AI for Sugarcane Stem Node Quality Assessment and Bud Germination Suitability: A Review

Abstract

Sugarcane (Saccharum officinarum L.) is propagated vegetatively through stem cuttings (“setts”) containing nodes that each bear a lateral bud, making stem node quality and bud germination suitability first-order determinants of field establishment and eventual cane yield. Commercial seed-cane selection still relies predominantly on manual visual inspection, a process that is subjective, labour-intensive, difficult to scale, and blind to sub-surface defects. This review synthesizes the current state of sensing technologies (RGB, multispectral, hyperspectral, thermal, X-ray/CT, 3D, and acoustic sensing) and artificial intelligence methods (traditional machine learning, deep learning-based detection and segmentation, transfer learning classification, vision transformers, and multimodal fusion) that could be applied, or have begun to be applied, to sugarcane stem node quality assessment and germination-suitability prediction. Dataset and annotation practices, regional deployment contexts spanning manual and mechanized seed-cane systems, and an integrated sensor-method-application framework are discussed. Persistent challenges include the absence of public sugarcane node/bud datasets, the difficulty of defining an objective ground truth for germination suitability, environmental variability affecting sensor readings in field conditions, and the interpretability of deep learning-based grading decisions. The review concludes by identifying a specific research gapthe lack of an integrated, annotated-dataset-driven, AI-based system for combined node quality assessment and germination-suitability classification that motivates need for research in dataset construction, model development, and systematic performance evaluation.