Impact of Dimensionality Reduction Techniques on Classification of Hyperspectral Images

Applications of Hyperspectral Imaging (HSI) has been phenomenally increased in recent times, there by establishing the need for accurate HSI analysing techniques. The major and primary challenges observed in analysing HSI is that the data is of very high spatial and spectral resolution and hence, generates the need for dimensionality reduction techniques in the pre-processing stage. This paper focuses on understanding the role and impact of different dimensionality reduction techniques when working on different machine learning techniques in classifying the hyperspectral image. We used agricultural, urban and mixed land cover datasets to evaluate the performance of dimensionality reduction techniques and their impact on machine learning based classification.