Monitoring, control and regulation are essentials for pharmaceutical manufacturing processes. The proportions and quantities of active substances in pharmaceuticals are subject to serious international rules, since the production of a single defective product may result in death. It’s a huge responsibility for the system that monitors and controls this. A new drug has to go through countless phases until it advances from the research lab to the market. It’s an enormous risk for research development, which is further increased by unsuccessful attempts. After the preparation and inspection of the active ingredients a carrier is added. This will result in a powder mixture. It’s granulated and homogenized by appropriate machines. Before converting them into pharmaceutical forms, it’s possible to take pictures, videos and process them.
The image processing of active substances carries many difficulties and pitfalls. This is a non-trivial task that requires constant compromises. An optimal balance must be found between processing speed and accuracy. For identifying and locating the blobs on the images, I used the OpenCV library. I've also tried several algorithms found in various image segmentation methods. It’s difficult to make polygons out of blobs because they have blurry boundaries. Well-aligned rectangles are required in order to examine the shape of the active ingredients. From these size ratios, we can conclude the deformation of the blobs. If a spot is bigger, longer than it should be, an error message must be sent to the processor of the industrial process. These have been integrated with many other services in my application named PharmaTech, which communicates with the user through a user interface (UI).