Abstract: In response to the problem that it is difficult to accurately assess the maturity of Camellia oleifera fruits and traditional discrimination methods have limitations, this paper explored the feasibility of combining hyperspectral microscopy imaging (HMI) with grey level co-occurrence matrix (GLCM) to assess its maturity. Samples were collected at different maturity stages. Microscopic images of fruit shell slices were collected for extraction of spectral and textural features. Principal component loading (PC Loading) and two-dimensional correlation spectroscopy (2D-COS) were introduced to select characteristic wavelengths. Different classification models were developed. Results showed that the models fusing spectral and textural features performed better than did the single-feature models. The model developed using support vector machine (SVM) combined with quantum particle swarm optimization (QPSO) achieved the best classification accuracy (87.0%). In summary, the maturity of C. oleifera fruits was closely related to fruit shell microstructure, as well as texture changes in spectral images. This study also confirmed the feasibility and superiority of HMI combined with data fusion in maturity assessment.

Key words: Camellia oleifera fruit; hyperspectral microscopy imaging; grey level co-occurrence matrix; maturity