Abstract:

A fast robust adaptive nonlinear scale feature detector (FRANSFD) is proposed. Noise is wiped off and edge response is guaranteed through the fast solving of the nonlinear scale space. Adaptive selection of the number of scale space and the adaptive and generic corner detection based on the accelerated segment test (AGAST), combined with frame Laplace filter via removing edge response take account of the detection accuracy and real-time performance. Compared with to feature detectors of scale invariant feature transform (SIFT), speeded up robust features (SURF), KAZE, and binary robust invariant scalable keypoints (BRISK) experiments, the FRANSFD reveals stronger robustness with five kinds of changes, and its registration speed is faster. Compared with KAZE, comprehensive robustness is increased about 5.76%, and the speed is increased about  47%.