A single vision system is the simplest vision-based method of simultaneous localization and mapping (SLAM) but it cannot reduce navigation errors that occur in the direction of the optical axis of vision sensors. Moreover, the limited observation area of vision sensors sometimes leads to failure in tracking a feature point. A vision sensor-based inertial navigation system INS/SLAM integrated system performs better with multiple vision sensors than with a single vision sensor because the observation area is wider and the impact of the sensitivity of the vision sensors is considerably smaller. However, the geometrical arrangement of multiple vision sensors induces navigation errors. This paper analyzes how the performance of the vision sensor-based INS/SLAM integrated system varies in relation to the geometrical arrangement of multiple vision sensors and the observation area of the vision sensors. The analysis shows that vertical navigation errors decline when the vision sensors are aimed horizontally, and that horizontal navigation errors decline in a similar manner when the vision sensor is aimed vertically; moreover, this behavior is especially evident when the level of sensitivity peaks in the horizontal direction.