Adaptive Iterative Closest SURF for visual scan matching, application to Visual odometry

Laser scan-matching is frequently used for mobile robot mapping and localization. This paper presents a scan-matching approach, based, instead on visual information from a stereo system. The Speeded Up Robust Feature (SURF) is used together with optimization tools constraints to get high matching precision between the stereo images. Calculating the 3D position of the corresponding points in the world results in a visual scan where each point has a descriptor attached to it. These descriptors can be used to associate scans observed from different positions. Just like in the work with laser based scan matching a map can be defined as a set of reference scans and their corresponding acquisition point. In this paper a robust Visual odometry and 3D reconstruction algorithm based on Adaptive Iterative Closest SURF for scan matching is proposed. This algorithm combine the robustness of SURF to detect and match good features and the accuracy an Adaptive ICP algorithm in which, the 3D point are weighted with their inverse depth to give more importance for near points. The proposed algorithm is validated and compared with two other optimization techniques based on Singular Values Decomposition (SVD) and Quaternion. Experimental results using Pioneer 3AT demonstrate that our algorithm can work robustly in indoor and outdoor environments and produce accurate results in static environments.