Citation

Sungheetha, Akey and Blake, John and Sharma R., Rajesh and Yogarayan, Sumendra and Kannan, Subarmaniam (2026) Advanced Computational Analysis of Martian Boxwork Formations: A Multi-scale Approach for Geological Feature Extraction. Lecture Notes in Networks and Systems, 1953. pp. 102-114. ISSN 2367-3370

Text 5.pdf - Published Version Restricted to Repository staff only Download (230kB)

Abstract

Martian boxwork formations represent critical paleoenvironmental indicators whose automated detection remains constrained by morphological complexity spanning 0.3–2.5 m ridge widths and polygonal compartments measuring 11.2–47.8 square meters. This study presents an integrated multi-scale meta-learning framework processing HiRISE imagery across five scale factors ( ) achieving 95.3% precision, 93.2% recall, and 89.1% intersection-over-union through weighted ensemble integration ( , ). Individual scale optimization reveals 0.707 downsampling maximizes network connectivity assessment at 94.7% while 1.414 upsampling captures fine features at 94.2%, with baseline 1.0 resolution providing 93.6% balanced accuracy. Inverse rendering reconstruction attains 0.18 m depth precision for simple rectilinear networks and 0.29 m accuracy for dust-covered formations, maintaining sub-0.5 m errors across 60–195 m distances. Paleoenvironmental recovery from 127 sites yields diffusivity square meters per second, precipitation rate per second, water activity , ridge relief heights 1.3 ± 0.4 m, and mineralization durations years under neutral-alkaline conditions (pH 7.75 ± 0.24, temperature 21.05 ± 3.2 C) compatible with sustained habitability.

Item Type:	Article
Uncontrolled Keywords:	Computational analysis
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	02 Sep 2026 07:20
Last Modified:	02 Sep 2026 07:20
URII:	http://shdl.mmu.edu.my/id/eprint/16538

Downloads

Edit (login required)