View my account

Depth values affected by spotlight

Comments

14 comments

  • MartyX Grover

    Hi @Iolcoz  As a starting point in investigating your case, may I ask whether you have auto-exposure enabled, or if you are instead using a manual exposure value that cannot adapt to changing light conditions.

    0
    Comment actions Permalink
  • Iolcoz

    Thank you for your fast answer. Yes, I have auto-exposure enabled, both for rgb and depth channels.

    0
    Comment actions Permalink
  • MartyX Grover

    Thanks very much for the information. What is the source of the spotlight, please?   Is it natural sunlight or an artificial lamp?

    0
    Comment actions Permalink
  • Iolcoz

    It is an artificial light in the ceiling that illuminates more the surface of object A than the surface of object B. Some areas of the image illuminated by the spotlight seem to be quite saturated, but not enough for considering a bad or burned image.

     

    By switching off this artificial spotlight in the ceiling (the global illumination keeps being enough), this weird effect in the affectation of the depths values of the area that is receiving the spotlight disappears and everything works fine.

    0
    Comment actions Permalink
  • MartyX Grover

    If the spotlight is blotting out the texture of Object A and making it more difficult to read depth detail from, you could try using the camera's projector (its IR Emitter) to project a dot pattern onto the scene.   The camera can use the dots cast over the surface of objects as a texture source to analyze for depth detail in situations where a surface appears to have low texture detail or no texture.

    0
    Comment actions Permalink
  • Iolcoz

    Yes, I already had the IR Emitter enabled. Sorry for not adding this information in the issue description.

    0
    Comment actions Permalink
  • MartyX Grover

    No problem at all.   How near is Object A to the camera, please?

    0
    Comment actions Permalink
  • Iolcoz

    In the normal axis, object A is about 50 cm to the camera and object B is about 53 cm to the camera. In the vertical axis, object A is about 3 cm above object B.

     

    The ceiling spotlight is behind the camera. Due to the spatial distribution, object A is receiving the spotlight in a more direct way than object B.

    0
    Comment actions Permalink
  • MartyX Grover

    Is there any possibility of moving the objects away from the spotlight's location so that object A is not receiving the spotlight directly?

    0
    Comment actions Permalink
  • Iolcoz

    The problem is that in the context of the application where I am applying this, I cannot ensure or demand a specific location for the two objects because is something that I might not control. What I have observed is that when the objects A and B are positioned in a location that doesn't suffer from the direct spotlight, the behavior is coherent.

     

    For me it is important to know if you strongly recommend that a direct spotlight shouldn't point to some of the objects in order to avoid this issue, because in that case I might re-think the boundary conditions of the application.

    0
    Comment actions Permalink
  • MartyX Grover

    In some cases a strong light source can be useful for obtaining depth information.  This includes dark grey / black objects that the camera may have difficulty observing, because dark objects absorb light.  Projecting a strong light source onto such surfaces, such as a black office chair, can bring out depth that would not otherwise be available.

    If a strong light source is projected onto a light colored object though then it can result in a pool of light that blocks out the detail of a surface, making it look like a plain white or black area on the camera image.

    It may be worth performing depth to color alignment so that RGB pixels are mapped to the depth coordinates.

    0
    Comment actions Permalink
  • MartyX Grover

    An easy way to test coordinate measurement with depth to color alignment without any programming is in the RealSense Viewer.   If you go into '3D' point cloud mode in the Viewer, enable depth and RGB and then select the Color option from the Texture control at the top of the Viewer window then you can map RGB to depth.   You can then move the mouse cursor over the image and read coordinates from each point.

    The Measure option also enables you to use the mouse to drag out a line between two points on the image and receive a measurement value.

     

    0
    Comment actions Permalink
  • Iolcoz

    Thank you. Yes, I was already aware of depth to color alignment importance and it is something that I was already applying in all I have reported.

     

    On the other hand, yes, the troublesome surface of object A is quite light and clear.

    0
    Comment actions Permalink
  • MartyX Grover

    If the disturbance can be characterized as glare from light falling onto a reflective surface then the negative effect on the image could be significantly reduced by purchasing a physical optical filter called a linear polarization filter and applying it over the top of the lenses on the outside of the camera.  Section 4.4  When to use polarizers and waveplates of Intel's white paper document about optical filters provides detail about this.

    https://dev.intelrealsense.com/docs/optical-filters-for-intel-realsense-depth-cameras-d400#section-4-the-use-of-optical-filters  

    If you agree that it is glare from a reflection, might use of an external filter on the camera be an option for you?

    0
    Comment actions Permalink

Please sign in to leave a comment.