This paper describes simple artificial landmark model and robust landmark detection and tracking algorithm for mobile robot localization. Artificial landmark is a very simple and powerful tool for accurate mobile robot self-localization. We make landmarks using color patches with symmetric and repetitive arrangement. Such arrangement of color pattern is for more robust performance against photometric and geometric problems in indoor environments. Also, to overcome the brightness variation due to the illuminant change, we use chromaticity color space. To detect and track the landmark robustly, we adopt a stochastic approach that is based on the CONDENSATION – conditional density propagation – algorithm, which is capable of tracking object in real time. We use global histogram intersection and projected histogram intersection together as the similarity measure between landmark model and sample block. These features are robust against scale change, rotation and illumination change because the landmark has symmetric and repetitive structure. Experimental results show that proposed landmark is enough to be used in cluttered environment. Also, proposed algorithm detects and tracks the landmark in cluttered scene in near real-time robustly.