Computer vision applied to identify behaviors of broiler breeders

Abstract

The utilization of images recorded by video camera in the broilers breeders' behaviors analysis, eliminates the effect of the human presence in poultry's behavior, but it doesn't solve problems related with wrong human interpretation, which are caused by human imprecision or human tiredness when the data analysis demand long time. It's proposed in this research the creation of a new methodology based on frames sequence analysis to differentiate broiler breeder's behaviors which weren't totally distinguished using the methodology based on only frame analysis, made by Miyamoto (2008). The tree decision developed in Miyamoto's (2008) research and the new methodology based on frames sequence analysis, will contribute to the development of a machine vision system, which will be developed in a virtual environment facing instrumentation, which will be able to classify broilers breeders' behaviors and will contribute to the building of computing models to estimate automatically poultry's welfare state.