Experimental Model for Determining Developmental Stage of Chicken Embryo Using Infrared Images and Artificial Neural Networks
Conference Paper
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
Development of a chicken embryo is conventionally assumed to follow a set growth pattern over the course of 21 days. However, despite identical incubation settings, many factors may contribute to an egg developing at a different rate from those around it. Being able to determine an embryo's actual development instead of relying on chronological assumptions of normal growth should prove to be a useful tool in the poultry industry for responding early to abnormal development and improving hatch rates. Previous studies have used infrared imaging to enhance candling observation, but relatively little has been done to implement infrared imaging in problem-solving. The purpose of this research is to construct a quantitative model for predicting the development stage and early viability of a chicken embryo during incubation. It may be noted that a similar project was conducted previously using different input parameters. This study seeks to improve upon the results from the earlier project. In this project, infrared images of eggs were processed to calculate air cell volumes and cooling rates, and daily measurements of egg weight and ambient temperature were compiled. Artificial neural networks (ANNs) were "trained" using multiple input parameters to recognize patterns in the data. Various training functions and topologies were evaluated in order to optimize prediction rates and consistency. The prediction rates obtained for the ANNs were around 81% for development stage and around 92% for viability. It is recommended for future research to expand the potential combinations of input parameters used in order to increase this model's versatility in the field. 2013 SPIE.
