LED Light-Pipe Hyperspectral Technology for Visualizing Apple Quality

Abstract

A monitoring solution for the spatial distribution visualization of fruit quality is crucial for developing intelligent drying strategies. Hyperspectral imaging is one of the most representative approaches in this field; however, its high cost has limited widespread adoption. To overcome this limitation, a highly cost-effective hyperspectral imaging technology based on a compact light guide system is proposed. The technology relies on a novel optimized compact optical light guide to eliminate the spectral non-uniformity of the target surface caused by the different light fields of multiple monochromatic LEDs. During the design process, an embedded microprocessor-based control unit is developed to synchronize LED flashing with image acquisition. Based on this, a prototype system is constructed, covering the 400–1000-nm range with 28 spectral channels. In practical application, the hyperspectral optical performance of this system is tested, and it is further integrated with a PLS model to visualize the moisture content and SSC distribution in apple slices. For moisture content prediction, the training set achieved an R² value of 0.977 and an RMSE of 3.14, while the test set achieved an R² value of 0.972 and an RMSE of 3.62. For SSC content prediction, the training set yielded an R² value of 0.979 and an RMSE of 1.62, while the test set produced an R² value of 0.973 and an RMSE of 2.35. The results indicate that this simpler and more cost-effective hyperspectral imaging technology still achieves remarkable accuracy and is an important step toward dynamic quality monitoring of smart dryers.