We are using an optic 3D scanner to have a 3D view of soil surface and then using GeoMagic software to have some quantitative results.
The problem is to find volume of some rills on the soil surface or the volume of soil decreased between two 3D images.
Is there any function in GeoMagic or other 3D softwares to differentiate the volumes between two 3D images?
If you know the pixel/voxel size in real world coordinates (or the size of the field of view - area being shown by the image), the simplest way to measure the volume of a part of the image is to count the voxels and then multiply the number of voxels by the volume of one voxel. No complex function needed. If you have tomographic images, individual voxel volume would be pixelSize X pixelSize X sliceThickness.
If you don't know how to obtain only the voxels that are different between volumes, you can simply subtract one from another (if they represent the same thing in different moments of time).
to add to the above, before you do any manipulation of your images, you will need to first segment the image to extract the soil rills. Simply subtracting the raw images will probably not give you anything of value I would imagine. Basically, you would want to take a bunch of raw images, and manually identify which voxels are rills (giving them a label of 0) and soil (giving them a label of 1) to give you your "correct" outputs. There are a number of tools to do this; I know of several people using ITK snap (http://www.itksnap.org/pmwiki/pmwiki.php) and they seem to like it.
If you are concerned with applying this procedure to tons and tons of new data, you may be interested in automating the task (though for a single field study, this would not be necessary) with a segmentation model. You would feed your model the raw input data and the segmented outputs in a supervised context. For development of a model, you could use something like tensorflow if you really wanted (deep learning toolbox), or could use other non-deep methods that would probably perform well if your sample size is small. The best models for segmentation tasks when you have loads of data are typically convolutional neural networks (CNNs). Once you have trained your segmentation model, you can then simply apply it to new data to segment the rills from the soil (ie, given your initial image, you would then get an image that is is identical dimensions and takes a 0 for rills and a 1 for soil).
Once you have segmented outputs, simply sum over the entire image to obtain the volume of soil in voxelspace. Note that this also translates very easily to computing the volume of the rills in the image; you can simply take (total number of voxels) - (sum(segmented image)) and will be left with the non-soil volume. You can obtain the true volume by just multiplying volume in voxel space * (volume per voxel) which will be a function of the camera resolution.
You could try our GigaMesh Software Framework: http://gigamesh.eu
After loading your PLY/OBJ choose "Info -> Mesh (Volume) dx, dy dz", which computes the sum of all prisms defined by the triangles of the mesh using the three directions of the coordinate system. The base of the prism is the triangle projected to the e.g. xy-plane for dz. This is inspired by the Divergence theorem by Gauss and can handle concavities like your cracks.
If your mesh is not aligned you can also use an arbitrary plane as base for the integrated prisms. The so-called Mesh plane is shown and can be manipulated, when you hold the "shift" key. Within the "Select" menu you find about six different methods to define your plane (three points, view direction, HNF, etc.).
dears
Thank you all.
To be honest, I am beginner in 3D image analysis.
I will provide an snapshot with some more explanations about my case.
As can be seen in attached photo, we used Shining 3DScanner (Optic) to prepare 4 shots from various aspects and then to merge them to have 3D scan of soil surface. Then we used GeoMagic software to have some quantitative results. The problem is to find for example, volume of roughness on the soil surface or the volume of soil decreased between two 3D images as a result of rain splash and etc. Is there any function in GeoMagic or other 3D softwares to differentiate the volumes between two 3D images?
It seems impossible to me, to calculate the volume of the peaks and throughs from renderings of those surfaces. You need the measured data from your 3D scanner, hopefully already a triangle mesh, and go from there.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing