For a while now I have been playing with structure from motion
Typically till now all of the SfM models I have generated have come from a simple single GoPro (Hero 3 Black or Hero 4 Silver dependent upon when I lost my 1st camera) but I have been starting to move on to multi camera systems to try and get more images and or better coverage of wreck sites. Due to this I have been trialling a couple of different camera systems looking more from cost effectiveness point of view if I want say 4or 6 or 8 cameras
In short my needs are pretty simple the best image quality I can get (subject to realistic $’s) camera suitable for taking a Time Lapse Photos @ 0.5, 1 seconds Intervals to generate the SfM models, a wide Field of Vision to get good coverage of the site suitable to at least 60m of water (happy to push the 45m units to the greater depth) and takes good images under low light conditions
This weekend just passed I used a camera rig with 2 cameras spaced 1.2m apart with a GoPro Hero 4 Silver (16MP FOV 122˚ vertical say $US400) and a Xiaomi Yi sports camera (16MP similar Vert FOV say $US90 with housing). I had wanted to do some comparisons stuff but we had pretty good vis on the wreck so I just swam the rear half of the wreck.
In short I am disappointed with the Xiaomi performance but not writing it off quite yet. Basically it has near same performance as the GoPro specs (16MP stills @ 0.5 images/second etc etc) and on the surface all of the images I have done with it look fine and no different to the GoPro.
But during the dive I was capturing a still every second but in the lower light all of the images were pretty much black whilst the GoPro images ran fine (see side by side performance bellow) The Xiaomi Yi sports camera was (I think) using stock standard settings so I may have to tweak a low light setting before I write it off.
I know I can add lights (and most likely will) and it will help things (and bring out more colours) but I was hoping that they may not be too needed in short term till I have settled on camera configurations (and number of cameras) given say a requirement of say 80 Lux (equivalent amount of light in a darker warehouse) and the FOV of the GoPro https://en.wikipedia.org/wiki/Lux per camera and assuming the FOV of the OpenROV External Light Cube is equivalent to the GoPro’s FOV (XX Brian do you know the FOV for the light cube?)) and knowing I typically produce SfM say 2 to 4m off the bottom
Anyway the semi worked results from the GoPro are bellow (I still have to clean up a few mesh points to get a better image -there are WAY TOO many fish causing issues with false heights) but here is the 1st cut out of the software as it stands
PS the wreck is the SS Tuggerah in 46m you can see more details of the vessel and site here
In short anyone got any good options given $'s and low light