IoT-Based Plant Identification Using Multi-Level Classification

Accurate plant identification is critical for applications such as automated agriculture and plant monitoring systems. However, traditional classification methods often face challenges in balancing accuracy and computational efficiency, particularly when handling large datasets or real-time processing. This research aims to develop a classification scheme that efficiently identifies plant types based on color and shape attributes, achieving high accuracy with minimal computational complexity. To address this, we propose a two-level classification approach using a Naive Bayes classifier in a hierarchical structure. The first stage utilizes simple color features to categorize the majority of images with high accuracy and low computational overhead. In cases where classification remains uncertain, the second stage extracts additional color and shape attributes, offering a more refined analysis of complex samples. The scheme is implemented within an Internet of Things (IoT)-enabled data acquisition framework, enabling real-time image data collection. The system was evaluated using four types of artificial plants placed in a growth chamber equipped with image sensors and LED lighting, with data processed through a cloud service. The results demonstrate that the two-level classifier outperforms single-level approaches, maintaining high accuracy by deferring more complex samples to the second stage without significantly increasing computational costs. This hierarchical classification scheme successfully balances efficiency and accuracy, making it well-suited for large-scale applications such as smart greenhouses, where reliable and rapid plant classification is essential.