Abukmeil, Reem2024-04-162024-04-162024-04-15http://hdl.handle.net/10222/83913In this research project, a sensing system based on spectroscopy was developed to offer a solution to the problem of slow and tedious method of understanding the status of nutrients in potato plants using chemical tissue analysis. In the past several researchers have explored the use of ground-based sensors to estimate nutrient content in different crops based on leaf chemical content. However, in potatoes, petiole rather than leaf is the main organ required for tissue testing. Hence, the step in this research investigated the relationship between the chemical composition of petioles and the reflectance spectra of leaves. Initially, a linear relationship was assumed between the chemical content of petioles and leaf spectra within the Vis to NIR spectral range. Two modes of lab spectral analysis, dried and fresh, were considered. Univariate linear regression models were constructed to estimate each nutrient individually. The results showed that a correlation between the chemical contents of potato petioles and the leaf spectrum was confirmed for 12 elements. The results further showed that the predictive power was higher in the dried mode, but also N, P, K showed a high potential to be estimated based on fresh spectra. Next, interlinkages between the nutrients were further assessed to enhance the predictive performance. Once the interlinkage was confirmed, a machine learning pipeline based on multivariate stacked regression was developed and implemented as it outperformed the univariate regression in estimating all nutrients. Finally, a full solution was developed based on portable spectrophotometer which can be deployed in the field. The spectrophotometer connects to a cloud computing, where the pipeline is stored, once internet is available to provide the farmer with nutrient status of plants in a matter of minutes in the case of fresh leaves and within 24 hours in the case of dried leaves. The rapid estimation of the nutrient status will help in better management of the amount of fertilizer applied during the potato growing season once the system is deployed in field.enSpectroscopypetiolemacronutrientsmicronutrientsleaf reflectancevisible and near-infrared spectral instrumentsneighbor-based variable selectioncorrelated nutrientsmultivariate regressionunivariate regressionnon-destructive sensing solutionless-destructive sensing solutionThesis