Abstract:

Abstract Plant image analysis plays a pivotal role across diverse domains such as agriculture, botany, and environmental monitoring. The accurate identification and classification of plant species from images are fundamental for tasks like biodiversity assessment, disease detection, and crop management. This study introduced an innovative approach to plant image analysis through the Multilayer Hybrid Deep Convolutional Neural Network (MHDCNN), a novel architecture that synergizes the strengths of CNN and LSTM. The objective is to amplify feature extraction capabilities and elevate classification accuracy, thereby achieving exceptional precision in plant species identification Various activation functions like “Tanh”, “ReLU”, “softmax”, and “sigmoid” are integrated into the architecture to impact learning dynamics. Extensive experiments using a diverse plant image dataset validate the approach. The MHDCNN achieves an impressive 99.8% classification accuracy, highlighting its effectiveness in handling complex plant images. By blending CNN + LSTM architectures and carefully selecting activation functions for enhanced feature extraction, this research advances plant image analysis techniques. This novel approach not only contributes to deep learning (DL) in plant biology but also paves the way for future innovations in image-based plant analysis methods.

Keywords: