Similar work has been done in [8] where a sensor used in optical mouse is adapted for an outdoor environment on a mobile robot. The camera and image processing is incorporated in a chip which gives a frame rate at 6 kHz. Two sensors are used for calculating heading and traveled distance, but each sensor assumes planar ground. This assumption also makes the system sensitive to tilt of the robot
where one sensor is closer and the other is further away from the ground. Reported error after calibration on a straight planar surface was 0.3%. In [9] two cameras are used, also mounted on each side of
the robot. Each camera are measuring distance as a mono camera using phase correlation technique and the system are analyzed from an odometry perspective obtaining position and heading. The error reported was similar to wheel odometry. The objective with this paper is to demonstrate the feasibility of a system for high precision visual odometry in agricultural field environment. The contribution of this paper is a visual odometry system for uneven terrain and an evaluation of the feasibility by
experiments on real field situations.