Abstract: The omni-directional vision sensor can provide an image of 360 degree around the sensor and the directional angles of objects obtained from the image are relatively accurate. This study proposed a localization system for agricultural machinery navigation utilizing characters of omni-directional vision sensor. The system consisted of four artificial landmarks, an omni-directional vision sensor, PC, PC software and operating vehicle. The algorithm mainly included that the principal point of camera was obtained via calibration; the landmark positions were extracted by image processing to get the direction angles between landmarks from the image. Camera position was estimated according with circumferential theorem and geometric transformation based on the direction angles. Experiments were conducted on the level cement ground under daylight lamps in a 5 m×3 m rectangular area in the laboratory. Ten test positions were chosen equably. Results showed that the distance mean absolute error between estimated position and test position was less than 3 cm. In conclusion, the system is feasible.