. Nonetheless, here it is:

(more…)

I’ve recently been continuing work on my webcam tracking project and just implemented a lightweight version of the Pong game I created for Sony HD-Test by vektorfarm.

Just track an object with your custom settings and press any key to release the pong ball.

Happy New Year! To master my hangover I updated the tracking project for my studies and added some new features to the test environment:

• custamizable tracker size, tolerance and minimum number of pixels found to recognize an object
• automatical adjustment of tolerance and min. px found relative to the brightness of the webcam image
• histogramm showing the actual brightness distribution
• panel showing the tracked object’s path

I’m currently working on object detection and tracking in Flash for a project in my studies. The current implementation uses the Meanshift algorithm (without color adaption) and delivers a decent performance while only looking for certain colors in the local tracking area. Tweaking is still necessary, but I thought I might already show it as is.

Here is the Meanshift version (click on your hand or face if neither is found at startup):