Coin Counting & Classification System
Automated Coin Detection, Marker-Controlled Watershed Segmentation & Valuation
Executive Overview
NumisVision is an advanced classical computer vision pipeline developed to detect, segment, classify, and calculate the total monetary value of overlapping and adjacent coins on unconstrained surfaces. The system combines morphological filtering, distance transform algorithms, and watershed segmentation to resolve touching coin boundaries, and evaluates performance using rigorous mathematical metrics.
The Challenge & Solution
The Core Problem: Standard circle detection algorithms (such as naive Hough Circle Transform) fail catastrophically when coins overlap, touch, or exist under uneven ambient lighting with harsh specular glares on metallic edges.
The Engineering Solution: Formulated a robust multi-stage segmentation pipeline using Gaussian blurring, comparative Otsu vs adaptive thresholding, Euclidean distance transforms to identify local maxima centers, and marker-controlled watershed segmentation to partition merged contours with high geometric precision.
System Architecture & Data Flow
Key Engineering Highlights
Robust Overlap Separation
Marker-controlled watershed successfully splits touching coins where Hough circles fail.
Mathematical Evaluation
Evaluated across variable lighting datasets with MAE, RMSE, and overall count accuracy metrics.
Calibrated Currency Summation
Converts detected coin radii into actual currency denominations (piastres / pounds).
Comparative Analysis Notebook
Interactive Jupyter Notebook documenting step-by-step image transformation stages.
