A fuzzy rule-based system for predicting the live weight of Holstein cows whose body dimensions were determined by image analysis

Abstract

The aim of this study was to determine the body measurement of Holstein cows through image analysis (IA) and to estimate their live weight (LW) by means of a fuzzy rule-based model using the body measurements. For this purpose, a photography environment was established at a dairy cattle farm where a large number of cows were kept. First, digital photographs of each animal were synchronously taken from different directions with Canon EOS 400D cameras. At the same time, body dimensions, namely wither height (WH), hip height (HH), body length (BL), and hip width (HW), of the cows were manually measured using a laser meter and measuring stick. The LWs of the cows were found with a weighing scale and the data were automatically saved on a computer. In the second stage, the photos were analyzed by IA software developed in the Delphi programming language and body measurements were computed. Manually measured values were very close to IA results. Finally, a fuzzy system was developed by using these body measurements. This fuzzy system was developed by using MATLAB software. Weights that were estimated with the developed knowledge-based system were compared with those found by the platform scale. The correlation coefficient was calculated (r = 0.99). There was a statistically meaningful relationship between the compared data. The developed system can be used confidently, and the system on which the experiments were performed can be modeled successfully.