A COMPUTER VISION-BASED ROBOTIC WEED SPRAYER FOR MAIZE FARMLAND PRECISION FARMING

Abstract

Maize, a staple crop of great worldwide significance, often experiences large production losses due to weed competing with its nutrients. It is important to treat these weeds without any harm on the maize crop. Existing approaches to weed management relied on traditional application of the herbicide which is marred by wastages, and could damage the crops also causing health issues to the consumers. In this research, an advanced robotic weed sprayer that uses deep learning and computer vision to solve the ubiquitous problem of weed control in maize farmland is proposed. The research employed an advanced deep learning algorithm that was trained on a large image dataset of common weed species and maize, allowing for accurate weed identification and focused herbicide application. The system’s real-time image analysis guarantees efficient weed control. The system performs exceptionally well, with 75% precision, 80% recall, 77% F1-score and 85.12% mean Average Precision (mAP) in weed recognition. This highlights its potential to completely transform conventional weed control techniques and represents a significant advancement in precision farming technologies as well as a promising option to improve productivity and sustainability in maize cultivation by minimizing crop damage through precise herbicides usage.