Establishing a Historical Fine-Resolution Urban Database from KH-9 HEXAGON MCS and PCS Imagery Including an Urban Landscape Map, Stereo Anaglyph Image and Digital Surface Model

Shahtahmassebi, A. and Shahtahmassebi, G. and Moore, N. and Atkinson, P.M. (2025) Establishing a Historical Fine-Resolution Urban Database from KH-9 HEXAGON MCS and PCS Imagery Including an Urban Landscape Map, Stereo Anaglyph Image and Digital Surface Model. PFG - Journal of Photogrammetry, Remote Sensing and Geoinformation Science.

Text (AmirReza_KH9_Round4_Final_4_AM_PMA_reference_Modified1)
AmirReza_KH9_Round4_Final_4_AM_PMA_reference_Modified1.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (3MB)

Abstract

Historical, fine-scale spatial information on urban regions such as heights, spatial patterns and categories are critical for various disciplines. However, obtaining such historical information prior to the year 2000 is challenging. Although declassified analog images from the KH‑9 HEXAGON (hereafter, KH-9)’s mapping camera system (MCS) and panoramic camera system (PCS) at meter to sub-meter resolutions during 1970s are available, the absence of camera meta parameters (e.g., focal length, extrinsics, lens distortion) and the single film-based panchromatic image format pose significant limitations. Herein, a multifaceted analytical framework is developed to reconstruct a historical fine-scale urban database from KH‑9 MCS and KH‑9 PCS images which relies on minimal technical information. For the KH‑9 MCS, a deep learning classification framework is proposed to retrieve historical urban landscape features from KH‑9 MCS imagery. Also ablation studies are proposed to examine the impacts of deep learning hyperparameters on historical urban landscape classification and to assess the efficiency of deep learning models in comparison with benchmark machine learning techniques. For the KH‑9 PCS, first a framework is developed using SuperPoint + SuperGlue (SPSG) models to create historical ground point controls (GCPs) from KH‑9 PCS for accuracy assessment purposes and estimating the exterior orientation of the stereo cameras. Structure from motion (SfM) techniques are adopted to create an ultra-fine spatial urban stereo-anaglyph image and digital surface model (DSM). Finally, ablation investigations are conducted to compare the performance of the developed techniques with benchmark approaches. Our results suggest that KH‑9 MCS and PCS images, when analyzed with appropriate methodologies, can significantly enhance the reconstruction of a historical fine-scale urban database. Moreover, both KH‑9 MCS and PCS images can contribute to revealing super-urbanization (i.e., unprecedented urban development) including super-new urban development and super-urban redevelopment as well as super-urban no-change (i.e., stable over the four decades, 1975-2023).