Appropriate approach for matching an image on database using SURF Algorithm

Hi,

I'm trying to develop an application to compare the given image (a cluttered scene taken with camera) with the images which are stored in database and find most matching one in database. For example;

My Input Image will be like this:

And I want to match this input image with the reference image which is stored in database. Reference image in database will be like this:

I applied and understood the technics which are in the Matlab Tutorial: Object Detection in a Cluttered Scene Using Point Feature Matching. I could match these images using SURF Features.

Now, i want to use this approach and try to create an image matching application. My questions are:

1. What is the best approach for image comparison? Should i store the surf points of reference image to the database which are the output of detectSURFFeatures function?
2. What kind of query should i perform for this data type? It seems if i have thousand of images' surf points in my database, one to one comparision of every record in database with input image's surf points is costly.
3. What kind of database type is appropriate for this application if i consider performance issues? (RDMS or whatever)
4. Should i use bag of features approach? (According to tutorial it seems not necessary but i'm not sure)

I would like to get any suggestion regarding this topic.

Answer 1

[Raabii Ansari]

Have you got the answer? if yes then let me know what to do if you have to detect a traffic cone from the image of a road having a cone?