Estimation analysis in VSLAM for UAV application

This paper presents an in-depth evaluation of filter algorithms utilized in the estimation of 3D position and attitude for UAV using stereo vision based Visual SLAM integrated with feature detection and matching techniques i.e., SIFT and SURF. The evaluation's aim was to investigate the accuracy and robustness of the filters' estimation for vision based navigation problems. The investigation covered several filter methods and both feature extraction algorithms behave in VSLAM applied to UAV. Statistical analyses were carried out in terms of error rates. The Robustness and relative merits of the approaches are discussed to conclude along with evidence of the filters' performances.