This document describes an image preprocessing scheme for line detection using the Hough transform in a mobile robot vision system. The preprocessing includes resizing images to 128x96 pixels, converting to grayscale, performing edge detection using Sobel filters, and edge thinning. A newly developed edge thinning method is found to produce images better suited for the Hough transform than other thinning methods. The preprocessed images are then used as input for line detection and the robot's self-navigation system.