This website is using cookies
Cookies Policy Last updated: (10/08/2020) VRGeoscience Limited ("us", "we", or "our") uses cookies on www.vrgeoscience.com (the "Service"). By using the Service, you consent to the use of cookies. Our Cookies Policy explains what cookies are, how we use cookies, how third-parties we may partner with may use cookies on the Service, your choices regarding cookies and further information about cookies. What are cookies Cookies are small pieces of text sent by your web browser by a website you visit. A cookie file is stored in your web browser and allows the Service or a third-party to recognize you and make your next visit easier and the Service more useful to you. Cookies can be "persistent" or "session" cookies. How VRGeoscience Limited uses cookies When you use and access the Service, we may place a number of cookies files in your web browser. We use cookies for the following purposes: to enable certain functions of the Service, to provide analytics, to store your preferences, to enable advertisements delivery, including behavioural advertising. We use both session and persistent cookies on the Service and we use different types of cookies to run the Service: - Essential cookies. We may use essential cookies to authenticate users and prevent fraudulent use of user accounts. Third-party cookies In addition to our own cookies, we may also use various third-parties cookies to report usage statistics of the Service, deliver advertisements on and through the Service, and so on. What are your choices regarding cookies If you'd like to delete cookies or instruct your web browser to delete or refuse cookies, please visit the help pages of your web browser. Please note, however, that if you delete cookies or refuse to accept them, you might not be able to use all of the features we offer, you may not be able to store your preferences, and some of our pages might not display properly. Where can your find more information about cookies You can learn more about cookies and the following third-party websites: ● AllAboutCookies: http://www.allaboutcookies.org/ ● Network Advertising Initiative: http://www.networkadvertising.org/
×
VRGeosciece Limited
VRGeoscience
01 August

Tips for better Geological SFM models

Some tips for collecting better digital outcrop models using SFM.

There are many good resources out there on creating SFM (Structure From Motion) models, but here are a few tips based on common errors I have seen over the past few years. Though far from an exhaustive list it will be enough to ensure you get good results.

Make sure you are aware of the local laws regarding flying your drone.

Look at drone videos on YouTube and you will find many examples where people have, quite often unintentionally, recorded themselves flying in a manner which breaks the law.  Spend a few minutes reviewing the laws for the area your working in. For the UK, more information on using a drone can be found here, or another good resource is https://dronesafe.uk/. The UK Civil Aviation Authority also has a good video on how to fly safe https://www.youtube.com/watch?v=HVJxeyEu7R8

Think about why you are collecting the data and plan your data collection accordingly. If you do not you may miss something important.

Do you need to identify sedimentary structures or measure fractures? Are you more interested in large scale geobody geometry? The scale at which you need to work will affect the way in which you collect the data. Factors such as time available or the number of batteries will also come into play, but the main focus is to get the data you need for the project you are working on. If you can get more data than you need within your logistical constraints then that is a bonus. 

Use the right equipment.

Equipment choice is important. A good camera, and in particular a good quality lens, will help you get better results. Fixed focal length lenses tend to give better results than zoom lenses.  I use a Nikon DSLR and a variety of lenses (28, 35, 50) for ground-based work. The Camera body is not particularly recent, but it still works well.

If it is not in the photographs it will not be in the model.

Seems obvious, but it is not uncommon for someone new to SFM modelling to wonder why the structure they are interested is not well resolved in the model, only to find their outcrop is a tiny blob in the middle of the images. This is quite often seen with drone data, and people simply being too far away with a wide-angle lens.

Do a trial run and check your camera settings.

Test, Test, Test. Take a few photographs and check the exposure and focus is as good as you can get. Yes, you can correct poor exposure in the image to some extent, but it will be time-consuming and the results are never as good as getting it right in the first place.  Which leads up on to:

Work in small batches.

One from Lean Manufacturing principles and useful in many circumstances. At regular intervals check the data you are collecting, and correct the settings, or modify the way you are collecting the data if needed. It is disappointing, not to mention wasteful of time, to collect many photographs only to find the exposure was wrong for most of your images as lighting conditions had changed.

Use a spotter.

I find this one especially useful for drone work, particularly when working at long range. In the UK you can fly line of sight up to 500m Horizontally and 120m (400ft) vertically, but at 500m (and much less in fact) a Phantom drone is a tiny speck. It is your responsibility as a pilot to keep the drone in sight, but a momentary glance at the controller can mean you cannot find your drone in the sky. A spotter standing with you also keeping an eye on the drone can help with this.

Remember the resulting model may not be scaled or georeferenced.

I have had questions about models being the wrong size or incorrectly oriented. This is always a georeferencing problem, perhaps your SFM package does not georeference data, or you did not collect GPS data for the purpose. There are ways of getting a reasonable georeferenced, but again planning at the data collection stage is important. VRGS provides tools for georeferencing your data which you can find here if you have GPS control points, or if you do not have accurate GPS data then you can get a reasonable orientation using straight edges on the outcrop (like rulers) and a compass.  

Take more images than you need.

It is far better to take more image than you need and then select a subset for processing than not having enough images in the first place. I work on an 80% overlap. For a terrestrial-based model I work out how many paces will give me the overlap I need then use that to work out where the next image will be (assuming you stay approximately the same distance from the outcrop). For a drone survey, I use the live feed to work out how fast I need to fly given a timed shot interval (a photo every 5 seconds for example), then keep at or below that velocity. Fly close – Fly slow. Remember a steady camera will give you sharper images.

Don’t stand in one place, move around.

SFM relies on the images being taken from different positions (its called Structure From Motion for a reason). If you stand in one place and take lots of images then this will not reconstruct. Move between taking each photograph. Plan a route along which you are going to take the images, this helps ensure you get good coverage.  If you have a good survey of photographs, then taking a set of images from one position will work as the algorithm has the other images in the survey to match tie-points to.

I hope this helps some people on their journey to digital outcrop modelling.  As with all tips, there may be times when some of these are not possible or appropriate (apart from obeying the law of course), in that case, use your judgement and remember the small batches principle.

×

TRY IT OUT

Register for a free one-month trial of VRGS. Try it out with your own data or with one of the demo datasets

FREE TRIAL


TOP