An enhanced tunicate swarm algorithm with deep-learning based rice seedling classification for sustainable computing based smart agriculture

Abstract

Smart agricultural techniques employ current information and communication technologies, leveraging artificial intelligence (AI) for effectually managing the crop. Recognizing rice seedlings, which is crucial for harvest estimation, traditionally depends on human supervision but can be expedited and enhanced via computer vision (CV). Unmanned aerial vehicles (UAVs) equipped with high-resolution cameras bestow a swift and precise option for crop condition surveillance, specifically in cloudy states, giving valuable insights into crop management and breeding programs. Therefore, we improved an enhanced tunicate swarm algorithm with deep learning-based rice seedling classification (ETSADL-RSC). The presented ETSADL-RSC technique examined the UAV images to classify them into two classes: Rice seedlings and arable land. Initially, the quality of the pictures could be enhanced by a contrast limited adaptive histogram equalization (CLAHE) approach. Next, the ETSADL-RSC technique used the neural architectural search network (NASNet) method for the feature extraction process and its hyperparameters could be tuned by the ETSA model. For rice seedling classification, the ETSADL-RSC technique used a sparse autoencoder (SAE) model. The experimental outcome study of the ETSADL-RSC system was verified for the UAV Rice Seedling Classification dataset. Wide simulation analysis of the ETSADL-RSC model stated the greater accuracy performance of 97.79% over other DL classifiers.

Keywords

• smart agriculture
• crop monitoring
• rice seedling
• computer vision
• image classification
• deep learning