Advice on Using Stereo Vision for Volume Measurement

[ccp1999]

Hello LabVIEW community, I'm planning a project where I aim to measure the volume of an object using stereo vision and 3D processing techniques.

Here is my proposed approach:

1. Stereo Vision: I'll use two cameras to setup my stereo vision system and capture images of an object. This will allow me to compute a disparity image and extract depth information.
2. Point Cloud Processing: Using the depth information, I will generate a point cloud representing the 3D space around the object. I plan to filter the point cloud to focus on the object of interest.
3. Mesh Generation: Once I have the filtered point cloud, I intend to create a 3D mesh (surface) of the object using appropriate algorithms.
4. Volume Calculation: My ultimate goal is to measure the volume of the object from the 3D mesh.

I'm reaching out for advice on a few specific points:

1. Point Cloud Filtering: What are the best methods for efficiently filtering a point cloud to isolate the object in LabVIEW?
2. 3D Mesh Creation: Which LabVIEW-compatible methods or libraries would you recommend for generating a 3D mesh from the point cloud? I discovered an add-on called '3D Vision Basic Tools' but it doesn't have a straightforward VI to do this.
3. Volume Measurement: Do you have any tips or resources for accurately calculating volume from a 3D mesh in LabVIEW?

Any guidance, advice, or examples on how to approach this process in LabVIEW would be greatly appreciated. Thank you!

[Quiztus2]

Do you have a realistic picture of how bad depth information from real world objects look like, when retrieved by passive stereo vision? Even active stereo vision can be bad, e.g. Intel RealSense.

Please share more details about your setup and the object you try measure.

[Andrey_Dmitriev · Accepted solution]

---

@ccp1999 wrote:

Hello LabVIEW community, I'm planning a project where I aim to measure the volume of an object using stereo vision and 3D processing techniques.
...

---

As very first step to start is Using the Stereo Vision Example. If this technically will work with your images, then half problem is solved. If not, then you should find "third-party" library which will compute acceptable disparity map. Writing such algorithm "from the sratch" may be complicated, based ony my experience from the past.

Once you will get disparity map, then you can continue with OpenCV reprojectImageTo3D. (such functionality is missing in VDM, AFAIK), and after that if point cloud is OK, then volume calculation is relative simple calculation.