Comparing Four Dimension Reduction Algorithms to Classify Algae Concentration Levels in Water Samples Using Hyperspectral Imaging

Reducing dimensions of hyperspectral data is very important as the removal of high-dimensional spectral variables could improve the predictive ability of the model. In the current study, four different linear dimension reduction algorithms, including principal component analysis (PCA), local preserving projections (LPP), neighborhood preserving embedding (NPE), and linear discriminant analysis (LDA), were used to reduce hyperspectral dimensions, and their classification performances on the algae concentration levels in water samples using hyperspectral imaging were compared. The LPP model showed satisfactory classification accuracy of 94.296 %, which was superior to the results based on reducing spectral dimensions with LDA (94.118 %), NPE (93.353 %), and PCA (90.588 %). The results demonstrated the potential of hyperspectral imaging coupled with dimension reduction methods in classifying water bodies with different algae concentration levels.