Simple Summary
Measuring the live weight of chickens is essential for proper feeding, health monitoring, and production planning on poultry farms. However, manual weighing of chickens can be time-consuming for farmers. This study investigated the use of computer-based methods and digital image analysis to estimate the live weight of broiler chickens without the need for direct weighing. In computer-based methods, the weight of broiler chickens was estimated using simple body measurements, such as the length and width of the back. In digital image analysis, the number of body surface pixels and the day of age were used to predict live weight. The results showed that both computer-based methods and digital image–based models provided highly accurate predictions, whereas traditional linear methods were less effective. These findings demonstrate that modern computer vision and machine learning tools can offer fast, reliable, and non-invasive alternatives for monitoring broiler growth. This approach supports smarter and more sustainable practices in precision poultry farming.
Abstract
Accurate estimation of live weight is essential for efficient management and precision control in poultry production. This study evaluated the potential of supervised machine learning (ML) algorithms and digital image analysis for non-invasive prediction of live weight in Anadolu-T broilers, a locally developed genotype in Türkiye. A total of 4200 records were collected from 100 broilers (50 males and 50 females) over 42 days, including daily measurements of back length, back width, and live weight. Five ML algorithms—Random Forest (RF), k-Nearest Neighbors (KNN), Support Vector Regression (SVR), Extreme Gradient Boosting (XGB), and Multiple Linear Regression (MLR)—were trained and validated to estimate live weight based on morphometric traits. Among all algorithms, KNN achieved the highest accuracy (R2 = 0.982, RMSE = 111.509 g, MAPE = 8.205%), followed by RF and XGB, which also produced stable and reliable predictions. The image-based models using log-transformed regression between body surface pixel area and live weight yielded similar accuracy (R2 = 0.989, RMSE = 101.197 g, MAPE = 7.266%), confirming that projected surface area can effectively represent growth progression. The results demonstrate that integrating ML algorithms with digital imaging offers a practical, cost-effective, and non-invasive approach for real-time broiler weight estimation. This approach supports the advancement of precision poultry farming through automated, data-driven growth monitoring.