My PhD dissertation was about automated lineament and geological structure identification using multi-source data, using an object-oriented image analysis approach.
My study was not based on proposing an automated lineament extraction algorithm, it was rather based on providing a methodology which (a) took into account all the maps derived using remote sensing and image analysis techniques (NDVI, PCA, DEM -derived maps) (b) incorporated them into a fuzzy knowledge database and (c) produced a map of identified geological & topographic lineaments (fault lineaments, drainage pattern segments, man-made linear features). The fuzzy knowledge database was produced in the environment of the eCOGNITION software.
To whom is interested, part of this work can be found in
Chapter: An object-oriented image analysis approach for the identification of geologic lineaments in a sedimentary geotectonic environment (https://www.researchgate.net/profile/Od_Mavrantza).
Greetings: As Gamal suggested, subtle hints in the visual record might best be identified by manual methods. I have great respect for the possibilities of automated identification algorithms, but they certainly are challenging to develop and proof. For my needs in identifying and measuring Carolina bays, I use the "Ground Overlay" element in Google Earth as a yardstick. Create a transparent .png file for the structure's planform, then place that on Google Earth and fit it to a specific target. The handles allow for sizing in x & y, while the overlay can be rotated to capture the alignment. From the DOM, the overlay's meta data can be captured and processed to yield the measurements. I've measured over 45k bays this way. The process can be seen in one of my submitted papers.
Article Measuring the Carolina Bays Using Archetype Template Overlay...