Informatika Tanszék

Vígh

Balázs

Preliminary Study on Mold Colony Segmentation Using K-means Clustering

CINTI IEEE 25th International Symposium on Computational Intelligence and Informatics

IEEE Hungary Section; Óbuda University; Hungarian Fuzzy Association; John von Neumann Computer Society

Budapest

Magyarország

2025.11.18.

2025.11.20.

Manual evaluation of microbiological samples has some disadvantages, such as being time-consuming and giving results that can depend on the person doing the analysis. To address these issues, digital image processing and computer vision methods are becoming more common. The goal of this pilot study was to create an automated system for analyzing mold colonies using image processing and clustering algorithms. The system can segment and count colonies reliably under different lighting conditions, using color-based segmentation in the RGB color space. During image preprocessing, the contours of the Petri dishes are automatically detected using the Hough transform method. For validation, 10 images were chosen to represent different stages of mold growth and lighting conditions. The pilot study showed good accuracy for colonies that were clearly separated. However, it was sensitive to background noise and had difficulty analyzing colonies that were close to each other.