Non‐Destructive Evaluation of Whole Kernel Proportion in Mechanically Cracked Walnuts Using X‐Ray Imaging
Journal of Food Process Engineering · 2026-07 · 49 (7)
DOI: 10.1111/jfpe.70713 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Food Process Engineering, online erschienen am 24.07.2026. Verfasst von Xianghao Chen und 6 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben.
Abstract (Originalsprache)
The walnut whole kernel proportion is a core indicator for evaluating the effectiveness of cracking processes. To address the subjectivity and secondary damage associated with traditional manual grading, this study proposes a non‐destructive evaluation method based on X‐ray imaging. X‐ray images of walnuts were collected under various experimental conditions. Using YOLOv8n‐seg as the baseline model, the D‐LKA and DCNv2 modules were introduced to construct an improved YOLOv8n‐seg‐DN model for segmenting kernel regions in X‐ray images. Based on the structural characteristics of the kernels, two image feature indicators—the maximum kernel block area proportion ( P max ) and the relative kernel block count index ( RI kbc )—were proposed. Ridge regression was then employed to establish the correlation between these image features and the whole kernel proportion, with the model's predictive accuracy evaluated through validation experiments. Results show that YOLOv8n‐seg‐DN achieved an mAP of 91.0%, Precision of 91.1%, Recall of 86.2%, and F1 score of 88.6%, significantly outperforming the original model. As compression deformation increased, P max decreased while RI kbc rose, accurately reflecting the structural transition from intact to fragmented states. Correlation analysis revealed that P max was significantly positively correlated with the whole kernel proportion ( r = 0.89), while RI kbc showed a significant negative correlation ( r = −0.87). The ridge regression model achieved an R p 2 of 0.88 and an RMSE p of 7.60% on the test set. In validation trials, the overall predicted mean was 61.75% with an average error of 4.08%, verifying the accuracy of the method. This study provides technical support for optimizing low‐damage walnut cracking processes.
Zitieren
APA
Chen, X., Li, L., Chen, S., Shu, H., Hu, W., Zeng, Q., & Liu, T. (2026). Non‐Destructive Evaluation of Whole Kernel Proportion in Mechanically Cracked Walnuts Using X‐Ray Imaging. Journal of Food Process Engineering, 49(7). https://doi.org/10.1111/jfpe.70713
Vancouver
Chen X, Li L, Chen S, Shu H, Hu W, Zeng Q, et al. Non‐Destructive Evaluation of Whole Kernel Proportion in Mechanically Cracked Walnuts Using X‐Ray Imaging. Journal of Food Process Engineering. 2026;49(7). doi:10.1111/jfpe.70713
BibTeX
@article{Chen2026,
author = {Chen, Xianghao and Li, Lixia and Chen, Shenao and Shu, Han and Hu, Weiming and Zeng, Qingyu and Liu, Taoyi},
title = {Non‐Destructive Evaluation of Whole Kernel Proportion in Mechanically Cracked Walnuts Using X‐Ray Imaging},
journal = {Journal of Food Process Engineering},
year = {2026},
volume = {49},
number = {7},
doi = {10.1111/jfpe.70713}
}
EUVIMED