The acquisition software I wrote has a lot of cool features. It tracked the sampling and reported under-runs (delay). It constantly updated a webpage with status statistics and the latest data products. It's also robust; running for years at various shore stations (accumulating unmanageable TB's of data ;). The image protocol was switched from NTSC (coax -> framegrabber) to 1394 (firewire) to GigE, as we acquired more modern cameras. The code is modular, so it's pretty easy to switch protocols. GigE is great because it allowed us to have long cable lengths for cheap. The latest version of the code is using a paid Active X plugin to handle the GigE protocol. Recently, Matlab has released GigE support, so an update could drop the paid plugin. The mid-range machine vision lenses I worked with were always a disappointment in terms of quality and performance. An additional difficulty was using machine vision cameras in an outdoor (variable) environment. This equipment is designed to work under fixed lighting and object distance, etc. No auto-exposure. The cctv cameras offer auto-exposure, but the combination of price, image quality, and analog interface format wasn't acceptable at the time (~2006). The increasingly digital market has undoubtedly changed. A consumer-grade personal camera would offer far superior images, but at 2-3 s intervals at best. A professional SLR can deliver < 1 s interval, but I doubt the moving parts would last more than a week of continuous operation.  Another issue is aperture, which directly affects the depth of field and exposure time. Depth of field was never a problem since the image object was always at a large distance. Finding an aperture that would allow for both high-noon and sunset light levels always required a bit of adjustment. My method was to maximize the aperture at peak sunlight with nearly the shortest exposure time. This gave the sharpest images and allowed for operation earlier/later in the day. |
