Application of X-ray imaging and convolutional neural networks in the prediction of tomato seed viability

• Published in: IEEE access 2023-01, Vol.11, p.1-1
• Main Authors: Hong, Suk-Ju, Park, Seongmin, Lee, Chang-Hyup, Kim, Sungjay, Roh, Seung-Woo, Nurhisna, Nandita Irsaulul, Kim, Ghiseok
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Convolutional neural networks; deep learning; Image processing; Model accuracy; Neural networks; Nondestructive testing; Plants; Prediction models; Predictive models; seed viability; Seeds; Seeds (agriculture); Structural integrity; tomato; Tomatoes; Transformers; Viability; X ray imagery; X-ray image; X-ray imaging
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2023.3265998

Tomato fruits are consumed worldwide owing to their health benefits, taste, and flavor. In tomato cultivation, seed viability is directly related to crop productivity. Currently, the methods used to evaluate seed viability involve destructive sampling tests; accordingly, nondestructive methods for predicting seed viability are urgently required. This study aimed to develop X-ray imagery-based models capable of predicting the viability of tomato seeds. Particularly, X-ray-imaged seeds were grown to the seedling stage, and seedlings were classified following their condition. The structural integrity of the seeds was calculated from the X-ray image processing, and an integrity-based viability prediction model was evaluated. Furthermore, convolutional neural network (CNN)-based viability prediction models were developed and evaluated. Both models showed strong performance in distinguishing germinated and non-germinated seeds. However, the CNN-based model revealed greater accuracy in seed viability prediction than the image-processing-based model. The CNN-based model accuracy was 86.01%, with an F1 score of 92.11%, indicating the usefulness of the developed nondestructive testing approach for evaluating tomato seed viability.