Image Distortion

Types of Image Distortions 

Image distortions happen when images are not well matched or stitched during the photogrammetry process. This usually happens due to a lack of clear distinguishable common features in each image.  

One of the most common errors we see happens during stitching due to wrong overlap settings. For images to be stitched together successfully, there need to be lots of common points within them to ensure that your photogrammetry software can place them correctly. This often results in distortions at the edge of sites, as well as around tall objects. Setting correct image overlap settings is particularly important when you have tall objects in your site, as the taller an object is, the less likely it is that you will have an image with a similar viewpoint of said object.  

Another cause for distortions in images is errors in stitching images of homogenous sites, such as dense forests, large bodies of water, etc. In these cases, there are usually not enough distinguishable features between the images to match them correctly during the stitching process. A third common reason for image distortion is poor image quality, which can arise for a variety of reasons. If the wrong settings are selected during the data collection process, images collected during your flight could end up distorted. Images that are grainy, blurry, or too dark will not stitch correctly and cause issues. 

Please refer to our section on image overlap settings to learn how to avoid distortions in different site settings. 

How are artifacts treated in the data processing stage?  

If your images happen to have a large number of artifacts, we, unfortunately, will not be able to safely distinguish between artifacts and true site features.  To deliver accurate and valid results, the site should be surveyed again with the correct image collection settings ensured.  

Currently, the only artifact that the AirWorks software can identify is terrain which is obstructed by foliage.  In these cases, AirWorks communicates with you to inform you that while we can proceed to process a site that is obstructed with foliage, the 3D contours may not be robust. 

Avoiding Image Distortions 

There is no one-size-fits-all approach to avoiding image distortions, but it ultimately comes down to proper image collection and choosing the right settings. These will all vary depending on the type of site, time of day, type of camera, etc. You should consult with your drone pilot or service provider before flying a site to communicate the necessary setting and parameter requirements for successful image-stitching and photogrammetry, as well as for autonomous drafting with AirWorks.  

Please refer to our section on data collection to see our recommended flight parameters for the best CAD output results. 

Point Cloud (3D) Point Spacing

Please note that you should only use point spacing for LiDAR data.  

In a LiDAR point cloud dataset, points are located throughout the site with a fixed distance between them as they repeat in a grid-like pattern throughout the site. This point spread metric measures the largest distance between either the rows or columns of points, which ensures full coverage of the site and guarantees accuracy.  

This image shows a LiDAR point cloud file with good point spacing.:

ex: bad point cloud spacing.

This image shows a LiDAR point cloud with bad point spacing: 

ex: good point cloud spacing.

Noise & Ground Thickness 

Every point cloud file will have a thickness that determines its accuracy. This thickness should be measured on a hard surface like a paved road and should be less than 1/5 of the desired contour interval.  So, if 1-foot contours are requested, the thickness should be under 0.2 feet.   

Currently, there is no simpler metric to follow in terms of thickness, so the 1/5 rule should be maintained to ensure that the data is accurate.  

ex:

File Size Limits

Technically, there are no file size limits when preparing data to upload to AirWorks. That being said, large files negatively impact workflow and will increase file upload and processing times. Anything larger than 4GB will likely incur this issue, which is why we recommend that you tile both your orthomosaic and point cloud files of large sites so that each file remains under 3GB. This will allow for easier upload and processing.  

File Naming Conventions

Please only use the following characters in your file names, as other characters will be replaced and you may encounter data upload issues when starting a project: 

Character Type Examples 
Uppercase letters A-Z 
Lowercase letters a-z 
Numbers 0-9 
Dash – 
Underscore 
Period 

The following is a list of illegal characters in our file naming conventions, and files that contain these will not be processed, which will result in upload issues and processing delays. 

Pound (#), percent (%), ampersand (&), curly brackets ({}), slashes (\ /), angle brackets (<>), brackets ([]), asterisk (*), question mark (?), currency signs ($ € ), exclamation point (!), single or double quotes, at symbol (@), plus sign (+), equal sign (=), pipe (|), comma, colon, and, semicolon. 

Please also note that blank spaces are not allowed in file names, and are some of the more common issues encountered when uploading datasets to the AirWorks app. 

Accepted File Formats

Now that you have collected your dataset and processed it with your choice of photogrammetry software, it’s time to prepare the files themselves for upload and processing through the AirWorks app. 

Currently, we accept the following types of files: 

Orthomosaic (2D)

Multiple versions of geotiffed .tif files only. Keep in mind that we currently do not accept .tiff files.  

  • 0.01 < GSD < 0.25 feet. 
  • GSD< 0.05’ feet for best results. 
  • Must be georeferenced to a recognized coordinate system, meaning it contains an EPSG code. 
  • Minimal distortions. 

*If your data does not meet the minimum GSD threshold, you will receive a warning in our app that looks like this:

GSD warning in AirWorks app.

Point Cloud (3D)

Multiple versions of .las and .laz are accepted.  

  • Point cloud density (PCD) should be above 400 points per square meter unless LiDAR has been used. 
  • PCD ideally above 800 points per sqm 
  • If LiDAR has been used, the space between points should be less than 1 ft in both columns and rows. 
  • Ideally, the space between points should be less than 6” 

Boundary

The AirWorks web-app allows you to draw the boundary of the site area that you would like processed. Doing this in the app reduces misalignment issues.

If you already have a boundary file and would like to upload it, we accept a single version of a .kml.  

  • The file should be a single feature boundary, fully enclosed.  
  • Ensure that the boundary is a closed polygon, (I.e. first point = last point).