Visual Odometry by Multi-Frame Feature Integration

The video shows the results of the MFI method applied to data set number 13 of the KITTI Benchmarking Suite (http://www.cvlibs.net/datasets/kitti/eval_odometry.php). Features are detected using Harris corner detector, and associated by brute force matching of FREAK descriptors in consecutive images. Features are integrated over time generating an augmented feature list. The inter-frame feature matches and the augmented feature list are used by the visual odometry method to compute the optimal rotation and translation. Hernan Badino, Akihiro Yamamoto, and Takeo Kanade, "Visual Odometry by Multi-frame Feature Integration". In International Workshop on Computer Vision for Autonomous Driving (CVAD 13) @ ICCV 2013, Sydney, Australia PDF: http://www.lelaps.de/papers/badino_cvad13.html Hernan Badino http://www.lelaps.de/ This video was generated using my computer vision development framework QCV, available at SourceForge.net http://sourceforge.net/projects/qcv/ .