Regresar

Prediction Model for Chicken Egg Fertility Using Artificial Neural Network

Abstract:

Prediction of fertility status of chicken eggs is an important process in the hatchery industry. Currently, the Philippine poultry sector relies on the traditional manual candling procedure which is subjective and laborious. For these reasons, an attempt to improve to the candling procedure, reduce loss and establish a standard platform that classify the fertility status of chicken eggs was conceptualized. The primary objective of this endeavor is to create a predictive model for early detection of the fertility status of chicken eggs. An experimental the imaging system set up was constructed to captures the image of a five (5) day old chicken egg without damaging the eggshell. A total of one hundred fifty (150) images were transferred to a computer. Images were preprocessed and undergone the color segmentation process in order to extract the color space parameters. One hundred (100) out of the one hundred fifty (150) images were fed directly into the classification algorithm. Matlab R2018a neural network toolbox was used, specifically the pattern recognition to train the dataset. Results of the accuracy of the system was shown in the confusion matrices, training is 98.6% accurate while validation accuracy is 93.3 % and testing has an accuracy of 93.3%. The predictive model has an over-all accuracy of 97%. The remaining fifty (50) chicken eggs were used in the final testing. Result of the comparison also revealed that the predictive model has a lower mistake ratio compared to the prediction made through manual candling process.