emesent HOVERMAP-ZOE - Scanning Patterns and Flight Parameters

Hovermap-Zoe user manual
Release date: 01 Nov 2021
Revision: 1.0
17
•
•
•
•
•
2.4.4 Smaller scans
Hovermap uses a powerful embedded computer for data storage and autonomy algorithms. It has a
memory storage capacity of 480 GB. This means that Hovermap can store up to 10 hours of scan data
internally.However, there are many good reasons for breaking down the scanning project into more
than one scan mission:
Multiple smaller scans are more easily copied, processed, and managed.
Smaller scans greatly reduce processing and RAM requirements.
Smaller scans are often more accurate than larger ones and have less accumulated drift.
A slip in a small scan will have less impact on a large multi-scan job.
Breaking the survey into smaller scans allows processing to begin in parallel with scanning. This
allows for a faster overall process, and gives you the ability to check quality throughout the job.
When considering surveying a large area in multiple small missions, it is important to plan how you will
break down the area, as well as the scanning pattern you will use. Take into account all the
recommendations in this section.
Aim for scans to be under 35 minutes. You should also ensure that there is some overlap between
scans that you intend to merge. There should be approximately 25% overlap from scan to scan.
2.4.5 Scanning patterns
2.4.5.1 “Closing the loop”
Over time, Hovermap will also accumulate an error known as "drift". The best way to eliminate drift is
through a practice called "closing the loop". Closing the loop works by helping Hovermap to connect
features at the end of a scan, where the most drift has accumulated, to the features at the start of a
scan, where there is no drift.
One of the most effective ways to increase SLAM accuracy is to “close the loop.” This means that you
should stop the scan in the same general area that you started it. This location does not have to be
exact.
Try to close the loop as often as possible during the scan itself by creating many smaller loops along the
way. By doing this, SLAM will build a number of smaller “local” maps, and will then attempt to align
these (based on overlapping features) to produce a more accurate “global” map.