This thesis presents an image-based system for measuring target objects on an oblique surface based on pixel variation of CCD images. By using the zooming function of a PTZ camera, we can simulate two images at two different photographing distances. By using the different pixel counts of the target object in the CCD image frame, two-dimensional localization of the target object on the oblique plane can be accomplished. To validate the effectiveness and viability of the proposed method, practical experiments are conducted in the thesis. Experiment results demonstrate that the proposed measuring system has a satisfactory accuracy in self-localization for mobile robots.