IMAGING AND PATTERN RECOGNITION METHODS FOR FEATHER SEX SEPARATION OF BROILER CHICKS

摘要

Recognition and separation of newly hatched broiler chicks by sex require intensive human labor in hatchery operations. Segregation of chicks by gender and grow-out in gender specific management schemes offer significant benefits to the broiler industry. In this manuscript, methods for automated sex separation using spectral imaging and pattern recognition are presented. A UV optical system was developed to distinguish chick feathers from down of the same color, and to extract the feather sex feature characteristics from baby chick wings. To recognize sexes by analyzing feather patterns, blob analysis and curve fitting methods were developed. The result showed a distinct contrast between feathers and down by the optical system, which produced clear signals for image processing and automated feather classification. The curve fitting method yielded a correct classification rate of 93% for female chicks and 94% for male chicks based on a test of 104 baby chicks. The imaging techniques presented in this article serve as a step forward to the development of an automated baby chick sex recognition system, which will result in enhanced worker welfare, labor savings, and feed conversion and production efficiency.