Presentation Title: Feature Extraction Techniques for Object Recognition

Committee:

• Dr. Jian Zhang(Major Professor)
• Dr. S.S.Iyengar
• Dr.Bijaya B. Karki

Date: Apr 11, 2011
Time: 1:00 PM
Location:256 Coates

Abstract:
Object recognition in computer vision is the task of finding a given object in an image or video sequence. Most Important and complicated module of this object recognition deals with Feature Extraction. In our Project we compare three different Feature Detection methods: Scale Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), and Automated Feature Detection, In-order classify the image into one of the pre-determined class of objects. We use different methods for classifying the image into a class of objects. KNN Search is one ofthe methods we used, also classified based on calculating of Euclidian distance and finding the minimum distance. In this project we compare the performance of the system with all the three feature extractions techniques, based on the Accuracy of results, and time taken for execution.

After analyzing the three Feature Extraction methods used, we found that SURF is recognizing the objects pretty well when compared to the other two methods. If Time is the factor of consideration Automated feature extraction is found to be faster. Where as SIFT feature extraction showed consistent performance both in terms of time and the quality of recognition.


All are invited.