Accurate assessment of forage quality is essential for ensuring optimal animal nutrition. Key parameters, such as Leaf Area Index (LAI) and grass coverage, are indicators that provide valuable insights into forage health and productivity. Accurate measurement is essential to ensure that livestock obtain the proper nutrition during various phases of plant growth. This study evaluated machine learning (ML) methods for non-invasive assessment of grassland development using RGB imagery, focusing on ryegrass and Timothy (Lolium perenne L. and Phleum pratense L.). ML models were implemented to segment and quantify coverage of live plants, dead material, and bare soil at three pasture growth stages (leaf development, tillering, and beginning of flowering). Unsupervised and supervised ML models, including a hybrid approach combining Gaussian Mixture Model (GMM) and Nearest Centroid Classifier (NCC), were applied for pixel-wise segmentation and classification. The best results were achieved in the tillering stage, with R2 values from 0.72 to 0.97 for Timothy (α = 0.05). For ryegrass, the RGB-based pixel-wise model performed best, particularly during leaf development, with R2 reaching 0.97. However, all models struggled during the beginning of flowering, particularly with dead grass and bare soil coverage
