Coin recognition based on geometric features

In this study, it is aimed to count the total amount in an image of coins using machine vision methods. The extrinsic parameters of the camera whose intrinsic parameters are already known are estimated relative to the plane on which the coins lie. Textural details are removed from the image using bilateral filtering and circles are detected. Reciprocal points from the detected circles are back projected using the estimated pose. By calculating the metric diameters of the detected coins, the coins are recognized and counted.