Photogrammetry recommendations for simulations creators

We are frequently asked to identify resources, guidelines and starting points when it comes to leveraging Reality Capture to create rich environments based on the real world by Unreal Engine simulation creators.

While the information is already available in the forums and delivered by tech writers and documentation teams, it looked like a where to start was missing.

This short post is meant to simply orient you to navigate through these resources.

:white_check_mark: First let’s start with high level recommendations from our Capturing Reality team to simulation creator we shared recently :

DO
• Use the camera with the highest resolution available
• Use high-quality prime (fixed focal length) lenses
• Use tripod, gimbal or other camera/lens accessories to help you
• Keep the object lighting consistent
• Ensure that the light source and lens are cross-polarized or using consistent diffuse / flat lighting
• Ensure that area where you are shooting is well lit
• Set your exposure triangle correctly
• Try to achieve 80% forward/backward and side/height overlap between your images (in all axis)
• The object of interest should cover 60% or more of the image frame
• Each point of the scanned object should be visible in at least two images
• Stop down your aperture to get rid of shallow depth of field
• Shoot Raw images and tweak them in postproduction software, if you want to get the best image quality

DON’T
• Don’t lose focus
• Don’t crop your images
• Don’t limit image count (shoot as many images as possible)
• Don’t undistort your images
• Don’t crop in your images
• Don’t oversample your images using super-resolution or AI algorithms
• Don´t use in-camera sensor-based stabilization and digital stabilization
• Avoid capturing photos with motion blur, bokeh, low shutter speed values
• Avoid object movement (i.e., due to wind, human interaction)
• Avoid capturing transparent, reflective and shiny objects if possible (use vanishing sprays if necessary, or a different type of coating to create features)
• Avoid featureless textures

:next_track_button: Now that you that we have your interest let’s dive more, you want to dive into these specific photogrammetry guidance content. Let’s start with these 3 essential lessons you can find bellow.

Start from the beginning : The following links are where you can find how to handle the photography skills you need to know about photogrammetry. The quality of the images taken from a digital camera differs between cameras and between photographers. (Main photographic principles explained in there such as the effects of Aperture on Depth of Field - following picture)


Using identical equipment, the quality of an image depends entirely on the skill and experience of the photographer. By adjusting the camera settings, a skilled photographer can produce high quality images sufficient for photogrammetry from almost any digital camera. You will find it all in the following link “Introduction to Photogrammetry”

Next you will dive in the steps to guide your scanning process so you establish the right pipeline. The photogrammetry process is summarized as reconstructing a subject from digital images. Digital images are captured in a process called “scanning”. The subject is scanned by taking digital images, using a digital camera, until every side, corner, and detail is captured. This “Principles of scanning for Photogrammetry” course covers the basics of scanning for photogrammetry.

After scanning a subject, the collection of scan images is called “scan data”. The scan data undergoes a quality review to remove blurry or out-of-focus images and then the remaining images are processed. The “processing RAW images” course covers the image corrections necessary to produce high quality reconstructions.

You may still have questions on these best practices to up your game with photogrammetry pipelines to create the right simulation applications, there is full curriculum training course available with samples you can get from our team.

The course content covers the entire workflow from inputting source data to exporting 3D models for use in other software applications. The course is intended for new users of RealityCapture and provides instruction at a Fundamental level.

To access this course ping our teams either simulation@epicgames.com or sales@capturingreality.com

If you have a high-resolution sensor camera, don’t forget to add in additional computing power. Two thousand images times 64 megapixels is going to take time to process. It is a good idea to capture a color card and a scale bar, if possible, for reference.

1 Like