Wednesday, June 10, 2009

White Dot 0001

This evening I cracked open what will probably become White Dot 0001, the first camera unit to be part of the program. Last week I took some test footage of birds with the two cameras I had available, a NexxusTech webcam ($15) and a Pentax Optio digital camera in movie mode (640x480).

I have quite a few birds in my backl yard, owing to some large trees surrounding our property. Birds are one class of aerial object that I want to rule out, hence the test footage.

The Pentax shoots in 640x480, but took surprisingly crisp images of birds as they zipped overhead, even birds that passed through the field of view in an eighth of a second were clearly identifiable as birds against the evening sky.

The cheapo webcam .. not so much. Just pointing it at the sky was enough to overwhelm the poor thing; all it showed was an enormous hazy mass, not distinct enough to capture clouds.

* * *

Enter the Logitech QuickCam Pro 9000. Following a recommendation from Andrew Kilpatrick that I consider some higher-end webcams, I picked up one of these from TigerDirect for just over $100 (that's Canadian). This afternoon it arrived, so I unpacked it and gave it a whirl.

This QuickCam was well reviewed because of its Carl Zeiss lens and its 2 megapixel CCD, which (in typical video chat settings, at least) produces very crisp video. My quick tests in my home office were encouraging - much clearer than the cheapo.

It's a much sturdier-looking device than the cheapo, with a folding base that is steadier than it looks. The cord is much more supple, also, which helps the camera stay where you put it.

As with many consumer peripherals, the drivers do a lot of the work. The driver software isn't too overbearing, and has relatively simple controls for controlling the camera's night-time settings, auto vs. manual focus.

The bundled motion detection is hopeless for my application; I suspected it would be. Even at its most sensitive, the camera wouldn't record until my fake-UFO LED, reflected in the window the camera was filming through, was close-encouters-of-the-third-kind huge.

Incidentally, this showed me just how surprisingly easy to use this approach to fake an interesting-looking light moving around in the camera's field of view. I'll post some soon to demonstrate. I think this validates my intention to have multiple, independent cameras tracking our white dots in the sky.

I started too late in the evening to get any good test footage of birds - birds to to bed around the same time as my kids, it seems - who knew? But just now I was able to capture a few frames of some air traffic from the direction of Pearson airport.

This video isn't exactly overwhelming me with its quality. The QuickCam's software only captures video to WMV, so despite its 2MP native resolution, the poor frame rate (I have yet to make a solid attempt to get it recording at the advertised rate of 30 frames per second) noise from the low lighting conditions, and the compression artifacts from WMV, the roving white dot in that video could be just about anything; the naked eye could clearly distinguish several flashing lights.

Here's a best-quality nighttime still:

If this is the best we can get for $100, that's going to be a bit disappointing; this means that cameras at this price range are going to be good for little more than spot-locating. And any prospect of star-based calibration is right out. Here's a close-up of a later shot, showing a bright star or planet:

Can you see it? No? You have to squint pretty hard at the point at the centre of the crosshairs. No way in hell is software going to pick that out of the data! It barely registers above the noise even when you know it's there.

And this is before I try it under Linux, where I assume the camera's performance will be hampered by using a non-Logitech driver, which (I wonder) may not be able to control the focus, night mode, etc.

Clearly I have my work cut out for me.

But I remain optimistic!