How we generate the photo database map
The Process Behind the Maps
Curious about the process behind our photo database map creation? It’s a fascinating blend of data science and imaging technology!
There are several layers that are visible on the map. The first layer, called 'geotagged images', is an automatically generated layer based on the geotags associated the imagery. The second layer, called ' Sfm_Areas' is a manually maintained layer that catalogues all the processed data in the photo database folder. If several features are overlapping, try right-clicking on the feature and the fill should empty leaving only the outline. You should be able to select layers underneath.
A screenshot of the online map, showing the various layers that are available.
Geotagged images
The 'geotagged images' layer makes us of a really useful feature of most imagery taken by drones (and some other cameras) is that the camera stores location data in a special format on the photos (called exif). We scan our database for files with a jpg extension. The exif data from these image files are then extracted. Did you know that there are over 200,0000 images with a jpg extension in the SEE photo database( only half of them are geolocated though!). It takes around 8 hours to do a full scan of the database and extract data. Following that, several steps to clean up and organise the data. For example, images without GPS data are filtered out. The remaining data are grouped and categorized by:
- The camera model, and make 📷
- The folder path where the photos are stored 📁
Every entry has a unique combination of the four parameters above. Each entry corresponds to a polygon on the map, and represents that group of images. There is also an opportunity to attach "hand" written data by adding [metadata files](\file\Shared\SEESPhotoDatabase\Example of metadata.xlsx) to the database. If a folder has a metadata file associated with it, we incorporate that data into the map and table.
Another neat feature is that *.psx (Metashape) files, often used for drone imagery processing are scanned and located. If we find any matches between the image scan results and the photos in those discovered psx files, we add the path to the psx file to the metadata.
Do you have geotagged images and would like try to scan your own photos yourself? Have a look at the instructions on the UC gitLab repository
SFM Catalogue
The SFM catalogue layer looks very similar to the geotagged images layer, but it has the benefit of having been checked by a human, which means it is likely to be more reliable. It also covers some datasets that are not picked up by the geotagged images layer. This layer also provides more detailed information about the datasets, e.g. the quality of the images or type of terrain.
Both of these layers create the interactive map. Using both layers together will give you the best possible information the data that is available.