Artificial ripening, frequently performed with the use of calcium carbide, speeds up the process but harms fruit and its nutritional value. Spotting unnaturally ripened fruits is crucial for the quality and safety of food, but current models are inefficient or too complicated, especially for diverse fruit varieties. Long-Short-Term Memory (LSTM) as well as Gated Recurrent Unit (GRU) extraction and selection of features techniques are combined using Extended Gradient Boosting (XGBoost) with Convolutional Neural Networks to deal with these challenges. The proposed structure segments fruit thermal pictures employing Saliency Maps to draw attention to relevant areas. LSTM and GRU models are fused to produce multiscale sets of characteristics, allowing the model to record temporal along with geographic characteristics. A Bacterial Foraging Optimizer (BFO) built around variance maximization retains high-density as well as discriminative features during feature selection. A novel Convolutional XGBoost Network (CXGBN) combines CNN's completely connected layers with XGBoost classifications for enhanced efficiency. On Mango as well as Apple data sets, precision, reliability, as well as recall improved 8.3 %, 4.9 %, and 3.4 %. The model efficiently identified artificially ripened fruits, decreasing classifying delays by 6.5 %. This study presents a novel hybrid framework for spotting prematurely ripened fruits using LSTM-GRU fusion and XGBoost CNN. The proposed model outperforms and scales existing methods by solving inefficiencies and using advanced optimization as well as classification techniques. It will be tested for adaptability to other fruit types and real-time applications using low-complexity feature sets along with advanced methods like Q-Learning as well as Auto Encoders that will enhance dynamical performance.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号