Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: TECHNOLOGIES FOR QUALITY MEASUREMENT AND GRADING OF FRUITS AND VEGETABLES

Location: Sugarbeet and Bean Research

Title: Comparison of optimal wavelengths selection methods for visible/near-infrared prediction of apple firmness and soluble solids content

Authors
item Zhu, Qibing -
item Huang, Min -
item Lu, Renfu
item Mendoza, Fernando

Submitted to: ASABE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: July 8, 2013
Publication Date: July 21, 2013
Citation: Zhu, Q., Huang, M., Lu, R., Mendoza, F. 2013. Comparison of optimal wavelengths selection methods for visible/near-infrared prediction of apple firmness and soluble solids content. In: Proceedings of the American Society of Agricultural and Biological Engineers Annual International Meeting, July 21-24, 2013, Kansas City, Missouri. Paper #13-1595860.

Interpretive Summary: Visible and near-infrared (Vis-NIR) spectroscopy is now being used for nondestructive quality measurement of fruits and other food products. To implement the technology, it is necessary to develop an effective calibration model relating the acquired spectral data to the quality attribute(s) of interest for a set of calibration samples. Vis-NIR spectra often contain a large amount of information, much of which is redundant or irrelevant. Hence, it is critical that an effective mathematical method be used to extract useful information from the spectral data with the minimum data representation. Currently, two schools of data processing techniques are being used to extract useful information or spectral features from the Vis-NIR spectral data, i.e., full-spectrum processing like partial least squares (PLS) and wavelengths selection algorithms, such as uninformative variable elimination (UVE), partial least squares projection analysis (PLSPA), standard genetic algorithm (SGA), successive projections algorithm (SPA), and affinity propagation (AP). This research evaluated and compared these five wavelengths selection algorithms with the full-spectrum PLS method for predicting the firmness and soluble solids content (SSC) of more than 6,500 ‘Delicious’, ‘Golden Delicious’, and ‘Jonagold’ apples that were tested in a two-year study in 2009 and 2010. Results showed that in most cases, UVE, PLSPA and SGA performed better than SPA and AP in predicting fruit firmness and SSC, although no single wavelengths selection method was consistently better than the other methods. The prediction results for firmness and SSC from using each wavelengths selection algorithms were generally not as good as those obtained by the full-spectrum PLS models. However, a simple fusion method, which averaged over the prediction results from the five wavelengths selection algorithms, consistently improved firmness and SSC predictions by 0.4%-4.8% and 0.4-5.6%, respectively, compared with the full-spectrum PLS models for the three varieties of apples. This fusion method provides a robust means for improving firmness and SSC prediction results for apples.

Technical Abstract: Visible and near-infrared (Vis-NIR) spectroscopy is now being used for nondestructive quality measurement of fruits and other food products. To implement the technology, it is necessary to develop an effective calibration model relating the acquired spectral data to the quality attribute(s) of interest for a set of calibration samples. Vis-NIR spectra often contain a large amount of information, much of which is redundant or irrelevant. Hence, it is critical that an effective mathematical method be used to extract useful information from the spectral data with the minimum data representation. Currently, two schools of data processing techniques are being used to extract useful information or spectral features from the Vis-NIR spectral data, i.e., full-spectrum processing like partial least squares (PLS) and wavelengths selection algorithms, such as uninformative variable elimination (UVE), partial least squares projection analysis (PLSPA), standard genetic algorithm (SGA), successive projections algorithm (SPA), and affinity propagation (AP). This research evaluated and compared these five wavelengths selection algorithms with the full-spectrum PLS method for predicting the firmness and soluble solids content (SSC) of more than 6,500 ‘Delicious’, ‘Golden Delicious’, and ‘Jonagold’ apples that were tested in a two-year study in 2009 and 2010. Results showed that in most cases, UVE, PLSPA and SGA performed better than SPA and AP in predicting fruit firmness and SSC, although no single wavelengths selection method was consistently better than the other methods. The prediction results for firmness and SSC from using each wavelengths selection algorithms were generally not as good as those obtained by the full-spectrum PLS models. However, a simple fusion method, which averaged over the prediction results from the five wavelengths selection algorithms, consistently improved firmness and SSC predictions by 0.4%-4.8% and 0.4-5.6%, respectively, compared with the full-spectrum PLS models for the three varieties of apples. This fusion method provides a robust means for improving firmness and SSC prediction results for apples.

Last Modified: 9/23/2014
Footer Content Back to Top of Page