In demosense/archeospec: Analysis of the Reflectance Spectra From Paintings: Classification and Endmembers.

library(archeospec)

# Set a seed value to ensure reproducibility
set.seed(934587)

Loading the data

The package includes a small set of signatures called signatures. Load the data

# Loading data
data(signatures)

Spectral class

Let's see the information of the signatures

print(signatures)
summary(signatures)
is.spectral(signatures)

Cleaning

We can filter signatures in three different ways: 1. We have used three different sensors for different wavelengths. The change of sensor introduces some error, which can be corrected by smooth_leaps function: just tell the first wavelength of each sensor (but the first one). 2. The lowest wavelengths might include more noise than the rest of the measures. We can filter out the lower wavelengths by using remove_head. 3. The highest wavelengths might include more noise than the rest of the measures. We can filter out the higher wavelengths by using remove_tail.

signatures <- smooth_leaps(signatures, leaps=c(1001, 1831))
signatures <- remove_head(signatures, head=400)
signatures <- remove_tail(signatures,  tail=2400)

Simple plots

We can see the simple plots like signatures and intracorrelation among wavelengths.

plot_signatures(signatures)
plot_intracorrelation(signatures)

Clustering + endmembers

If we are not sure aboud the possible number of endmembers, we can use the plot_elbow to check the K-Means intercluster error for different number of groups (k). For report purposes, we might want to select a specific value of k in the plot.

plot_elbow(signatures, k=2:4, selected=3)

VCA + K-Means

Apply VCA to compute the endmembers using k=2. Then, apply K-Means clustering using a number of clusters k=3.

signatures <- unmixing_vca(signatures, k=2)
signatures <- clustering_kmeans(signatures, k=3)

VCA results in tabular format

See information of VCA in tabular format

table_endmembers(signatures)
table_residuals_summary(signatures)
table_weights(signatures)

VCA and clustering results as plots

See information of VCA and K-Means as plots

plot_endmembers(signatures)
plot_clusters(signatures)
plot_mutualinfo(signatures)
plot_endmember_cluster(signatures)
plot_endmember_density_bar(signatures)
plot_endmember_density_box(signatures)
plot_residuals(signatures)

demosense/archeospec documentation built on May 28, 2019, 8:39 a.m.