Aarhus Universitets segl

Improving tomato seed quality- challenges and possibilities

Publikation: Bog/antologi/afhandling/rapportPh.d.-afhandling

  • Santosh Shrestha
The thesis investigates the possibility of using single seed near-infrared (NIR) spectroscopy, multispectral imaging (MSI) and NIR hyperspectral imaging (NIR-HSI) in combination with chemometrics for rapid determination of the tomato seed quality. The results of the PhD study are compiled in four manuscripts (MS). These non-destructive methods show the potential of sorting tomato seeds as per their viability and varietal identity. The results are discussed in the context of possible contribution from these methods in the improvement of the seed quality in Nepal.
In MS I, potential application of NIR spectroscopy in combination with chemometrics for prediction of tomato seed viability is demonstrated. The work in MS I also emphasises on identifying the important NIR spectral regions for the chemometric model that are relevant to the separation of viable and non-viable seeds. The NIR-HIS method was also investigated but showed poor separation between viable and non-viable seeds.
In MS II, the prospects of using NIR-HSI in varietal identification of seeds were investigated. The tomato seeds from same harvest year were identified with higher accuracy. However, ability of the model to identify the tomato seeds was reduced when the diverse seed lots of each variety were analysed simultaneously. The study revealed that chemical fingerprints from NIR-HSI are sensitive to the variation present in the tomato seeds from growing conditions and seed deterioration. Therefore, thorough validation of model before the use of NIR-HSI for varietal identification is very important.
The work on MS III demonstrates the utilisation in various scenarios for classification of tomato seeds using MSI. The results displayed that MSI can be used in confirming the hybridity of the hybrid seeds. Manuscript III showed that accuracy of the chemometric model to identify the seeds may decrease when many varieties are analysed simultaneously. Therefore, the study suggested pairwise analysis and stepwise classification of the seeds in order to increase the accuracy of the chemometric model.
In MS IV, visible near-infrared spectra from multispectral images of tomato seeds were extracted and investigated for varietal classification of tomato seeds. The results showed high accuracy of varietal identification for tomato seeds. The classification results from the spectra were comparable to the ones from the multispectral images (MS III). Therefore, the study shows the possibility of using either whole images or spectra without losing the accuracy for varietal classification.
ForlagAarhus Universitet, Institut for Agroøkologi
Antal sider124
ISBN (Trykt)978-87-93398-67-2
StatusUdgivet - 27 feb. 2017

Se relationer på Aarhus Universitet Citationsformater

ID: 105239035