Robot localization is an important issue for the soccer robot. The robot has to decide strategy based on its position on the soccer field. In the robot soccer competition, the robot only can use its sensor information to achieve its localization. In this thesis, an omnidirectional image-based localization method design by the particle filter is proposed to achieve the localization of robot on a soccer field. The particle filter is also known as Monte-Carlo method. The multi-sensor information is used to be conditions to calculate the posterior probability of robot’s position and estimate its position in the particle filter. In this thesis, the information obtained by two sensors (an odometer and an omnidirectional image system) is used to be conditions. The odometer is used to detect the position and orientation of the robot. The omnidirectional image system is used to observe the marking lines on the field around the robot and the field information in the captured omnidirectional image can be quickly obtained by the proposed method. Some simulation results are presented to illustrate the proposed localization method. The proposed method is built based on the probability distribution of particles, so many experiments on the real environment are needed to improve the success of the proposed localization system. From the experimental results, we can see that the proposed method can estimate the location of the robot effectively.