Precision Agriculture: Computer Vision-Enabled Sugarcane Plant Counting in the Tillering Phase

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Sugarcane is a globally significant crop, vital for sugar, ethanol, and other industrial products.
• Accurate pre-harvest yield estimation is crucial for sugarcane industry planning and farmer agreements.
• Current methods lack precision in estimating plant numbers during the critical tillering phase.

Purpose of the Study:

• To develop and present a novel methodology for estimating sugarcane plant numbers in the tillering phase.
• To introduce a new, publicly available dataset of sugarcane fields for research purposes.
• To improve pre-harvest planning for the sugarcane industry.

Main Methods:

• A modified Faster R-CNN architecture was employed for plant detection and classification.
• Feature extraction was enhanced using VGG-16 with Inception-v3 modules.
• A sigmoid threshold function was integrated for improved detection accuracy.

Main Results:

• The proposed methodology achieved a promising accuracy of 82.10% in detecting and classifying sugarcane plants.
• The developed dataset, captured during the fall season, supports further research in sugarcane crop monitoring.
• The model demonstrated viability for real-world application in sugarcane field assessment.

Conclusions:

• The research presents a viable AI-driven methodology for accurate sugarcane plant estimation.
• The new dataset and model contribute to advancing precision agriculture in sugarcane production.
• The findings support better production and purchase planning for the global sugarcane industry.