A Rapid Non-destructive Detection Method for Wolfberry Moisture Grade Using Hyperspectral Imaging Technology

Improper drying processes or poor storage conditions negatively affect the storage period and quality of wolfberries and consequently affect their long-term utilization. As a result, to effectively, fast, and subjectively determine the degree of moisture content of dried wolfberry fruits, this study investigated a rapid nondestructive method for the detection and classification grounded on hyperspectral imaging (HSI). Ten wolfberry's hyperspectral image sample sets of varying moisture grades were collected using an HSI device within a spectral range of 400.885-1002.193 nm. First-order derivative (FD) with Savitzky-Golay (SG) preprocessing method was selected for data cleaning and a support vector machine (SVM) and sparrow search algorithm-SVM (SSA-SVM) were utilized for the detection of wolfberry moisture gradient content. Also, competitive adaptive reweighted sampling (CARS) was applied to the preprocessed data for wavelength selection to simplify the calibration model, and the characteristic wavelengths were settled as 27. From the result, the SVM model built in the integrated FD-SG smoothing algorithm and CARS showed a 100% training set, and a prediction set accuracy of 84.00%. The SSA improved the conventional SVM classification accuracy, and consequently, the SSA-SVM model provided the best prediction accuracy, improving from 91.66 to 100% at an elapsed time of 1.76 s. The kernel parameter g and penalty factor c were set at 0.1641 and 0.2176, respectively. Conclusively, a combination of HSI technology and the CARS-SSA-SVM model could precisely detect wolfberry moisture gradient content and can be potentially developed for the detection of other related food/fruit in the processing industry. The moisture content of wolfberry has a direct connection with its quality. Dried wolfberry storage time is reduced when it contains a considerable level of moisture content in the fruit either due to improper drying process or bad storage conditions, making the wolfberry more likely to deteriorate quickly in quality. Therefore, limiting its long-term use. Traditional methods for detecting wolfberry moisture focus mostly on attractiveness and rely on the opinions of individuals. These methods are arduous, require a lot of time, and are highly impacted by biased elements. HSI technology, on the other hand, is non-destructive, quick/fast, subjective, reproducible, accurate, and pollution-free. The findings revealed that the HSI technique for wolfberry moisture detection is viable and capable of determining wolfberry moisture gradient content. Hence, could be used in the fruit processing industry for fast and accurate moisture content determination for safe storage, shelf-life extension, and processed products' quality preservation for long-term utilization.