I’ve decided to skip the intrinsic calibration and go right on with steps 4 and 5. I have run into a couple of things that have me stumped:
Step 4: What sequence of events would lead to “Extract planes” NOT outputting anything to the terminal? The first time I tried it it worked. Now that I’ve moved the Kinect a bit, it doesn’t. I have a feeling that I started a terminal the first time in some subtly different manner than all the other times, but I haven’t been able to get any coordintes since then.
Step 5: when I run KInectViewer I get a gray trapeziod, not an image that looks like the actual base plane of the sandbox. In the video, Oliver’s trapezoid looks like his floor, so I am expecting mine to look like the bottom of our (empty) sandbox. Well, the bottom with the computer’s desktop projected on it. But no. Here’s a hint: While I can manipulate my gray trapezoid using the Z key, I cannot orient it so that the intersection point of the axes is in the same plane. If I zoom in/out as much as I can, the trapezoid always appears in front of the crosshairs, never behind. Hopefully that’s a clue!
KinectViewer needs an already-working intrinsic calibration to map the color camera’s image onto the depth camera’s geometry. If no intrinsic calibration parameters are present, the 3D geometry will appear in some random color. You can bootstrap the calibration procedure by downloading your Kinect’s factory calibration data using sudo ./bin/KinectUtil getCalib 0
In general, skip the intrinsic calibration procedure (and any other instructions contained in the Kinect and SARndbox packages themselves) and stick with the complete AR Sandbox installation instructions. Factory calibration is good enough for the AR Sandbox, as the sandbox completely ignores the color camera.
To your issues in step 5: Check the “Using Vrui Applications” guide in Vrui’s packaged HTML documentation, or online here.