农业机械学报2025，Vol.56Issue(5)：159-166,8.DOI:10.6041/j.issn.1000-1298.2025.05.016
基于Stacking集成的籽棉回潮率信息融合检测方法研究
Moisture Regain Detection of Seed Cotton Using Information Fusion Based on Stacking Ensemble
摘要
Abstract
Moisture regain is a critical indicator of cotton quality,playing an essential role in cotton trading and processing.With the mechanization rate of cotton harvesting in Xinjiang exceeding 85％,accurately detecting moisture regain in machine-harvested seed cotton has become indispensable for transaction settlement.However,existing resistive-based methods for detecting seed cotton moisture regain exhibit low accuracy and poor robustness.An information fusion detection method leveraging resistive technology to achieve precise moisture regain measurement was proposed.Seed cotton samples,specifically the Xinluzao 80 variety,were collected from a cotton processing enterprise in Changji City,Xinjiang.From the same batch of seed cotton,totally 200 g per group was randomly selected.Under varying environmental temperature and humidity conditions,totally 50 g from each group was dried by using constant-temperature ovens to determine the true moisture regain.The remaining 150 g was tested in a constant temperature and humidity chamber by using a resistive sensing technique with pressure compensation.In total,517 sets of resistance values and corresponding moisture regain data were obtained.The relationship between seed cotton moisture regain,environmental conditions,and cotton density was analyzed,determining the influence of density on resistive measurements.Using environmental temperature,humidity,seed cotton resistance,and density as feature variables,five regression models were established:multiple linear regression(MLR),support vector regression(SVR),random forest(RF),backpropagation neural network(BPNN),and K-nearest neighbors(KNN).Additionally,a stacking ensemble model was constructed to integrate data-level and decision-level information fusion.The dataset was split into training,validation,and test sets at a ratio of 6∶2∶2,and hyperparameters were optimized by using grid search.RF,BPNN,and KNN served as base learners,while MLR was employed as the meta-learner in the stacking ensemble model.Comparative analysis revealed that the stacking ensemble model outperformed the individual models,achieving a coefficient of determination(R2)of 0.994,a root mean square error(RMSE)of 0.151％,and a mean absolute error(MAE)of 0.104％on the test set.These results validated the reliability of the proposed information fusion detection method.The stacking ensemble model demonstrated superior performance and stability across validation and test sets compared with single models.This approach was well-suited for moisture regain detection during cotton harvesting,baling,and trading,providing robust data support for trade settlement and process optimization.关键词
籽棉回潮率/信息融合/堆叠集成融合模型/电阻检测/回归预测模型Key words
seed cotton moisture regain/information fusion/stacking ensemble model/resistance detection/regression prediction model分类
农业科技引用本文复制引用
钱一夫,黄杰,方亮,段宏伟,张梦芸..基于Stacking集成的籽棉回潮率信息融合检测方法研究[J].农业机械学报,2025,56(5):159-166,8.基金项目
国家重点研发计划项目(2022YFD2002400)、兵团科技计划项目(2023AB014、2022DB003、2023ZD053)和兵团研究生科研创新项目(BTYJXM-2024-K38) （2022YFD2002400）
