Table 13.

Comparison between ML and non-ML techniques for HSI classification.

Methods	Advantages	Disadvantages
Classical state-of-art techniques	(i) Simple structure and design	(i) High space complexity due to the storage of bulk data
	(ii) Less time consumption	(ii) Based on empirical identities, hence a tedious workpiece
	(iii) Easy to implement	(iii) Feature selection and extraction are not accurate
	(iv) Dimension handling skillfully by PCA and ICA	(iv) Suffers from limited labeled sample problem, Hughes phenomenon, and noise
	(v) Better binary and moderate multiclass classification by kernel and SVM
Advanced machine learning techniques	(i) Easy dealing with high-dimensional data, that is, troubles of Hughes phenomenon removed	(i) The construction of the model is difficult due to its complex network-alike structure
	(ii) Equally manipulative to labeled and unlabeled samples	(ii) High time complexity due to training and testing of the huge amount of raw HSI data
	(iii) Precise and meticulous choice of features	(iii) Extremely expensive design
	(iv) High-end-precise models to deal with real hypercubes, hence, top-notch classification accuracy	(iv) Strenuous to implement
	(v) Removes overfitting, noises, and other hurdles to a much greater extent
	(vi) Mimics the human brain to solve multiclass problems