The classic application of scene recognition algorithms is the classify the image content into semantically meaningful entities in order to understand the image content. Then one could use content-based indexing to categorise the images for latter retrieval. Imagine a google for images which is not based on the name of the image and the context in which the images is found (such as page it's on and the tags), but rather based on the image content.
One older reference is A. Vailaya, M. Figueiredo, A. Jain and H.-J. Zhang, Image classification for content-based indexing, IEEE Trans. Image Processing, 10(1):117-130, 2001.
Being able to recognise images has applications for automatic tour guides too. For instance, this paper explores such a scenario: see the Eiffel tower, your camera phone will give you information about the Eiffel tower.
Another one is Lost Robot Localization.
There are undoubtedly many more applications, if you do some reference/citation chasing from these papers.