ANIMAL BODY ANATOMICAL DIMENSION CALCULATION SYSTEM USING COMPUTER VISION

• Main Author: Rizqi, Raditya
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/77382
• Institution: Institut Teknologi Bandung

As the human population increases, the need for food becomes crucial. The consumption of livestock meat is expected to rise by 14% in 2030, prompting the livestock industry to be more proactive in boosting production. Anthropometric measurements of livestock become a crucial step in maintaining the quality and quantity of production, monitoring animal health and growth, and supporting economic aspects. However, traditional measurement methods using contact tools have limitations and can induce stress in animals. In this research, the proposed solution is the utilization of computer vision technology. A computer vision-based livestock anthropometric measurement system is designed to enhance the efficiency of measuring livestock dimensions. This involves an object segmentation model using Mask R-CNN to predict the anatomical dimensions of the livestock. The IoU evaluation method is used to measure the segmentation quality, resulting in bounding box AP and segmentation values of 77.9% and 73.8% respectively. In anthropometric measurements using computer vision technology, the RMSE value for body length measurement is 6.8 cm with a background. For chest girth, the RMSE value is 11.7 cm with an elliptical approach and background, and 14.8 cm with a circular approach without a background. Abdominal girth has an RMSE value of 7.5 cm with an elliptical approach and background, and 12.9 cm with a circular approach without a background.