Abstract: “This paper presents a novel approach to scene localization and mapping in indoor environments from the concept of the Image Bag of Words (BOW) technique, where a group of native feature descriptors represents images and are subsequently transformed to a separate set of image words. This approach uses the famed algorithm called Scale invariant Feature Transform (SIFT). To extract distinctive invariant feature for reliable matching, we normalized the images as illumination changes affect the feature extraction. To achieve robust and efficient matching the environment is modelled. Clustering is shown to be appropriate for quantizing these descriptors into clusters based on selected threshold. In this work, we developed an efficient SLAM using images captured from a highly cluttered background. The result indicates a promising trend in using the camera for SLAM implementation.”

Keywords: mage Bag of words (BOW), Scale Invariant Feature Transform (SIFT), Clustering, Simultaneous Localization and mapping (SLAM), Images, Environment