Development of an IoT-Enabled Smart Scarecrow with Resource-Efficient Computer Vision

Modern agriculture faces a substantial number of challenges in protecting crops from wildlife, especially birds. Crop loss is caused by damage from wildlife, and the consequent economic loss and concerns about food security are high. The traditional scarecrows are ineffective against intelligent intruders, and this calls for something more advanced. In this paper, we present an IoT-enabled smart scarecrow that integrates detection of harmful animals in real time and deterrence using the YOLOv8s object detection model. The smart scarecrow gives movements through servo motors, sounds from buzzers, and lighting using the LED lights when it identifies threats, not unnecessary actuations from different entities that aren't threats, and it prevents any actions that could annoy an individual. The system integrates an advanced computer vision approach with the lightweight ESP32-CAM module, and achieves a detection speed of $\mathbf{3 1 5}$ frames per second, with mAP of $\mathbf{5 4. 5 \%}$. Results from the experiments show that this system significantly improves crop protection by successfully completing 97 % of trials. Due to its use of low-cost hardware components and a lightweight, edgecompatible neural network, the system offers a cost-effective solution that minimizes economic expense and avoids the need for power-hungry GPU resources. The findings indicate that such intelligent solutions are applicable and practical approaches in agricultural grounds that minimally involve manual labor and maintain healthy growing conditions for the environment.