Submitted to: Transactions of the ASABE
Publication Type: Peer reviewed journal
Publication Acceptance Date: 6/28/2010
Publication Date: 9/1/2010
Publication URL: hdl.handle.net/10113/46696
Citation: Huang, M., Lu, R. 2010. Optimal wavelengths selection for hyperspectral scattering prediction of apple firmness and soluble solids content. Transactions of the ASABE. 53(4):1175-1182. Interpretive Summary: Hyperspectral scattering is a promising technique for characterization of light scattering and absorption features in fruit. The technique is useful for assessing fruit firmness and soluble solids content, two important quality attributes for apple. Extraction of the most useful information from the hyperspectral scattering data is critical to accurate prediction of fruit firmness and soluble solids content. This research applied an artificial intelligence method, called hierarchical evolutionary algorithm, to select optimal wavelengths from the hyperspectral scattering data for 'Golden Delicious' apples for predicting fruit firmness and soluble solids content. The prediction models developed using the optimal wavelengths gave better prediction results for both firmness and soluble solids content, compared with partial least square method with full spectra, which is widely used in spectral data analysis. This artificial intelligence method provides an effective means for selecting specific wavelengths to develop prediction models for fruit firmness and soluble solids content. The method can be integrated with the hyperspectral scattering system to achieve the goal of sorting and grading apples for internal quality.
Technical Abstract: Hyperspectral scattering is a promising technique for nondestructive quality measurement of apple fruit, and extraction of the most useful information from the hyperspectral scattering data is critical to accurate assessment of fruit firmness and soluble solids content (SSC). A hierarchical evolutionary algorithm (HEA) approach coupled with subspace decomposition and partial least squares regression was proposed to select the optimal wavelengths from the hyperspectral scattering profiles of ‘Golden Delicious’ apples for predicting fruit firmness and SSC. Seventeen optimal wavelengths were selected for firmness, which nearly spanned the entire spectral range of 500 - 1,000 nm, and 16 optimal wavelengths, all of which were above 600 nm, were selected in the SSC prediction model. The model using the 17 optimal wavelengths for predicting firmness yielded better results (r = 0.857, root mean square error of prediction or RMSEP = 6.2 N) than the full spectrum model (r = 0.848, RMSEP = 6.4 N). For predicting SSC, the model using the 16 optimal wavelengths model also yielded better results (r = 0.822, RMSEP = 0.78%) than the full spectrum model (r = 0.802, RMSEP = 0.83%). The HEA approach provided an effective means for optimal wavelengths selection and improved the prediction of firmness and SSC in apples compared with the approach of using full spectrum.