Abstract: The world heritage site of Mogao caves,
along the ancient Silk Road, consists of 492 richly painted Buddhist
cave temples dating from the 4th - 14th century. Cave 465 at the
northern end of the site is unique in its Indo-Tibetan tantric
Buddhist style, and like many other caves, its date of construction is
still under debate. The geopolitics of a historical period determines
the accessibility of artistic materials through trade and the
likelihood of technological and cultural exchanges through
communication. To resolve a long-standing debate on the date of Cave
465, ranging from the Tibetan period in the 9th century, the early
Tangut period in the 11th century to the Mongol/Yuan period in the
13th/14th century, the painting materials were analysed. Automatic
remote reflectance spectral imaging of large painted areas in high
resolution, from distances of tens of meters, a concept borrowed from
Astronomy, has made the imaging of entire architectural interior
feasible. However, it has significantly increased the volume of
data. Here we present a machine learning based method to automatically
detect ‘hidden’ writings and cluster the data into a materials cluster
map. Each spectral cluster was then analysed with non-invasive
complementary spectroscopic techniques for material
identification. The date was narrowed down to late 12th century to
13th century based on the material combination, paleographic analysis
of the revealed Sanskrit writings and archaeological evidence. This
study demonstrates a holistic interdisciplinary approach which can be
applied to all painted architectural interiors in the future.
|