06-18-2018 07:53 AM
Hi Bob Schor,
Thank you very much for sharing the information,
I can not do it manually.
In work I have to make a device which can detect the cells as more as possible then makes bounding box onto the detected cells and scan the area of the bounding box using dual axis Galvano mirrors.
06-18-2018 12:07 PM
Well, I hope you are up on the Visual Processing techniques, as it appears (to me) that what you are asking is near the current State of the Art.
While the solution that you appear to require is beyond my own ability, I do have a suggestion that may help you at least identify frames where the Bounding Rectangles are probably incorrect. From the examples that you posted, it appears that most cells are roughly circular with a diameter thabot varies by maybe 30%. This means that all of the Bounding Rectangles should have width and height in the same narrow range. Bounding boxes much larger probably contain multiple cells, as do bounding boxes where the ratio of wide-to-narrow side > 1.5, while very small Bounding Boxes may contain "junk".
Bob Schor