Object Recognition Using Bows Model

Soon, Wei Jun (2016) Object Recognition Using Bows Model. Project Report. UTeM, Melaka, Malaysia . (Submitted)

Text (24 Pages) Object Recognition Using Bows Model.pdf - Submitted Version Download (357kB)

Abstract

This paper presents the findings on Object Recognition using Bag-of-Words model done by the author. For the past decades, object recognition had been used in image processing of many fields to carry out tasks such as image classifying, video searching, robot localization and optical character recognition (OCR). Among all these object recognition method, Bag-of-Words (BOWs) or Bag-of-Features (BOFs) is a very popular approaches because of its simplicity in object recognition. Bag of words model is used for documents classification where the occurrence of every word is treated as a feature to be used for classifier training. Due to the fact that BOWs model is a multi-step process, there are still many possible combination of method that has yet to be tried. Hence, in this project, researching and developing a better object recognition application in terms of feature recognition and processing speed using BOWs model will be emphasized. Several tests would be run to prove the functionality of the system and also to acquire the value for library size and threshold value that would generate the best performance for the system. The conclusion will be including the best combination of feature extractor descriptor – FAST SURF model along with recommended library size and threshold value. In the end, FAST SURF was proven to be faster than SIFT+SURF and also 1.373% more accurate than current existing system.

Actions (login required)

View Item