Cascaded Keypoint Detection and Description for Object Recognition

Abstract

Keypoints detection and the computation of their descriptions are two critical steps required in performing local keypoints matching between pair of images for object recognition. The description of keypoints is crucial in many vision based applications including 3D reconstruction and camera calibration, structure from motion, image stitching, image retrieval and stereo images. This paper therefore, presents (1) a robust keypoints descriptor using a cascade of Upright FAST -Harris Filter and Binary Robust Independent Elementary Feature descriptor referred to as UFAHB and (2) a comprehensive performance evaluation of UFAHB descriptor and other state of the art descriptors using dataset extracted from images captured under different photometric and geometric transformations (scale change, image rotation and illumination variation). The experimental results obtained show that the integration of UFAH and BRIEF descriptor is robust and invariant to varying illumination and exhibited one of the fastest execution time under different imaging conditions.