Fertilizer Granulation

Aim

Devise an algorithm which will examine the granulation of the fertilizer and detect grains which sizes do not fit in specified range.

Single image of the granulated fertilizer.

Detected and circled correct and incorrect pills:

Performing some operations on the image will help you detect the center of each granule.

Labeling connections is explained in this article.

Add LoadImage filter to load the image from the disk directory.
To easily distinguish the dots use SmoothImage_Gauss to make the image blurred. Change inStdDevX to 2.5.
Add ImageLocalMaxima filter to detect all the centers of the circles-to-be. Change inConsiderPlateaus to true.
Create a circle using CreateCircle filter. To the inPoint connect outLocalMaxima.Point - detected centers of the circles. Set inRadius parameter in filter properties as 8.
The next step is to fit the circles to the original image. You can do it with FitCircleToRidges filter. As inImage use the original image. Connect outCircle positions to the inFittingField.Axis. Change of the following parameters will be required:
- inScanCount - 16
- inScanWidth - 1
- SmoothingStdDev - 0
- RidgeWidth - 2
- MinMagnitude - 3
- inRidgeSelection - First
- inMaxIncompleteness - 0.9
- inFittingMethod - AlgebraicKasa
- inOutlierSuppression - Turkey
- inFittingField.Width - 14
Create a formula using right-click and Add Formula Here option to compare outCircle.Radius. Set minimal value as 6 and maximal value as 10. Add ?? False at the end of it to filter out any Nil occurrences and set them as False by default.
Add DrawCircles_MultiColor. As inImage use the original image. To inCircles connect outCircle from FitCircleToRidges filter. Connect outIsInRange to inColorIds. You can change the marking hue in the properties of the filter. They are green and red by default.

Icon	Name	Description
	DrawCircles_MultiColor	Draws circles on an image with multiple colors.
	SmoothImage_Gauss	Removal of gaussian noise from images.
	FitCircleToRidges	Precise detection of a thin circular line, whose rough location is known beforehand.
	CreateCircle	Creates a circle from an aligned point and radius.
	ImageLocalMaxima	Detection of characteristic points, usually after some image transformations.
	LoadImage	Loads a single image from a file.