I have found it easy to keep "originals" (reference images) by sorting them so that the the reference images are last. Then I can select just the duplicates. The steps are laid out in this "It's Foss" article How to Find and Remove Duplicate Photos in Linux
Since all of your pictures are in one folder, you can simply do the following.
If you want to delete duplicate photos for [multiple] images at the same time, select all the images from the left sidebar.
Now, go to the file menu -> View -> Sort Items and select By Similarity.
This will put all the referenced images at the bottom. You can select all the images in the right sidebar that are NOT labelled as duplicate and press the delete button.
For those who are scanning across different folders, it is possible to "group" them in digiKam and do the same technique. The only downside is that you cannot choose which folder is the "original".
By default, the duplicate photos are divided into sections based on their location, i.e. folders. You may change that by going to View->Separate Items in the file menu.
I choose "Flat List" for the above.
Final notes, This takes some scrolling to select things the way you want, but it still is a time saver.
Warning! If you select many images on the left bar, give DigiKam some time to do the sorting before you try selecting.