Efficient sugarcane billet damage detection and classification a novel approach with caso dqnn and ImABO_Twin Cap_BiNet Models

Abstract

Sugarcane, a cornerstone of the agricultural economy in tropical and subtropical regions, faces a pressing issue, such as damage incurred during crucial stages like harvesting, transportation, and storage. These damages not only result in reduced yields, but also compromise the quality of sugarcane products, impacting the livelihoods of farmers and the profitability of sugar mills. Accurate and timely detection of sugarcane billet damage is imperative to curtail these losses. The proposed research includes two main contributions to overcome the issues associated with the existing methods of sugarcane billet damage detection and classification. First, the work employs a Deep Convolution Neural Network (DCNN) model for the detection of sugarcane billet damage. The Aquila Sailfish Optimizer (ASO) algorithm, inheriting the characteristics features of Aquila Optimization Algorithm (AOA) and the Sailfish optimization Algorithm (SOA) involve in tuning the weights of the DCNN model in such a way to enhance its efficiency. Damaged billets are categorized using a Deep Quantum Neural Network (DQNN) trained with Chronological Aquila Sailfish Optimizer (CASO) for accurate classification. In the second contribution, the research work propose the Improved Artificial Humming Bird based Twin Attention Capsule Bidirectional Long Short Term Memory Network (ImABO_Twin Attn Cap_BiNet) model. This model combines capsule networks with BiLSTM and twin attention layers, achieving comprehensive billet analysis.The Improved Artificial Humming Bird Optimization (ImABO) algorithm fine-tunes the model for superior performance. Results show CASO-based DQNN achieves 91% precision, 93.3% recall, and 92.1% F-measure. ImABO_Twin Attn Cap_BiNet demonstrates 96.35% accuracy, 99.61% recall, 96.35% precision, and 96.35% F1-Measure. These techniques significantly enhance sugarcane billet damage detection, ensuring high-quality crop yield. newline