基于高光谱成像技术的冷鲜猪肉水分和脂肪含量同步快速检测

打开文本图片集

Abstract：This studyaims to develop arapid method for determining the moisture and fatcontents in chilled pork using visible-near-infrared（VIS-NIR）/NIRhyperspectralimagingtechnology.Traditionallaboratorymethods wereusedtomeasure the moistureandfatcontentsof128chilledpork （longissimus dorsi muscle）samples.Hyperspectraldata werecolleted in the VIS-NIR （388-1045nm andNIR （930-1710nm ）ranges.Three prediction models were constructed using partial least squares regresion （PLSR），one-dimensionalonvolutional neural network （DCNN），and recurrent neural network （NN） basedonthe fullband spectra and were compared.Theeffects of diferentspectraldata preprocessngmethods： SavitzkyGolaysmoothing（S-G），S-Gfirstderivative（S-G1），andS-Gsecondderivative （S-G2）anddiferentfeaturewavelength extractionmethods： successive projectionsalgorithm （SPA）and two-dimensionalcorrelation spectroscopy （2DCOS）on the predictionaccuracyof thePLSRmodel.Among thethree models，thePLSRmodel wasselectedconsidering its stability asthebestmodelfor predicting the moistureandfatcontentinchilled pork.Forboth VIS-NIRandNIRranges，the model based on the raw data was more accurate than those based on S-G， S-G 1′ or S-G 2′ preprocessed data. To simplify the process，therawNIR data were selected for modeling.Comparedtotheful-bandPLSR models，thePLSRmodels built usingthefeature wavelengthsshowedslightlyreducedpredictiveperformance.UnderNIR，theoptimalmoisturepredictionmodel， builtsing S-G，SPAandLS，withoeffcntofeteationofpreictionofO71，pforedmariallyeteranidthe full-band model， indicating that feature wavelength extractioncan，incertain cases，improve modelconstruction.

Keywords： chilled pork; water content; fat content; prediction model; non-destructive testing

DOI：10.7506/rlyj1001-8123-20250609-169

中图分类号：TS251.5 文献标志码：A 文章编号：1001-8123（2026）05-0065-08

食品安全是当前全球食品工业发展的核心议题，而肉类新鲜度作为关键质量指标直接关系到公众健康安全]。（剩余15998字）