Determination of amylose content in fresh lotus by hyperspectral imaging

With the improvement of living standards, people have higher and higher requirements for the taste and nutrition of lotus seeds. Lotus seed as a medicine is also a kind of tonic, its amylose content directly affects the quality and taste of lotus seed. The amylose content of lotus seeds varies greatly among different varieties, so the determination of amylose content of lotus seeds is of great significance for subsequent processing. The traditional amylose detection is generally using iodine colorimetry, iodine affinity titration and cross-cutting infection method, these methods are time-consuming and laborious, and easy to be affected by experimental conditions!

Hyperspectral imaging technology is a non-destructive testing technology that can obtain rich spectrum and image information. Compared with chemical detection methods, it has the advantages of saving time, labor and environmental protection. In this paper, hyperspectral imaging technology was used to detect amylose of fresh lotus.

一、Materials and methods

1.1 Test materials

The samples were from Fujian province, and the varieties of Xuanlian, Guangchanglian, Jianxuan 36, Mantianxing, Space lotus and Xianglian were selected. After picking at maturity, the fresh lotus seed was stored in liquid nitrogen and transported to the laboratory, where it was refrigerated at 4 ° C for 12 hours.

1.2 Hyperspectral image acquisition and correction

The main components of hyperspectral imaging system include hyperspectral imager, light source, stage, black box and hyperspectral data acquisition software. The whole system can use the color spectrum hyperspectral camera fs-13, which can collect the spectral range of 400nm~1000nm, and the spectral resolution is 2.5nm. The hyperspectral imaging system is shown in Figure 1. The moving speed of the payload platform is set to 3.5mm/s and the exposure time is 30ms. The lens is 40cm away from the moving platform and straight down. Adjust the focal length of the camera of the spectrometer for black and white correction of the system.

1.3 Data Processing

Analysis software was used to extract the average spectrum of the region of interest (ROI) from the spectral image of lotus seeds. In order to eliminate the influence of noise and external stray light, the modeling effect of pre-processing methods such as first derivative, second derivative, SG smoothing, multiple scattering correction (MSC) standard normal variable conversion was compared, and the best pretreatment method was selected.

二、Results and analysis

2.1 Average spectrum of the region of interest

In this paper, the spectral curve of each pixel in the region of interest of a single sample is used for subsequent processing. The average spectral diagram after removing the head and tail noise (400nm~971nm) is shown in Figure 2. It can be seen from the figure that the variation trend of spectral values of different samples is consistent. The band has an obvious upward shift between 460nm and 570nm, which may be caused by the shift in the water band. The band has a relatively obvious absorption between 500nm and 920nm. It may be related to quaternary frequency doubling, O-H secondary frequency doubling and O-H primary frequency doubling of C-H group in amylose molecule.

2.2 Amylose content of lotus seeds

The results of correction set and prediction set of amylose content divided by SPXY method are shown in Table 1. It can be seen from the table that the amylose content of fresh lotus seeds varies greatly. The maximum value of amylose content of corrected lotus seeds is 227.90 mg/g, the minimum value is 100.82 mg/g, and the standard deviation is 44.73mg/g. The amylose content of the predicted sample is within the range of the correction set sample, so the sample division is reasonable.

三、Conclusion

In this paper, hyperspectral imaging technology was used to rapidly detect amylose content. The results show that the modeling effect is the best after using first derivative and multiple scattering correction MSC). Then SPA was used to extract 9 feature bands. The corrected set correlation coefficient (R) of the PLSR prediction model was 0.835, the corrected set root mean square error (RMSEC) was 1.802, the predicted set correlation coefficient (R) was 0.856, and the predicted set root mean square error (RMSEP) was 1.752. The relative analysis error (RPD) was 1.944. The correlation coefficient of prediction set of PLSR prediction model established by RC method (R. The prediction set root mean square error (RMSEP) was 1.897. The relative analysis error (RPD) was 1.761. This study provided a thought for further developing an on-line detection instrument for amylose content, and laid a good foundation.