Satellite-Guided Herding: Optimizing Pasture Selection for Efficient Livestock Management
摘要
This work presents a Python application to determine the best places to conduct herding by analyzing satellite images. Grazing monitoring provides information to identify, from the available zones, which ones are more suitable for maintaining better livestock feeding, thereby improving the quality of products such as meat or milk. To accomplish this, a vision-based system is trained using images from previous years, with ground-truth data provided by an agricultural application, SIGPAC, where farmers determine land use. Once the model is trained, it is applied to new satellite images to identify pasture areas and subsequently determine the coordinates to decide the path for the herd. Among the available options, those exhibiting the highest NDVI levels can be selected as optimal feeding areas for the livestock. This system can be implemented on a 4-legged robot to assist the shepherd in their daily tasks by herding the flock.