Anda di halaman 1dari 1

Full Body Capture Tips

Below there are several recommendations regarding full body


capture based on AgiSoft team experience with such data:
Capturing scenario:
Capture system should consist of 5! well synchroni"ed #S$% cameras&
'umber of (blind)"ones( should be minimi"ed since *hotoScan is able to reconstruct
geometry visible from at least two cameras+ Camera placement and zone overlap should be
planned carefully&
,ach photo should effectively use the frame si"e: ob-ect of interest should ta.e up the
maximum area+ /n some cases portrait camera orientation should be used&
0ood lighting is re1uired to achieve better 1uality of the results2 yet blin.s should be
avoided+ /t is recommended to remove sources of light from camera fields of view&
The ob-ect should contrast against the bac.ground& blin.s from the bac.ground on the ob-ect
surface should be avoided+ Starting from +3 version *hotoScan has mas.)by)bac.ground
feature2 that allows automas.ing for fixed cameras2 providing that (empty( shots for every
camera are ta.en without the present of the ob-ect of interest+ The bac.ground should be
.ept unchanged during ob-ect shooting&
#o not try to place full ob-ect in the image frame2 if some parts are missing it is not a
problem whereas these parts appear on other images&
Cameras should not be placed very close to each other2 recommended distance between
cameras is one tenth from the distance from camera to ob-ect+
Object specifics:
4air is difficult for reconstruction2 using well textured hat 5for example2 .nitting6 is
recommended&
7sing cloth with irregular texture pattern that could be distinguishable on photos is highly
recommended: -eans2 .nitting2 etc+ Too dar. as well as too bright cloth are not good+ Avoid
using synthetics2 blin.ing shoes and other cloth with poor texture+
Lens:
Fixed lens are preferred+ /f "oom lens are used 8 focal length should be set either on
maximal or minimal value2 if this condition is not provided each "oom)lens camera should
be calibrated separately&
Try to use minimal possible camera9lens models and different focal length+
Camera settings:
7sing %A: data losslessly converted to the T/FF files is preferred2 since ;*0 compression
induces unwanted noise to the images&
/mages should be ta.en at maximal possible resolution&
<inimal possible /S= value should be used2 otherwise high /S= values will induce
additional noise to images&
Aperture value should be high enough to produce the sufficient focal depth2 it is important
that most of ob-ect details are clearly visible and aren>t blurred&
Shutter speed should not be fast enough to prevent blur caused by possible movements+
Images:
/mages should not be resi"ed2 rotated or anyhow transformed in terms of geometry+
*hotoScan operate with the original images+