SUSTAINABLE WEED MANAGEMENT IN RICE ECOSYSTEMS USING AI-BASED DRONE SURVEILLANCE AND TARGETED HERBICIDE APPLICATION

Abstract

Weed infestation remains a major constraint in rice cultivation, leading to significant yield losses and excessive herbicide use. This study explores the integration of artificial intelligence (AI) and drone technology for sustainable weed management through targeted herbicide application in rice ecosystems. High-resolution multispectral imagery captured via drones was analyzed using a custom convolutional neural network (CNN), achieving a weed detection accuracy of 94%, surpassing alternative models like YOLOv5 and ResNet50. AI-generated weed maps were employed for site-specific herbicide spraying, resulting in a 59.2% reduction in herbicide usage compared to conventional blanket application. Under AI direction, tests in the field revealed that weed density fell by 85.3% and rice grew by 14.6% more compared to other groups. According to environmental studies, there was only a small amount (6.8%) of unintentional damage to vegetation, along with average soil and water contamination levels of 0.46 ppm which helps protect the environment. Researchers noted that economic analysis indicated a $25 per hectare benefit because both input prices and productivity improved. Moreover, the metrics from model training showed that the AI system became more stable; this indicates that the system is well-built. Despite facing costs and data problems, the study clearly shows that AI-driven drones are effective for farming. This research aims to improve sustainable, economical and data-supported weed management in rice farming which can help improve modern and greener rice growing.