An Improved Randomized Local Binary Features for Keypoints Recognition.

Jinming Zhang,Zuren Feng,Gang Li,Jinpeng Zhang

doi:10.3390/s18061937

Abstract

In this paper, we carry out researches on randomized local binary features. Randomized local binary features have been used in many methods like RandomForests, RandomFerns, BRIEF, ORB and AKAZE to matching keypoints. However, in those existing methods, the randomness of feature operators only reflects in sampling position. In this paper, we find the quality of the binary feature space can be greatly improved by increasing the randomness of the basic sampling operator. The key idea of our method is to use a Randomized Intensity Difference operator (we call it RID operator) as a basic sampling operator to observe image patches. The randomness of RID operators are reflected in five aspects: grids, position, aperture, weights and channels. Comparing with the traditional incompletely randomized binary features (we call them RIT features), a completely randomized sampling manner can generate higher quality binary feature space. The RID operator can be used on both gray and color images. We embed different kinds of RID operators into RandomFerns and RandomForests classifiers to test their recognition rate on both image and video datasets. The experiment results show the excellent quality of our feature method. We also propose the evaluation criteria for robustness and distinctiveness to observe the effects of randomization on binary feature space.

Highlights

We propose using more general randomized intensity difference sampling operator to construct binary feature space for keypoints recognition
After reviewing the existing randomized local binary feature methods, we find that the common ground of these methods is that they all use nonrandom-weighted aperture-fixed point-pairs single-channel basic sampling operators to construct binary feature space
We first detect 300 keypoints on each reference image and randomly generate 1000 image patch samples for each keypoint, evaluate the recognition rate of RandomFerns classifier trained upon the feature space with specified parameter settings

Summary

Introduction

AKAZE [14] that have achieved some success These methods usually need careful preprocessing and complicated hand-designed sampling patterns to resist deformations like rotation, zooming and view-point changes. Boltzmann machine [19] are proposed to recognize keypoints. These classifier-based methods focus their attention on classifier improvement but ignore improving the quality of binary feature space. The basic sampling operator they used to construct feature space are nonrandom-weighted aperture-fixed two-grids single-channel intensity difference operators. Their randomness only reflects in the distribution of sampling position within an size-fixed image patch

Methods

Results

Discussion

Conclusion