Hello everyone,

I searched the forum and the Internet for quite a bit...but I can't find a solution to this problem.

A single Kinect 2 is supposed to deliver better results compared to a single Kinect 1, correct ?

In my tests, I have huge amounts of yellow (no depth measured) in the frame, whenever I use Kinect 2. The yellow (no depth) is on the edges of the frame, on the floor....it looks almost as if the Kinect 2 infrared lights are way too weak to light the scene. Of course I tried everything: repositioning, different room, different lighting conditions, different computer....nothing, Kinect 2 gives much worser results than Kinect 1 for me.

I'm well aware of the USB 3.0 issues a lot of people have with the Kinect 2. But from my research and experience...bad USB 3.0 or graphics card configuration usually result in frame drops. At least I couldn't find anyone with yellow (no depth) issues because of this.

It could be, that I simply have a defective Kinect 2 sensor. Please, can anyone confirm that it is possible to get an image with floor, yellow free (i.e. correct depths for most of the pixels) with the Kinect 2 ?

It's not just that the capturing looks bad, also when processing the data (tracking with Ipi Studio) the Kinect 1 gives better Results.

Thanks for helping me out here. Anything is appreciated.

Hi
Please check out our sample videos
http://docs.ipisoft.com/Sample_Video_and_Project_(single_depth_sensor)
http://docs.ipisoft.com/Sample_Video_and_Project_(two_depth_sensors)
Do you have similar amount of yellow in your videos?

Hello,

thank you very much for the quick response. I will download the sample video and see if automatic processing will get through without issues.

For clarification, of what I believe might be an issue with my Kinect 2 I attach two images taken at the same location.

As you can clearly see, the kinect_1.png shows almost no yellow, whereas the kinect_2.png has a lot of yellow in critical areas (like the floor).

Kinect 1 (old)


Kinect 2 (new)


Do you believe this can be normal with the kinect 2 ? Or do you think it is malfunctioning. When I looked at the how to record video from ipi it was explained that yellow means no depth information and should be avoided....only, I have absolutelly no sucess avoiding this with the Kinect 2. The Kinect 1 shows no yellow in the exact same location.

Tracking wise I already switched back to kinect 2, as the results, even, with a might be defective sensor, still seem to be a little bit better.

I would like to figure if it would be worth it to go and get a new kinect 2, to solve this or if it would be a waste of money.

Thank you !

Now I, tested the single_kinect2_walking.iPiVideo, from your link. As expected...tracking went through flawlessly. I'm really not sure if my Kinect 2 would have produced the same quality. I could see, that on the edges of the frame, the sample video also has yellow...the main difference is really the floor. With my Kinect 2 I get also a lot of yellow in the middle of the image on the floor. My floor is not reflective (matte wood chip wood panels). I also tried to get rid of the yellow on the floor by using some cotton sheet, unfortunatelly this didn't made much of a difference. Is the kinect 2 much more sensible to even the faintest glimpse of day light ? There are some windows in the room, but I also tried to rule this out, by trying to track at night...unfortunatelly also this made no difference.

Thanks a lot for having a look.

Kinect v2 uses different depth sensing technology than Kinect v1, although both are based on infrared light. That's why they differ in terms of how the same environment affects yellow noise in the depth picture. Indeed, from our experience Kinect v2 picture may look more yellow under good conditions than Kinect v1. However, at the same time it is always more sharp and has less noise in measured depth data itself - so accuracy of depth is better after all.
Actually yellow noise on your picture for Kinect v2 does not look that bad to me. The noise is located mostly in the corners (usual for Kinect v2) and on the floor at far distance from the sensor. Near and middle floor surface is measured quite well and should be sufficient for proper detection of the floor plane. Other than that you should only worry about the noise on an actor. We noticed that some fabrics are poorly visible to Kinect v2 and produce much yellow noise. If actor's depth cloud is noiseless then you're good for the tracking.

Thanks for the clarification. So now I'm confident that my Kinect v2 is working.
I hope this thread might help others if they see their Kinect v2 producing more yellow than v1.

By the way, I said I tried to improove the situation by laying cotton sheets on the floor. This didn't do much good. Now I also tried hanging the sheets in Front of the wall and pulling them with weights towards the camera to form some kind of concave profile, instead of the cubic shape from the room. This improoves the situation in the middle of the frame. Still, it was unecpected, that the v2 needs this whereas the v1 shows no yellow just without any modifications. But as you said, for the tracking result, yellow on the environment might not even be that critical for the v2 as long as the actor is not yellow.

Thanks again for the great support !