Smart Farm Animal Intrusion Detection Using Multi-Modal IoT, Edge AI and Kafka Streaming

Abstract

Wild animals such as deer or rabbits cause significant crop losses worldwide and create major problems for farmers. Traditional methods, such as fences, are often too expensive, difficult to maintain and not consistently effective. This paper provides details about an animal intrusion avoiding system that uses a combination of sensors, including thermal cameras, microphones, and motion detectors. The sensors work in conjunction with artificial intelligence running on edge devices, sending alerts through a Kafka streaming system. Unlike existing systems that rely on only one type of sensor or fixed models, our method combines several signals and can also learn to recognize new animals with only a few examples. In simulation tests, the system achieved about ninety-four percent accuracy, reduced false alarms by more than a third, and responded in less than two hundred milliseconds. When compared with systems that used only motion sensors or only cameras, our approach proved to be more reliable. The work is still limited because it is based on simulations rather than real-world farm testing, but plans include real-world trials, adding long-range communication such as LoRaWAN, and utilizing advanced techniques like federated learning to make the system even stronger.