Abstract:China's tobacco plantation industry includes a large scale and numerous employees. Nevertheless, the current tobacco curing control technology has higher labor intensity, however, it also cannot be adjusted based on the conditions of different tobacco leaves batches reducing the quality of tobacco. The tobacco's quality can be improved after curing and labor intensity can be reduced by modeling the state prediction of the tobacco curing process, accurately predicting the state of the tobacco curing, and making timely adjustments to the curing process. The area, color, weight, and some chemical substances of tobacco leaves significantly change during the tobacco curing process. This can be theoretically used as the input feature of the state prediction model. However, it is difficult to calculate the changes of the area and the changes of chemical substances in real-time owing to the complexity of the intensive curing room environment. Only features such as color and weight are easier to extract indicating that the model has fewer available features, and the accuracy of prediction using a single model is relatively low. Considering this problem, a state prediction fusion model (SPFM) was proposed integrating Long-Short Term Memory (LSTM) and Extreme Gradient Boosting (XGBoost). At the same time, based on the characteristics of the data set, a new data processing is proposed for the tobacco curing data set. By denoising, the image, the characteristic values of the RGB (Red, Green, Blue), HSV (Hue, Saturation, Value) color space were extracted. Then, the data pre-processing such as standardization, data cleaning, and label digitization were performed on the data. Furthermore, an intelligent tobacco curing platform was designed to integrate data collection, online monitoring, data mining, and status prediction, and SPFM was embedded in the platform. A comparative test was conducted, based on the real data collected via a tobacco station in 2019. The results indicate that SPFM has a better performance compared to support vector machines, artificial neural network, and the base models of SPFM such as XGBoost and LSTM. The accuracy of SPFM is 0.974, with an increase of 4.8–59.7%; the macro recall of SPFM is 0.952, with an increase of 8.2%–49.9%; the macro recall of SPFM is 0.936, with an increase of 8.2–75.0%; and the macro F1-score of SPFM is 0.943, with an increase of 9.7–108.6%.

An Intelligent And Automatic Control Method For Tobacco Flue-Curing Based On Machine Learning

Enhancing prior-knowledge-based control with deep attention neural network for outlet moisture content of cut-tobacco

State recognition and temperature rise time prediction of tobacco curing using multi-sensor data-fusion method based on feature impact factor

Research on state prediction method of tobacco curing process based on model fusion

Intelligent large-scale flue-cured tobacco grading based on deep densely convolutional network

A predictive model for the sensory aroma characteristics of flue-cured tobacco based on a back-propagation neural network

Feasibility of an NIR spectral calibration transfer algorithm based on optimized feature variables to predict tobacco samples in different states.

A novel integrated system for real-time monitoring of tobacco leaf images in the bulk curing barn

Multi-channel and multi-scale separable dilated convolutional neural network with attention mechanism for flue-cured tobacco classification

Regression prediction of tobacco chemical components during curing based on color quantification and machine learning

The Application of Hyperspectral Images in the Classification of Fresh Leaves' Maturity for Flue-Curing Tobacco

Rapid Recognizing the Producing Area of a Tobacco Leaf Using Near-Infrared Technology and a Multi-Layer Extreme Learning Machine Algorithm

An Artificial Neural Network Model to Predict Efficiency and Emissions of a Gasoline Engine

Intelligent techniques for cigarette formula design

Predicting Tobacco Pyrolysis Based on Chemical Constituents and Heating Conditions Using Machine Learning Approaches

Characterization of Tobacco Leaves by Near-Infrared Reflectance Spectroscopy and Electronic Nose with Support Vector Machine

Identification of Disease Type of Tobacco Leaves Based on Near Infrared Spectroscopy and Convolutional Neural Network

An adaptive optimization method toward batch-wise variable set point of outlet moisture content for the tobacco drying process

Tobacco Leaves Maturity Classification Based on Deep Learning and Proximal Hyperspectral Imaging

Intelligent Grading of Tobacco Leaves Using an Improved Bilinear Convolutional Neural Network.

Automated Process Parameters Tuning for an Injection Moulding Machine with Soft Computing