Robust and Fast Template Matching Visual Odometry Based on IMU and Laser Range Finder Assistance

Reliable position estimation is significant for robot navigation and localization. The monocular visual odometry (VO) based on template matching is only suitable for robots running on relatively flat ground. Therefore, this article presents a robust and fast template matching VO. A laser range finder and inertial measurement unit (IMU) are added to the template matching VO. The camera image scale is transformed to reduce the impact of image scale uncertainty on robot motion estimation. Additionally, we introduce an adaptive search area method during the matching process, tailored to the robot’s speed, which substantially reduces computation time and enables real-time robot position. Experimental results on different types of pavements demonstrate that the algorithm is better suited to adapt to uneven pavements and has higher positioning accuracy. Furthermore, the proposed adaptive search area method reduces computing time by 59.92% when compared to the full search area.