Problems importing DJI Phantom images into QGIS

Question

I am trying to import DJI Phantom 4 multispectral imagery into QGIS. I have imported the imagery but the images are not in the right location (showing near the coast of Africa when they should be in QLD, Australia).

I have used the "Importphoto" tool which imported a point feature to actual right location of the map but the raster images from the drone imagery itself is not in the right place.

I am using an OSM or Google Satellite Images basemap.

Obviously, it is a projection issue and the solution is probably to use the right CRS but I'm unable to do so.

Answer 1

"Coast of Africa" sounds like Null Island. This is very likely a CRS-Issue.

However, your point's are at the right location so probably it worked just right. And the import-photo tool actually just creates point-features at the location where the picture was taken with all the relevant metadata (see the documentation ).

I think what happened was that your point-creation worked as intended and you may then have dragged the images itself into QGIS (since only points and no pictures show up). They are then treated like a raster-file but without associated CRS and hence they are all displayed at the same location, where Lat/Long is 0,0 (i.e. Null Island).

If you want to display your images at the location of the points where theywere taken you can link the file-path that is within the attribute table of your points to the images and display them at the given locations. There are many ways to do this, most via layer-styling and here you can finde some possible ways to do this: Show images related to features in QGIS?

Answer 2

The raster images from your multispectral camera are not georeferenced images they are geotagged images. The points created from Importphotos are going to be the point locations of the GNSS enabled camera at the time of image capture. These values are written to the EXIF data of your images and exploited by the Importphotos tool to create the point locations. You could use some application like EXIFtool, or gdalinfo to view the EXIF tags including the GNSS locations in the image files and confirm this. Linking your images to the points as mentioned by Vince is going to display the image in an image viewer. The images will not be drapped over the locations that they cover.

The images are not geotiff images, if they were, the location, pixel size, and rotation would be written to the geotiff and the images might display in the correct space (basically the georeferencing information would be stored inside the tif files). At best you have some associated world files with each tiff, jpeg, or png. If this is the case you could add your images to your QGIS project. You will need to ensure that your QGIS project is in the CRS as the CRS expressed in the world files since QGIS cannot project-on-the-fly these kinds of data. You can confirm the presents of world files by checking the directory and seeing if there is a world file for each image. For example. picture1.tif, picture1.twf.

I doubt you have world files in which case you have two options. Learn how to georeference images and do that work yourself or use a drone-based photogrammetry application to stitch the images together and georeference them. Some options are WebODM, Agisoft Metashape, and DroneDeploy. WebODM is freeware and the other two have free trials. The quality of these tools would depend on the amount of correspondence between images (observable objects like roads or fences) and the amount of image overlap.

There is a QGIS plugin called Vertical Photo Placer. VPP will take the location data, GNSS height, azimuth, focal length, and image size and roughly georeference your images. VPP does not have you CMOS sensor size, accurate elevations, and cannot handle any oblique angles. As such, the resulting product from VPP will likely be poor but it is a good start for further georeferencing.