Evaluation of Feature Detection Algorithms and Epipolar Geometry Based Camera Pose Estimation

Visual sensors are widely used for applications such as Autonomous Navigation, Localization, Mapping risky environments and the like. Vision based navigation involves various stages, in which Feature Detection, Extraction, and Matching are the primary steps. A number of algorithms are available for performing this stage in which SURF and SIFT algorithms are the most commonly used. This paper compares the algorithms based on certain parameters which includes the number of features detected, match points, and match rates etc.In order to estimate the camera motion after feature matching, different methods are used based on the available inputs. In this paper, Epipolar Geometry is used to estimate the camera pose for the available image points and intrinsic parameters.