CATFISH FEATURE IDENTIFICATION VIA COMPUTER VISION

Computer vision algorithms for automated channel catfish (Ictalurus pu nctatus) processing were developed to: (1) detect the orientation of a catfish; (2) identify the head, tail, pectoral, ventral, and dorsal f ins of a catfish; and (3) determine cutting lines for deheading, detai ling, and definning (dorsal and ventral fins). The algorithms are inva riant to translation, rotation, and scaling of a catfish and are robus t to noise. They may be applied to most fin fish processing, and are n ot limited to catfish. Canny edge detection and a labeling and trackin g algorithm were applied to locate the boundary of a catfish. A two-st age, model-based, catfish segmentation algorithm was proposed to locat e each part of a catfish. A dominant point detection scheme was propos ed and applied to find the points that connect each part of a catfish. Then morphological knowledge of the catfish was used to locate the fe ature points of each part of the catfish and to determine the cutting lines. The angle of the major axis and center of mass were used to rep resent the orientation of a catfish.