Code for the method "Directed Network of Angular Similarity (DNAS)", from the work "Complex Networks for Modeling Texture and Spectral Features of Hyperspectral Images for Environmental Analysis"

Usage

• Make sure you have a mex compiler and that you can run/compile the .cpp source files.

• The main script is run.m. It extract features with DNAS using the given rset (radius set, e.g. [1,4] as in the paper). base is the path to the dataset with .lsm image files with names class1_sample1.lsm, class1_sample2.lsm, ... , class2_sample1.lsm... etc.

[data, classes] = run(base, rset)

• It extract features, train classifiers and show classification results. If you want to use the DNAS features with other classifiers or applications, the script also returns data, a matrix with DNAS features for each image, and classes, a vector with the ground-truth class of each image (corresponding to rows of the matrix data).

Dataset

• You can also download the biological dataset used in the paper. It has two versions (RGB and LSM) containing 148 CLSM hyperspectral images from leaves of Jacaranda Caroba specimens under different pollutant levels. The samples of the RGB version are 8-bit .png trichromatic color-images, while LSM samples are .lsm 16-bit files with all the 32 spectral bands.
• https://drive.google.com/drive/folders/16cz58Aa1mNAhXTHlXyGEYp7aR8W9bIGt?usp=sharing