PENGEMBANGAN MODEL ESTIMASI BERAT AYAM BROILER MENGGUNAKAN PENDEKATAN PEMBELAJARAN MESIN BERBASIS CITRA

• Main Author: Electra, Jordan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/67195
• Institution: Institut Teknologi Bandung

Farm X, which operates in the field of broiler chickens, has low cage productivity, as measured by the food conversion ratio (FCR) metric which is 1,58. The ideal FCR value is in the range of 0.8 – 1.5 with a lower value indicating higher productivity. Farm X seeks to increase the productivity of the cage by allocating the amount of feed given according to the weight growth of broilers every day. The weight growth of broiler chickens in farm X is still measured manually so it requires a lot of time and labor and makes the chickens more easily stressed. This research presents a method to estimate broiler weightautomatically using computer vision approach and machine learning model. Thedata used in this research were obtained from video broilers cage and manual weight calculation on daily basis for 33 days. The broiler images is processed using several image processing method such as noise removal, adaptive thresholding, morphological operations, and image segmentation to generate morphological data of each broilers. The model development is carried out by utilizing morphological data of each broilers which increased along with the increase in weight. Five features of broilers were used, namely area, perimeter, mean radius, maximum radius, and major axis. These five features were used tobuild machine learning models using Multiple Linear Regression, Support Vector Regression, and Artificial Neural Network algorithm. The proposed model then was validated using a 10-fold cross validation method. The proposedmodel performance is calculated using the RMSE metrics, which shows SVR model has the best performance with RMSE value of 144,7 gram.