By my estimate, the map resolution is 5 cm per pixel. By comparison the best resolution available in Google Earth is 30 cm/pixel (and many land areas are available only in 2.5 m/pixel or 15 m/pixel resolution). As you would expect, flying close to the ground above a small area and capturing lots of images yields higher resolution.
Rock Hills Ranch is an Angus cattle ranch in South Dakota. Conservation is a big part of the owners’ philosophy to ranching, paying particular attention to the grasses that are growing and using rotational grazing to mimic the migration of the buffalo that once roamed the northern Great Plains.
The survey flight
Images were captured on a downward facing GoPro camera flown by a 3D Robotics Y6 hexacopter. I used Droidplanner 2 to identify the survey area, and the app calculated the striations and turning points of the flight plan which I uploaded to the Y6. I set the GoPro to capture images using the built-in intervalometer.
I did a pre flight check, then a manual take-off and left the drone in loiter mode 5 m above ground level. While in that position I checked the gimbal had the camera pointing straight down. When everything looked right I changed to autonomous flight mode and watched as the drone ascended to 60 m above ground and started executing the flight plan. The last step of the flight plan included returning to a grassy patch for an autonomous landing.
Creating the map
The survey yielded 346 images, 12 Megapixels each. That’s more than 4 Gigapixels combined and it is a bit over the top for mapping a small area.
Overlap between images is absolutely essential when combining individual images into larger mosaics. The overlap helps the software determine how images are positioned relative to each other. GPS coordinates would help too, but the GoPro does not have GPS and I did not invest time in using flight log data to infuse geo coordinates into the images’ EXIF data which is a handy workaround for enriching non-geotagged images.
Maps Made Easy opened for business this week. They currently offer only one service, Stitched Aerial Imagery, but are rolling out georeferenced maps later this month. Uploading 300+ images took about an hour and a half. It took the service a bit more than ten hours to process the images into a stitched image that can be panned and zoomed.
The map offers spectacular detail for everything at ground level that has been captured from directly above. Most of the map appears in the quality and detail shown in the example below.
Structures (like trees or buildings) that protrude above ground seem to confuse the stitching algorithm, and as a result they appear warped. The giant cottonwood in the top left corner of the image below is an example. The haystack in the same image doesn’t suffer from the same distortions, and the ladder leaning on the haystack is so detailed that it is possible to count the rungs.
Towards the edges of the image, the map uses imagery that has not actually been overflown but is captured due to the wide-angle lens on the GoPro. This is evident in the difference between the left side and right sides in the image below.
The Maps Made Easy service provides a page for each map you have created. A link to the map is included for sharing as well as an embed code.
Maps Made Easy uses a point system to calculate pricing of their service. The map at the top of this page took 600 points to process, equivalent to $30. However, in future surveys I will use the planning tool on the website. It suggests that similar surveys can be produced at the cost of about $5 by optimizing the overlaps.
Why not a 3D map?
The area surveyed is quite hilly. Together with characteristics such as trees and buildings, a 3D rendering of the images would be fascinating. Given the amount of detail captured, a 3D photogrammetric analysis should be possible. However, two different photogrammetry tools (one online and one running locally) failed to deliver any usable output. Both tools were fed a subset of the images to keep the task simple. The locally installed tool processed the images for 30 hours but the output was an empty xyz space.
Given those setbacks I was thrilled to see Maps Made Easy deliver on the promise encompassed by their name.