Estimation of Live Weight from 2D Digital Images of Simmental Cattle Grazing in Kars Province

Authors : Serkan Özkaya, Yalçın Bozkurt, Gökhan Tüzün

Pages : 84-91

Doi:10.51970/jasp.1748591

View : 63 | Download : 118

Publication Date : 2025-12-30

Article Type : Research Paper

Abstract :In this study, we aimed to develop regression models to estimate the live weight (LW) of Simmental cattle using image-based body area (BA) and back view area (BVA) measurements. A total of 42 healthy Simmental cattle were included in the study, and a total of 336 measurements of CG, HW, BA and BVA were taken during the whole grazing season. The collected data were subjected to linear and polynomial analyses, and the predictive power of different variables for estimating LW was compared. According to the results, the highest R2 value for LW prediction was provided by CG (R2=0.93). For BA, the linear model yielded an R2=0.90, while the cubic model improved the prediction accuracy to R2=0.92. The model using HW had relatively low predictive power (R2=0.46), whereas BVA showed a moderate relationship (R2=0.60). In multiple regression analysis, combining CG and HW increased the predictive power to R2=0.94. Similarly, combining BA and BVA resulted in an R2=0.93. A high correlation was found between LW and both CG and BA (0.97 and 0.95, respectively). These findings confirm that CG remains the most reliable traditional measurement for predicting LW. However, digital BA measurements obtained through DIA can also serve as a contact-free and practical alternative under field conditions. In particular, cubic polynomial models improve the accuracy of predictions derived from DIA. It was determined that study could be implemented under field conditions without causing any adverse effects.
Keywords : Canlı ağırlık, Digital görüntü, Vücut alanı, Arka görüntü alanı, Tahmin

ORIGINAL ARTICLE URL