Moldy core is a common disease of apples, and non-destructive, rapid and accurate detection of moldy core apples is essential to ensure food safety and reduce post-harvest economic losses. In this study, the acoustic method was used for the first time for the online detection of moldy core apples, and we explore the feasibility of integrating acoustic and visible–near-infrared spectroscopy (Vis–NIRS) technologies for precise, real-time detection of moldy core in apples. The sound and Vis–NIRS signals of apples were collected using a novel acoustic online detection device and a traditional Vis–NIRS online sorter, respectively. Based on this, traditional machine learning and deep learning classification models were developed for the prediction of healthy, mild, moderate, and severe moldy apples. The results show that the acoustic detection method significantly outperforms the Vis–NIRS method in terms of moldy apple identification accuracy, and the fusion of acoustic and Vis–NIRS data can further improve the model prediction performance. The MLP-Transformer shows the best prediction performance, with the overall classification accuracies for the fusion of Vis–NIRS, acoustic, Vis–NIRS and acoustic reached 89.66%, 96.55%, and 98.62%, respectively. This study demonstrates the excellent performance of acoustic online detection for intra-fruit lesion identification and shows the potential of the fusion of acoustics and Vis–NIRS.
Assessing the internal quality of fruits is crucial in food chemistry and quality control, and bruises on peaches can affect their edible value and storage life. However, the early detection of slight bruises in yellow peaches is a major challenge, as the symptoms of slight bruises are difficult to distinguish. Herein, this study aims to develop a more simple and efficient structured-illumination reflectance imaging system (SIRI) and algorithms for the early nondestructive detection of slight bruises in yellow peaches. Pattern images of samples were acquired at spatial frequencies of 0.05, 0.10, 0.15, and 0.20 cycle and wavelengths of 700, 750, and 800 nm using a laboratory-built multispectral structured-illumination reflectance imaging system (M-SIRI), and the direct component (DC) and alternating component (AC) images were obtained by image demodulation. A spatial frequency of 0.10 cycle and wavelength of 700 nm were determined to be optimal for acquiring pattern images based on the analysis of the pixel intensity curve of the AC image; then, the pattern images of all yellow peaches samples were obtained. The ratio image (RT) between the AC image and the DC image significantly enhances bruise features. An improved Otsu algorithm is proposed to improve the robustness and accuracy of the Otsu algorithm against dark spot noise in AC and RT images. As a comparison, the global thresholding method and the Otsu method were also applied to the segmentation of the bruised region in all samples. The results indicate that the I-Otsu algorithm has the best segmentation performance for RT images, with an overall detection accuracy of 96%. This study demonstrates that M-SIRI technology combined with the I-Otsu algorithms has considerable potential in non-destructive detection of early bruises in yellow peaches.