Gloss and roughness inspection of tablets:
Definition of problem:
Process/operation: tablet coating
Product: coated tablet
Defect types:
- uneven corners on square tablets,
- low or irregular surface gloss (microscopic roughness)
- large and small divots on top or bottom surfaces,
- uneven sides (not straight) on square tablets,
- overall outer dimensions out of spec
Consequences: Unevenly coated products affect product aesthetics and packaging efficiency.
Solution:
General overview:
A sample quantity (20 pcs.) of product is presented to two cameras with different lighting techniques. One camera handles macroscopic roughness and one camera handles dimensions and gloss. A Visual Basic front end stores images and data for each sample onto a PC. The system archives data and images on each sample based on its numbered position on the sample plate.
Fixturing & Lighting:
The system is built in the form-factor of a light-controlled booth with two stations. The samples are conveyed from the load-point to each of the inspection stations.
At station one, the samples are suspended above a back light where, using prisms, the camera is able to acquire two simultaneous views of the sample – one from above and one from the side and measure its length, width and thickness. Gloss measurement is also done at this station.
At station two, the samples are conveyed under a second camera where they are lit by intense low-incident lighting. The low-incident light causes shadows whose size and darkness are directly proportional to the surface roughness.
Vision Hardware:
Two Cognex Insight 5401 cameras with 1024x768 resolution. Two proximity sensors trigger the cameras. Backlighting and low-incident lighting are used for station one and two, respectively.
Programming:
Station one camera is programmed to measure the width, length and height of the sample using FindSegment tools that are fixtured in X, Y and O to the centroid and long axis of the sample
The uniformity of the four corners is measured by determining the shape of each and reporting the percentage difference from the smallest one to the mean value of the three remaining larger ones.
The straightness of the sides is measured by selecting pixel points along each side of the sample and running a linear regression routine to fit a straight line to the points.
Gloss is measured by projecting a specularity onto the surface of the sample and measuring the surface dispersion with respect to a calibrated gloss-standard that is also in the field of view.
Station two camera is programmed to measure the macro-roughness of the sample, i.e., how bumpy is the sample and also the graininess of the surface of the sample. Samples can be roughly oriented in the field of view because the FindPattern function fixtures the measurement regions onto the part wherever it’s found in the scene. The CompareImage function allows for image subtraction of any number of images that have been filtered at varying thresholds. The result of the image subtraction is a “relief map” of the sample that can be used to infer depth and quantity of bumps.
Results for the customer:
The system is much faster than the human method of manual comparison.
The system measures more parameters than was previously practical.
The system greatly reduces the subjectivity of grading the samples and provides the data needed for a quantifiable decision on process corrections.
The system is an archive for data and images.
