Real-Time Multi-Sensory Animal Intrusion Prevention in Crop Fields Using CNN and ESP32

Wild animals' damage to crops is a significant challenge in agriculture that leads to considerable financial losses. Most conventional methods of deterrence using fences, scarecrows, and noise-makers have been found ineffective, labor-intensive, or expensive-especially considering the presumed worth of farm produce. This paper discusses the creation of a multi-modal, real-time intruding animal prevention system for crops using Convolutional Neural Networks (CNNs) integrated with ESP32 microcontrollers. The objective of this work is to propose a cost-effective, low-power, and autonomous system that can reduce crop damage due to animal intrusions by detecting and preventing them. The system consists of a series of infrared, ultrasonic, and motion sensors offering persistent field monitoring, while the visual information is processed by a CNN model for the proper intrusion classification. The system also self-captures intruders and provides real-time notifications on the farmer's mobile phone using IoT technology. Experimental results confirm that the system attains good detection accuracy and reliability under diverse weather and illumination conditions, confirming its effectiveness in securing agricultural land.