Research On Adulterate Detection Of Wheat Flour Based On The Near-Infrared Hyperspectral Imaging

Wheat flour adulteration is an important issue for food quality and safety.Traditional chemical analysis methods are mature in the detection of wheat flour adulteration.However,these methods are time consuming,laborious,destructive,and technically demanding.The food industry needs a non-destructive,fast,and accurate method to detect the wheat flour adulteration.Hyperspectral imaging combines the advantages of spectroscopy technique and digital imaging,and has been widely used in various food quality and safety evaluations.In this study,near-infrared hyperspectral imaging was used to detect low-level(0.02%-5%)talcum and benzoyl peroxide in wheat flour.After normalization,the spectrum of each pixel is calculated as the first derivative band difference,which provides a categorizable result for the detection of talcum.That is,wheat flour and talcum powder can be classified and detected by threshold segmentation and their distributions can be displayed.The spectral correlation measurement and the band ratio calculation of the spectrum of each pixel provide a sortable result for the detection of benzoyl peroxide.In addition,the two methods are combined to achieve the simultaneous detection of talcum and benzoyl peroxide.Finally,the average spectra detected as talcum and benzoyl peroxide were extracted to verify the classification results.In the study of wheat flour adulterate detection,it was found that the "multi-layer effect" of hyperspectral imaging will affect the detection results,and the penetration depth affects the "multi-layer effect".Therefore,near-infrared hyperspectral imaging was used to detect talcum and benzoyl peroxide under different thickness(0.3-2 mm)of wheat flour to study the effect of penetration depth on adulterate detection.The results show that when talcum is used as a substrate,particles under 0.3 mm can be detected.When benzoyl peroxide of diferent particle sizes is used as a substrate,the penetration depth shows a certain trend.When the benzoyl peroxide is 20 mesh,the penetration depth is about 0.3 mm;when the benzoyl peroxide is 80 mesh,the penetration depth is about 0.8 mm.It shows that different underlayer powders will affect the penetration depth.In addition,the particle size of the substrate has an impact on the penetration depth,and may affect the results of adulterate detection.Near-infrared hyperspectral imaging was also used to study the effect of different particle sizes(20-80 mesh)on detection of benzoyl peroxide in wheat flour.Using the band ratio method to detect benzoyl peroxide,the results showed that 40 mesh and 50 mesh benzoyl peroxide can obtain the better detection results,which may because these two particle sizes are close to or slightly larger than the camera spatial resolution.In addition,the quantitative analysis method like partial least squares regression and extreme learning machine were compared with the band ratio method,all these methods have achieved similar results,which proves the accuracy from the side.All these results showed that the combination of near-infrared hyperspectral imaging and spectral analysis methods is an effective method to detect talcum and benzoyl peroxide in wheat flour.In addition,it was found that the penetration depth of hyperspectral imaging and the particle size of adulterants will affect the "multi-layer effect",which will affect the detection results of wheat flour adulteration.In future we need to do more research about the factors affecting the detection of powder adulteration by targeting more kinds of powders and combining the optical characteristics of powder particles.