- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi guys.
I'm using three Intel realsense r200 cameras and have some problems that I can't resolve.
The first one is that the point cloud I get is wavy, I have a feeling that sensor has some
noise dependant on the distance to the working surface. As shown in the examples below you can
see noise as waves. The closer I move to the camera the lower amplitude this noise has.
What is the reason of this noise and how can I get rid of it?
- close look(distance about 60 cm):
https://gyazo.com/0f67a5e3e8a72cb4295faef700ccfbf0
- distance about 1.5 meters (amplitude of the noise is much bigger):
https://gyazo.com/0818084b9f74ce2a6c38907edd2633a2
The second problem is that color images from different realsenses are different. For example,
the color checker from first camera looks different in the second camera. Sensors are placed
one above another. As you already guessed, I tried to use a color checker to compute the color
transform for each camera. But it works too bad. I tried to use different terms for finding
color transform(r g b rr rg rb etc). The best result I get when using 3x3 matrix color
transform with three terms r, g and b respectively. But this transform is still not that good.
When stitching images from different realsenses I get a differently colored halves
of the same object. Can some one give me an advice how to resolve this issue.
To be more precise, when I compute color transform with a man in blue t-shirt it works good.
But I need to recompute the color transform each time I have the man in white T-shirt in front
of the camera. Does it means that color checker approach need to be used each time the new
object appears in the scene? I thought I need to compute it only once under the certain
lighting conditions in the empty scene.
As for the lighting conditions, I have the big LED wall right behind the sensors, that gives
uniform white light that is almost shadowless.
Here on the models texture you can see the difference in color of the clothing's. I did some
blending to reduce the effect. But you still can notice the difference
between sensors(top/bottom of the shirt):
https://sketchfab.com/models/d826bea5a12749eba95f273a5cfb48cd
One last word. I am using librealsense, as it alows to use several r200 cameras at the same time.
I would appreciate any help.
Thanks in advance.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
No one camera of the same model will have completely the same values due to small variances in the 'intrinsic matrix' of the camera during the factory manufacturing process.
Also, I wonder if the LED light is affecting the results of your scans. The R200 has a component called the IR Emitter that assists the IR sensor with light exposure. Under very dark or very bright conditions, having the IR Emitter on can cause the IR sensor to become saturated. You can turn off the IR Emitter in Librealsense using scripting.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
No one camera of the same model will have completely the same values due to small variances in the 'intrinsic matrix' of the camera during the factory manufacturing process.
Also, I wonder if the LED light is affecting the results of your scans. The R200 has a component called the IR Emitter that assists the IR sensor with light exposure. Under very dark or very bright conditions, having the IR Emitter on can cause the IR sensor to become saturated. You can turn off the IR Emitter in Librealsense using scripting.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page