Leaf Segmentation Based on Thresholding Techniques Using Visible, Infrared, and Thermal Cameras

A real-time plant monitoring system using a camera system provides benefits in precision agriculture and other fields. In the system, leaf segmentation is an important task to extract the leaf from the background for advanced processes. This paper presents leaf segmentation based on thresholding techniques, namely Isodata, Otsu, Yen, and Li thresholding, using RGB, Infrared (NoIR), and Thermal cameras. In the experiments, images of the Vetiveria zizanioides plant captured by developed multi-camera systems from outdoor environments were used for evaluation. The evaluation results showed that the highest Vetiveria zizanioides leaf segmentation performance was achieved by the RGB camera using Li thresholding with a Recall of 0.933, Precision of 0.603, and F1 score of 0.742. Furthermore, the results also showed that the RGB camera achieved the best average Vetiveria zizanioides leaf segmentation performance among the other cameras, while Li thresholding achieved the best average leaf segmentation performance among the other thresholding techniques.