Smart Agriculture & Weed Management/Security

Weeds contend with crops for access to moisture, light as well as nutrients as farms have to deal with unauthorized access and vandalism by animals. Manual weeding and the application of chemicals are labor-intensive, costly and harmful to the environment while the efficiency of traditional on-farm security is limited by ineffective continuous monitoring. The emergence of new technologies, especially artificial intelligence, computer vision, and deep learning can offer a solution to this challenge. The paper reviews nine modern scientific works on the application of artificial intelligence technologies for weed control and security of farms. The reviewed works are constructed using different processes including object detection, instance segmentation as well as CNN classification, with the use of additional technologies such as Internet of Things sensors, thermographic cameras, drones, and edge computing in the process. All reviewed works are checked for methodology, results, and disadvantages found by authors. The results only allow stating that deep learning is helpful in the identification and segmentation of the objects of interest but show that some issues, such as high computational requirements, prohibitively expensive hardware, and the requirement of large labeled datasets, remain. Research gaps have been examined. These include lightweight edge deployment, multi-class detection, robustness to outdoor conditions, and the integration of weed management with farm security. The review then sets a foundation for researchers and students to build accurate, sustainable and affordable smart agricultural monitoring systems. Index Terms—Smart Agriculture, Weed Detection, Computer Vision, Deep Learning, YOLO, Mask R-CNN, Edge AI, IoT, Farm Security.