Near-infrared-based capsicum counting algorithm using YOLO11

Abstract

This work presents a novel near-infrared-based approach to capsicum counting in greenhouses that uses the advantages of NIR imaging to enhance detection in challenging lighting condi- tions. The proposed algorithm integrates the YOLO11 detection model for capsicum iden-tification and the BoT-SORT multi-object tracker to track detections across a video stream, enabling accurate fruit counting. Trained on a dataset of 611 labeled images captured in a greenhouse, the detection model achieved an F1-score of 0.82, while the tracker obtained a multi-object tracking accuracy (MOTA) of 0.85. The results demonstrate the effectiveness of this NIR-based approach in automating fruit counting in greenhouse environments, offering potential applications in yield estimation.