This paper documents a light-hearted machine vision system designed to recognize a small, fuzzy object, modelled after ``Doc'' in Snow White and the Seven Dwarves. As seen in Figure 1, some of Doc's primary features are the cartoon-like colors of the fabric from which he is constructed. It is hoped, that by recognizing these colors, we can recognize Doc in his entirety.
Below, we follow the development of the Doc Recognition System (DRS) in chronological order. In the final analysis, only one algorithm will be discarded, resulting in a system depicted by the block diagram in Figure 2. Doc is reduced to five defining components: his hat, his shoes, his skin, his suit, and his beard. Each component is separated from the test scene clutter by extracting its colors, modelled as a cylinder section in hue-saturation-value (HSV) space. The result is a set of monochromatic blobs which correspond to each defined component of Doc; these can be roughly bound by a contour extraction and bounding box algorithm.
