How to perform object recognition in MatLab?

3 views (last 30 days)
Igor P. Louzolo-Kimbembe
Igor P. Louzolo-Kimbembe on 25 Nov 2013
Commented: Image Analyst on 30 Oct 2014
I am currently implementing a program that will segment and identify objects placed inside a refrigerator (vegetables, drinks, food...). My goal is to be able to designate objects in the fridge. For example, given a refrigerator interior image, the program should be able to tell which object is an apple, an orange, a bottle of milk... My approach is to start with a couple of template images (in a database) and then match those against stuff in the refrigerator using normalized cross correlation. However function normxcorr2 only works best with templates that are sub images of a bigger one. I'd like to train my program to recognize and efficiently classify objects.

Sign in to answer this question.

Answers (1)

Image Analyst
Image Analyst on 25 Nov 2013
normxcorr2 is good when you have a template of what you want to look for. You probably won't have a template for the apple or the dish of butter. This is a very challenging task, one that goes far beyond the type of help we offer here. You're best off searching VisionBib for a suitable algorithm, or Google "image understanding" or "cbir" or things like that.

Categories

Find more on Semantic Segmentation in Help Center and File Exchange

Tags

Products

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!