This message was posted on behalf of Intel Corporation
Thank you for your interest in the Intel RealSense SR300 camera.
Could you please give us more details about your project? You mentioned that the depth values are inconsistent. What is the scene that you are streaming?
Thank you and best regards,
First of all, I would like to mention that I work on behalf of a company and we have a big stock of SR300. I performed tests with at least 9 different SR300.
For the test purposes I used a toy (stuffed animal, no reflective surfaces - so I could get more accurate results) placed on the distance about 60-70 cm away from the camera. Everything behind the toy is too far for the SR300 to see ("undefined" depth).
But I think I found the problem:
I tested the same code with the D435 camera firmware 05.10.13.00 (of course, with different preset), and solution works perfectly.
I only have problem when using SR300 camera.
When comparing the results, I noticed:
- When using SR300, distance (z) for the same point in space are "jumpy": 3-5 frames with distance 0.7m (correct), than next 1-2 frames with distance 0.2m (static scene, nothing moves), than again 0.7m (correct).
- When using SR300, after mapping depth to color there is a shift by x-axis (spatial alignment is perfect when I use the same code with D435 camera).
- When using SR300, during call to PointCloud.Calculate, color sensor stream (Bgr8 & Bgra8) often freezes (works fine with D435).
I read point 1 again. It sounds as though you are describing that the detected distance is fluctuating between 0.7 and 0.2 m and then back to 0.7 m, even though the stuffed toy that is being scanned is not moving?
In another case where the depth was fluctuating with an SR300, it was suggested that temperature may be a factor.
- yes, scene in front of the camera is static.
- camera temperature or environment temperature?
we made excessive Intel RealSense SR300 temperature measurements, obtained results never exceeded acceptable working range described in documentation.
we were able to heat camera up to ~47ºC, and from that point camera's temperature is pretty steady, as you can see on the chart below.
- on the other hand, in the usual scenario camera streams only for up to 30s, with intervals of minimum 30s
Both environmental temperature and operating temperature will be factors. The operating temperature inside the casing is likely to be the more important of the two. The camera can stand up to 35 degrees max operating temperature officially. I would expect temperature to rise as the camera uses higher settings.
Electrical equipment can also be affected by the height of your location above sea level though. The higher the altitude, the thinner the air is and the harder it is to ventilate a device effectively. An example is PC fans, which may be louder at high altitude because they have to work harder to keep the PC's internal temperature down.
The SR300 and 400 Series are different camera technologies. SR300 uses structured / coded light projection, and the 400 Series uses Stereo projection involving a pair of left and right imagers. So the different sensor hardware arrangements could be a factor in differences between the point clouds of SR300 and the 400 Series when aligning between depth and color.
I cannot think of further suggestions, unfortunately. It may be best to see what Intel support member's Eliza's next response is, or post your problem on the RealSense GitHub and seek the perspective of the RealSense engineers and developers who reside there.