LabVIEW
Pattern Matching
In order to get help, you'll either need to be wayyyyy more descriptive than "the particle that is in the images" or we'll need to see an example image.
Analyzing 1000 or 12000 images is exactly the same as analyzing one image, it just takes longer. 😄
Do you have the code to analyze one image?
- Do you have any time restrictions?
- Is there always exactly one particle of known anisotropic shape and size?
- Is the size constant? (If so, you could just do a 2D convolution with a black disk of correct size, similar to this where we look for a circle).
- Is the particle always in view?
- How much noise is there?
- How many pixels in each image? (is 146x146 typical)
- Can we assume that the particle moves slowly i.e. the positions in adjacent frames is similar?
- what resolution is required? (rounded to the nearest pixel? Subpixel resolution?)
- Once you express the xy positions as complex values, getting the direction and speed between two adjacent frames is trivial.
Well, you have a nice image, so that makes life a lot easier.
Which one is the particle?
A simple IMAQ Threshold and IMAQ object analysis or circle detection would detect both particles. A simple filter on the image or size on the results would give you the small or large particle.
You could use OpenCV with Python (called from LabVIEW) for this too.
Search LabVIEW like a graph! 
