Utilization of Advanced Slope-based Indexing Technique (ASIT) for the extraction of built-up land

https://doi.org/10.1080/01431161.2019.1585592

Try this

Normalized Difference Built-up Index (NDBI) is used to extract built-up features and have indices range from -1 to 1.

NDBI = (SWIR – NIR)/(SWIR + NIR)

or else

Plot the spectral curve of both LU types in ENVI. Refer to spectral libraries. on the basis of spectral curve create signature file for classification.

In my opinion, if you want to have the best result in land cover classification you can accept the fact that some objects that having the different about the spectral values is not much.Thus the application of algorithm to classify them will encounter with many difficults to discriminate among them. In your case is sand beach and build up area. You can use a Supervised Image Classification to create the polygons of region of interest based on spectral values to recognize two objects. Sometimes follow this way you will get a not very reliable results. The best way is to make a polygons for Sand Beach area then you this as the masks when you apply classification methods to create land use or land cover map. Good luck for you!

Here is the result:

https://doi.org/10.1080/10095020.2022.2066574