Firmness is an important quality attribute for assessing fruit maturity, postharvest management, and consumer acceptance. Texture profile analyser and penetrometers are two destructive and time-consuming traditional methods for measuring firmness. Hyperspectral imaging presents a potential non-destructive alternative by acquiring the spectral and spatial data of fruits. The ability of hyperspectral imaging to predict firmness of Haskap berries (Lonicera caerulea) at various ripening stages is investigated in this study. Hyperspectral images of the Aurora cultivar were collected at 3 harvesting stages (early, mid, late), 3 growing positions (high, middle, low), and different sunlight exposure conditions (partial shade, full sun). The spectral data was collected for a selected region of interest. The reflectance spectra (396.92-1033.95 nm) were analysed, and different models were developed using neural network (R2= 0.44; RMSE= 0.35), general regression (R2= 0.41; RMSE= 0.36), partial least squares (R2= 0.45; RMSE= 0.35) and bootstrap (R2= 0.63; RMSE= 0.28) predictive modelling methods. The preliminary results of the research study findings imply that hyperspectral imaging is a viable tool for promptly estimating firmness and classifying the ripeness stage of haskap berries. By integrating hyperspectral imaging and data-driven approaches, growers can significantly enhance fruit quality and optimize decision-making processes, enabling better pre- and postharvest management.
