Detection of early decay in dynamic oranges by structured-illumination reflectance imaging combined with Hough gradient method and YOLO v10 model

Early decay induced by fungal infections in citrus fruit can lead to severe economic losses. In order to detect early decayed oranges in a moving state, this study proposed a new methodology based on structured illumination reflectance imaging (SIRI), which uses a fixed phase stripe pattern for illumination and introduces phase shift into the obtained pattern images through sample motion. The Hough gradient method was used to align the regions of interest in acquired pattern images. The direct component (DC) and alternating component (AC) images were recovered by using the 2-phase spiral phase transition (SPT) demodulation algorithm. At a speed of 100 mm/s, the combination of AC images and YOLO v10 achieved the highest classification accuracy, which was over 97 %. This study indicated that the proposed methodology provided a valuable reference to the online detection of early decayed oranges. • SIRI detects early decay in moving oranges with high accuracy. • Hough gradient method effectively aligns images of dynamic fruit samples. • Two-phase SPT efficiently demodulates structured-light images in motion. • Optimal fruit movement speed identified as 100 mm per second. • YOLO v10 model achieves detection accuracy of approximately 97.2 %.