In this thesis, we present the design and implementation of a method for real-time person detection and tracking. Many current methods for detecting and tracking people rely on color contrast or movement to segment the image. Using color, however, requires the target and the background to be significantly different, and motion segmentation requires the target to be in constant motion relative to the background, often requiring stationary cameras. Pattern detection methods have also been applied to the problem of detecting pedestrians, but these approaches are slower and require stationary cameras to function. The method we present in this work does not require a color difference or constant motion to operate. We use Lucas-Kanade features to track feature points between left and right images, producing a sparse disparity map which is then segmented through the application of k-means clustering. We apply a Viola-Jones face detector to determine which, if any, of the resulting feature clusters represent a trackable person. This algorithm is tested using two identical standard cameras mounted on a mobile robot platform. Results are presented demonstrating detection and tracking of a person in several different situations, including partial occlusion and self-occlusion.