hybridroulettecomputer.com V3 Development

[Steve]

Development of the Hybrid V3 roulette computer is progressing very well (:.hybridroulettecomputer.com). The automated detection of the wheel in video, and automated prediction process is extremely complicated - especially since the angle of the video will vary, footage shakes, converting 60 frames per second video and ball position to precise timings etc.

One of the most challenging parts is for the computer to identify and track a particular wheel, and this part is now done. At :.hybridroulettecomputer.com/v3.wmv is a video demo of the Hybrid V3 tracking a wheel in a video. It demonstrates:

    * How the wheel is selected and tracked: Importantly for the tracking, I selected a wheel in the far background and NOT the one right in front of the camera bearer (to make it much more difficult) and still you see it is easily tracked. The spins from the far back wheel can comfortably be predicted with a suitable video resolution.

    * How the computer knows if an object moves in front of the wheel: You can plainly see when the person moves in front of the wheel, the computer knows this and keeps the detection box in place. In such a situation, the camera-bearing player can easily step around the person as needed because the computer automatically learns what the wheel looks like from multiple angles. There is no need to aim the camera, and as you can see from the video, even a distant wheel which appears small in the video is easily tracked. This is a significant improvement on earlier models.

While this development is completed sooner than expected, there is still an enormous amount of work to be done. Now that the wheel position can be tracked, the computer needs to observe inside the wheel detection area to track the actual ball and rotor speed. This is actually easier than what is done in the video demo especially as we can use programming source code from earlier versions.

Nevertheless, the actual hardware requires a lot of development to make it small enough to be bodyworn. Overall, I expect full completion in 3-4 months.

[Steve]

Overall we are about 50% complete. We tweaked some of the reacquisition algorithms - this is where the wheel is not in the video (if the camera wearer turns away), then the hybrid needs to determine where it is again. A video of this component in action is at:

:.hybridroulettecomputer.com/reacq.wmv

In this video, the left part shows the live video (in this case, just a video file). The top right shows what is being tracked. The bottom right shows the wheel being tracked in real-time.

During the video I click "Restart" to show reacquisition of the wheel. It is 100% reliable now. The end product will be able to track the wheel, ball and green zero even if the camera wearer does cartwheels, literally. The work achieved so far is military grade.

There is a lot more that is done that I wont show publicly. But I'm very happy with progress. Completion may be in roughly a month (although it is difficult to say for sure), but then we'll need to thoroughly test everything and tweak for improvements.

[speed]

Hi Steve,  I see the hybrid uses a camera to measure the speed of the rotor and ball, so I wonder is it  possible to make a some lite hybrid version for android or iphone mobile platform ?

[Steve]

we already did testing and no current phone is powerful enough to do what is needed

[subby]

With the new quadcore phones coming online it might be an option sooner than later

[speed]

How the phone should be powerful if it is not secret, there is some android phones that have 1GB OF RAM and can overclock to 2 GHz. Is your hybrid written in VB.Net 2008 ?

[Steve]

none of the current smartphones are fast enough without needing to take some shortcuts. yes that is likely to change soon.

it is not written in vb.