Temporal clustering of minced meat by RGB- and spectral imaging

The aim of this study was to cluster minced meat samples temporally by different imaging systems and to study if a normal consumer digital camera can be used to evaluate the quality of minced meat during storage period. A set of minced meat samples of differing age was imaged both by spectral imaging and by RGB imaging. Spectral images were also used as a reference data for reconstruction of spectral images from the RGB images. Hierarchical temporal clustering of the samples was made using normalized RGB image data, spectral image data and reconstructed spectral image data. Clustering results for spectral data were perfect. Reconstructed spectral data managed moderately, outperforming RGB data clearly. Finally, two critical wavelengths for temporal classification were determined. Bacterium concentrations were also specified for the samples, and the correlations between bacterial concentrations and image data were shown. It is shown in this study that with the proper calibration technique, a consumer digital camera can be used for coarse evaluation of minced meat containing time.

► Minced meat samples of differing age were imaged by spectral and RGB camera.
► RGB images were reconstructed to spectral images.
► Images were classified temporally by hierarchical clustering.
► Spectral images were clustered perfectly.
► The age of minced meat can be predicted by the ratio of two wavelengths.