Automated sugarcane stem node extraction can be extremely useful to promote single-stem node plantlets and tissue culture techniques. It can be utilized to reduce labor, plantations, and logistics costs while providing a high yield in sugarcane cultivation. Real-time stem node counting and monitoring can be utilized for controlling the production of billets and their quality. This study not only presents an approach for real-time sugarcane stem node detection and localization but also counting and tracking for production monitoring. Smaller and lighter scaled down light weight YOLOv8 model is applied to optimize the deep learning model. A dataset of 2870 images is collected in natural sugarcane field complex environments, with each image annotated in two classes in the YOLOv8 format. The model achieved a high mean average precision of 0.974 and a low detection time 7.17 milli seconds. The model achieved 94.08% overall accuracy. Additionally, automated sugarcane stem node counting and monitoring offers a scalable solution for real-time sugarcane stem node extraction. The trained model demonstrates exceptional performance in real-time scenarios, facilitating precise stem node detection, localization, counting, and monitoring.