Environment detection - How to detect "city" versus "landscape" background environment in computer vision?

Question

I'm searching for computer vision or machine learning algorithms/methods that are used to classify or differentiate two outdoor environments. Given an image with vehicles, I need to be able to detect whether the vehicles are in a natural open landscape (desert, in particular), or whether they're in the city.

I've searched but can't seem to find relevant work on this. Perhaps because I'm new at computer vision, I'm using the wrong search terms.

Any ideas? Is there any work (or related) available in this direction?

Answer 1

Think of what different features the desert has compared to an urban environment. The desert is probably restricted the a limited set of colors/hues. It is limited to mostly smooth textures. It lacks "man-made" geometries such as straight lines from building edges etc.

So you should be able to differentiate between desert and urban environment by looking at the color components (both the number of colors and the variation), textures and the amount of prominent straight lines/edges in the environment.

By googling "image processing classify outdoor environment" I found the paper Automatic Classification of Outdoor Images by Region Matching which can be a good starting point for reading up on environment classification. They seem to base the classification mostly on color and texture features.

These are just general ideas but hopefully they can help you get started. Good luck.

Answer 2

Perhaps these can be helpful (see also their references):

Aditya Vailaya , Anil Jain , Hong Jiang Zhang, On Image Classification: City Images vs Landscapes (1998)

and the more recent:

Li Fei-fei, Pietro Perona, A Bayesian Hierarchical Model for Learning Natural Scene Categories (2005)

Answer 3

I would suggest you to use a training classifier. You train the system by feeding in what you want and then you can do template matching or histogram matching to get good results.

Answer 4

There are many possible different approaches, each one based on a different classification technique. The following list of commonly adopted techniques, is a starting point to get you started studying the underlying relevant theory:

Principal Component Analysis;

Neural networks (probabilistic, fuzzy logic based, back propagation based, self-organizing maps);

Support vector Machines;

Nonnegative Matrix Factorization;

Wavelets;

Fisher Discriminant Analysis.