Drone mapping
Image capturing has critical meaning in UAV photogrammetry. If your photos aren’t good, then you can’t really expect high-quality output. Even the best photogrammetric software won’t produce high-grade orthophoto and Digital Surface Model if the quality of input data will be low. So to ensure the best possible outcome, it’s good to take some time and carefully plan the photo capture.
Apart from camera hardware and settings, the flight parameters have a strong influence on the pictures quality. If you want to make sure to get real quality results, consider every aspect from flight path, overlap between images, number of images to lighting conditions. After image capture, look through acquired photos and check whether they are in a good condition. For image processing purposes, you want to use only clear, not blurry pictures.
But even if the photos look good visually, that doesn’t always guarantee that photogrammetric software will successfully process them. Wouldn’t it be good to find out any data defects directly after data capture, while still on the site? In case any problems with photos arisen or an insufficient number of photos were taken, you could recapture images straight away. In that way, you wouldn’t have to come twice to the same location and save a lot of time. Sounds good, doesn’t it? But if you, like me had anything to do with image processing before, you know that it takes time. Most often a lot of time. And let’s be honest, no one wants to stay at the field with a laptop just waiting to process data.
Fortunately, there is a way to get around this issue. Firstly, low accuracy settings need to be used for image processing. Limiting the number of loaded photos will further reduce the processing time. You can achieve this in any dedicated program. Take for example DroneMapper. It’s a Windows software for photogrammetric imagery processing and simple GIS analysis. The Rapid version of DroneMapper is free and processes up to 150 images. It gives you Digital Elevation Model and orthomosaic as a final result. Today I’d like to show you how well the software performs and how useful it is for evaluating input data. I used a small dataset containing 70 images (Cadastre dataset; Courtesy of Pix4D/pix4d.com).
Image processing
What distinguishes DroneMapper software from others is its simplicity. The user intervention has been limited to a minimum. The whole processing can be done in just a few clicks. All you have to do is to upload the pictures and DroneMapper will do the rest. The locations of photos are displayed on top of a base layer. At this point, you can analyse how well the flight plan was executed and notice if there are some pictures missing.
The user can choose to run each processing step (alignment, DEM and orthomosaic generation) manually or use Overnight mode and run them automatically one after another. The latter function will be most useful for on-site image processing. Because of image limit, DroneMapper Rapid solution is intended for smaller areas. However, projects with a greater number of pictures can be managed as well. Instead of all the images, uploading only part of them will do the trick. DroneMapper Rapid generates DEM with resolution 8 times GSD, orthomosaic with resolution 4 times GSD, so you can’t expect high-quality outputs. Yet, the limited accuracy and number of photos are the reason why you get results in no time. It takes just about 20 min to be able to preview DSM and orthophoto and it definitely can be done on the site. If DSM and orthomosaic generation are errorless, you can expect that repeating data capture is not needed and that full processing will not fail. What is more, depending on the application of the results, maybe this resolution will be enough. If you need orthomosaic only for visual purposes or as a base map, it’ll do pretty well.
For my data, DroneMapper Rapid generated DSM with 40 cm/pix resolution and orthophoto 20 cm/pix (GSD was equal to 5cm). To give you an idea how DroneMapper Rapid’s results present in contrast to other software, I compared resulting DSM with the one calculated in Pix4D. For processing images in Pix4D, I chose the same resolution for DSM and orthomosaic outputs as in DroneMapper. Both programs deliver the results in GeoTIFF format. DroneMapper output coordinate system was WGS84, whilst Pix4D’s WGS84/UTM zone 32N. I opened DSM’s and orthomosaics in open source GIS application, QGIS.
Orthomosaics comparison
Orthophoto generated in DroneMapper has some noise along the borders of the area. The stitching between images is slightly visible, same as colour changes on objects surfaces. Pix4D created orthomosaic with even edges, without noise. Image stitching and colour changes between images are so subtle that no longer visible.
Digital Surface Models comparison
The first thing that I noticed analysing both DSMs is the difference in feature representation. The edges of objects like roofs are far less sharp and clear on DroneMapper DSM than on the other one. On the other hand, Pix4D didn’t do so well when reconstructing vegetation. The large portion in the bottom-right corner of the is covered with trees. This region shows well on DSM from DroneMapper but is missing on the other DSM. Next, I compared these two rasters by
Conclusions
DroneMapper Rapid is good for quickly checking on the site if the captured images produce the expected result. However, the quality of DSM generated by this software might not be sufficient for most applications. It gives a rough idea though how the surface looks. And of course, if this processing succeeds, we can assume that the full processing later will bring high-quality results. In case the software fails to process images, that is a sign that data set is not alright. It might be necessary to improve the flight plan and collect images again. The pix4D result works better in this case. When it comes to orthomosaics, if we need it only for visualisation purposes or as a base map, both will work equally well. DroneMapper Rapid has the advantage of being free solution.
What to read next
- Read the information about 3D reconstruction from images here
- Check the last text about integrating LiDAR and panoramic photos here
- To find out how machine learning is used in point cloud classification, check the previous text here
- Check how to learn more through point cloud visualisation options here
Test Pointscene today
Want to know how to include point clouds in your projects? Visit www.pointscene.com to explore many examples in gallery or start free trial and upload your own data within minutes.
