An Unsupervised Clustering Approach to Location Classification

Conventionally, tracking people through an environment has been achieved by monitoring a series of fixed cameras. With the advent of wireless technologies, the option of inverting the paradigm and monitoring instead, from each person’s pointof-view, has become more accessible. By taking video sequences from a person moving through various environments, this paper explores a process for classifying the different locations encountered using chromatic information gathered from images. This involves extracting a set of simple features from each frame, applying an unsupervised clustering algorithm and classifying new images with a nearest neighbour method.