Overlapping particles edge' cannot be detected with most image analysis software. Thresholding distorts data of particle size when particles are unequally bright. Non-spherical-shaped particles are difficult to be detected as-is. So what can be done with these images? Would I need to trace the images on grain manually and then go through particle size estimation?
EDIT: particle size obtainable from 2D image may be equivalent-area circle's diameter or largest distance between two parallel lines in between which the target particle can be snugly fitted (i.e. maximum width)
Sumit Bhowmick
One question I would ask is 'how do you define or measure 'particle size' for an irregular 3D object (particle) shown as 2D in an image'?
Let's imagine that you could deconvolute 2 overlapping particles - what dimension would you measure or define?
Sumit Bhowmick
I can't see a scale bar on your image.
If this is a representative image of your sample then you need to reduce the concentration dramatically on the slide or plate. 2 other issues are of much more importance:
- Representative sampling. If this image represents millimeter particles of sand then you may need many kg for a representative sample at the x99 part of the size distribution
- Depth of field. Larger particles may be in focus and then it's likely that the smaller ones will not be. The same applies in reverse
The questions you need to answer are attached in the 4Q's.
I have done similar analyses many times, in multiple ways. 1) The static method, as your example image indicates, would require a meticulous separation of each particle in the set to be image (by hand) so that no two particle are touching. Then the application of your image processing software of choice (via camera) so that it recognizes each particle independently. This may require an experiment with differing microscope imaging mode, to see which mode presents the particles to the camera most clearly.
Another approach would be to use a particle size, or size and shape analyzer which can work in one of two ways, 2) either in diffraction mode which use Mie calculations (internal to the instrument) of light scattering patterns, or 3) may present the particles to a camera in a flow condition (in air or within a liquid), from which any number of images can be recorded, processed and analyzed. Some of the latter actually analyze multiple images of each particle as it rotates in free-fall in front of a high-speed camera, creating a 3-D particle analysis. The latter is the most expensive to acquire, the former is the least expensive, but only looks at each particle from one perspective.
Too, one can often submit samples to manufacturers of particle size analyzers, or to independent labs that carry particle size analyzers.
Hope this helps!
I should add that none of the above methods will give the exact same size or size and shape distribution. Each will be slightly different, but will be very reproducible within the same methodology.
- Badges
- Science method