high altitude image acquisition w/ limited bandwidth

[Coke]

I am currently developing the image acquisition system for a ultra long duration balloon (ULDB) that will inspect the health of the balloon structure throughtout the mission. I have a very limited bandwidth to work with for image downlink, significantly less than 2400 baude at most times, and a complete image of the balloon must be available for review every hour. For sufficient resolution, and considering the balloon is roughly 400ft by 400ft when completely inflated, I desire roughly 5 pixels per foot or better. This adds up to rougly 4 Mega pixels to capture the entire balloon structure. My thought was to divide this image into four quadrants or eight divisions to get the complete image of the balloon every ho
ur. One image could possibly get us the resolution we desire, but the balloon gondola would be in the way. With multiple cameras that sit away from the gondola fixed, this isn't a problem. I have been looking at various packages to accomplish this task. The set-up I was most taken using Labview was the PCI-1422, two of them, each with two digital I/Os. That way, if a camera with sufficient resolution was selected we could get a greyscale resolution of up to 16 bits instead of being limited to 8 bits of another camera was added to the PCI-1422. I am attaching a document that outlines all of the requirements we are going to have (preliminary). Also, note that to reduce the possibility of a mechanical failure I am going to use fixed, fixed focal length camera for this project. I saw many possible roads to pursue this project with using NI Imaging software and one or more of the approved cameras. I would though, like to get an idea of what an NI engineer feels about this project,
and if it is indeed do-able. I have attached a preliminary set of requirements for the imaging system. These specs will be modified in the near future to narrow down the ambiguity. Thanks.

[BruceAmmons]

I'm not an NI engineer, but I think I can give some good advice.

Taking the image is a minor issue. You will need to figure out how many cameras you want and how much image bit depth you need. If possible, you should use an 8 bit camera to reduce the amount of data you need to transfer. Another option would be to use a 16 bit camera and convert the resulting image to 8 bit while retaining the maximum amount of detail available. If you want a lot of cameras, you might need to use the PCI-1409 and use 4 cameras per board. Make sure you don't run out of slots in your computer.

I would acquire the data and compress it onboard. For compression, it sounds like wavelets would be the best route. You get the best detail for the compression ratio. When sendin
g the image, you will get an overall image, then the detail is filled in. If you don't need more detail, you can stop transmitting the image and move on to the next image. Decompressing the image on the receiving end would be fairly easy.

I must admit I haven't done any calculations regarding how much information you can transmit in an hour using 2400 baud. I don't know how much detail you would be able to transmit in that hour. You would need to make sure this is feasible.

Bruce