In dice recognition, there are several situations that may cause dice image identification problems, including the color of background being the same as the dice, the shape of the dice being round, or the dice touching each other. Especially when the camera zooms in or out, the pip sizes become irregular. We analyze the dice structure features to develop an intelligent memorized least-distance search method (IMLDSM) that can achieve accurate dice detection. Compared to other schemes, our method has two major advantages: (i) Unlike other schemes using complicated classification methods to segment the dice, we adopt a novel, simple, and effective method to analyze dice structure features to achieve more accurate recognition results and (ii) the IMLDSM solves the problems of image zooming in and out, which other methods cannot. After over 250 test images that include different shapes, styles, sizes, and colors used in simulations, this scheme is proven to achieve 100% dice image identification.
