Virtual landforms

I've always been delighted and fascinated by the shape of the land. The delight part is pure aesthetics. The fascination part is all about how they achieve that form. To answer that question, one needs to study the processes that move sediment around, and that requires detailed measurements.

Lately, due to significant advances in photogrammetry, I can measure the shape of landforms relatively easily, and on a scale where very small changes can be measured. I've been more or less restricted to smaller landforms (from 1 m to 50 m) accessible from ground-based photography, but this year, I'll be working with UAS (unmanned aerial systems) to stretch to the 100 to 200 m scale features.

It's a continuous learning process, as the software and technology keep changing, and mastering suitable photo survey techniques for simple, robust, and reproducible results is full of trial and error. You can find more info and examples below.

Examples of Landform Models

Check out some of the models I have created and posted on SketchFab. You should be able to click and drag the model to view it from different angles and zoom levels.

Landslide activity in upstate New York

For the last few years, I have been conducting repeat photo surveys of several landslides close to SUNY Oneonta's campus, and incorporating them into my classes. There's a surprising number of landslides, and I've have been completely surprised by how much change happens between surveys.

SfM Procedures and Tests

SfM stands for structure-from-motion, which is a technique where objects viewed from different locations can be mapped in 3D. One needs a camera and software to create 3D virtual objects of the real world. It's not very expensive to get into it, and the ability to monitor an active area quickly and repeatedly makes it possible to capture geomorphic events when mass is moving, and gain insight into how it all works.

Repeat surveys require consistent methodology, particularly when the whole landform is changing from one time to the next. I am trying to establish simple and robust methods for these photosurveys. With SfM, simple while maintaining reproducibilty and high fidelity is usually not possible, but I am making some headway.

If one is satisfied with creating a 3D object which can be rotated easily and used mostly for visual portrayal, then SfM surveys can be as simple as taking several photos with substantial overlap of the object of interest. More is better, and from all angles around the object. I've created a bunch of these models, and they can be quite instructive. See my models on SketchFab. Most of these are unscaled and not aligned to gravity or north.

If one needs to extract measurements from the 3D model, then a scale object must be included in the scene. This is an easy addition. Just place a ruler, or something which has known length, into the scene when you photograph it. The software will permit you to scale the model.

If one needs to know which direction is up or down (that is, the direction of gravity) and which direction is north, then several points with known distance between points and direction of north must be included. A flat-lying compass with measured locations on the compass; a flat-lying 8.5" x 11" sheet with printed markers at known distances on the sheet aligned to north; a flat-lying ruler folded into a square and aligned to cardinal directions--I have found that all of these can provide enough information to scale and orient a model. There are some exaggerated errors if one extrapolates too much from a small scale object. But to first order, having a model aligned within a couple degrees of north and with distance errors of a few % is suitable to answer many questions about geologic shapes, rock and fracture orientation, etc.

If one needs a 3D model that can be combined with other spatial data, then scale and orientation and location of the model must be well established. A whole new system has to be brought into play--GPS. Depending on the level of precision, this gets complicated and while better cheaper devices continue to come on line, including high precision GPS location systems is expensive, both in time and money.

Drones

SfM can work remarkably well, and the more points of view of an object you obtain, the better the 3D model of the object becomes. Enter a mobile platform for a camera: drones. I currently am learning how to fly small quadcopters, and running tests on the quality of models one can reconstruct from them.

What I have learned thus far:

• Flying is a hoot!
• Elfie (JJRC H37) drone video can be used to reconstruct an object! (~$40 drone). It's also very challenging to fly in the wind--kind of like riding a wild stallion. The controller is an app and a smartphone, and I keep losing track of where my finger is on the screen. Transfer of imagery to the phone is very convenient. My Nexus 5 uploads the imagery to Google Photos automatically.
• Holy Stone F181C is a very stable flyer, but the imagery, which touts 720 p, is less sharp than Elfie. The controller is great, especially after having learned to fly on the smartphone. It's a lot of fun to fly...I think this is a very good trainer drone, and for around ~$100, it's ok. I hope it lasts--I've only flown it for a couple days (as of May 13, 2017).
• Syma X8HG
• MJX Bugs 3 with a Campark Extreme I+ action camera
• Parrot Bebop 1 is a very stable flyer, though on a breezy day, altitude hold means +/- a meter or two. Nice clear imagery, very suitable for object reconstruction, and the ability to set up a flight plan along with pans, tilts, etc is very useful. Not bad for under $200, though that is without a dedicated controller. Flying with the phone isn't bad, though.
• DJI Phantom 3 Standard
• DJI Mavic Pro