KOMPARATIVNA ANALIZA ALGORITAMA ZA DETEKCIJU KRUŽNIH OBJEKATA: PRIMJENA NA AUTOMATSKO BROJANJE KOVANICA
COMPARATIVE ANALYSIS OF CIRCULAR OBJECT DETECTION ALGORITHMS: APPLICATION TO AUTOMATIC COIN COUNTING

Summary/Abstract: This paper presents a comparative analysis of three approaches to circular object detection in digital image processing, focusing on automatic coin counting. The main contribution is not a globally novel algorithm, but an experimental comparison of three classical methods under the same controlled scenario: a domain-tuned morphological pipeline, the Hough circle transform, and blob detection (SimpleBlobDetector). Evaluation uses a small manually annotated set of six images, a unified framework based on IoU and F1, and a practical Django web application for testing and education. Results show that the morphological approach achieves strong performance (F1=1.0, average IoU=0.77) in controlled conditions with consistent backgrounds, while Hough and Blob show weaker aggregate results (F1=0.382 and 0.404). Performance differences are analyzed in detail, emphasizing HSV background segmentation as a key factor for the morphological pipeline. The implementation supports interactive testing, algorithm comparison, and parameter tuning via a web interface. The work illustrates the applicability of classical methods in controlled industrial settings and provides an educational tool for digital image processing.

• Details
• Contents