With Chinese poultry production scale and quantity expanding, intelligent breeding is the current hot research direction of the regulation of poultry. Computer vision technology can be used to provide a noninvasive, non-invasive, low-cost and highly effective way to identify animals behaviors to detect the activity level, diagnose the diseases of chickens and find the dead ones. The visual systems of chicken detection and behavior recognition were summarized, and the correlativity between the phenotypic parameters, behavior parameters and reproductivity respectively, phenotypic feature extraction and recognition and behavior recognition algorithm were reviewed; the problems in visual system were analyzed, the optimization strategy was put forward; the feasibility of using computer vision technique to select high quality breeder rooster was discussed, and the rooster selection algorithm framework was preliminarily proposed. Finally, the applied prospects and the optimization directions of computer vision technology in the poultry industry were expected.