Analysis of feature detector and descriptor combinations with a localization experiment for various performance metrics

The purpose of this study is to give a detailed performance comparison about the feature detector and descriptor methods, particularly when their various combinations are used for image matching. As the case study, the localization experiments of a mobile robot in an indoor environment are given. In these experiments, 3090 query images and 127 dataset images are used. This study includes five methods for feature detectors such as features from accelerated segment test (FAST), oriented FAST and rotated binary robust independent elementary features (BRIEF) (ORB), speeded-up robust features (SURF), scale invariant feature transform (SIFT), binary robust invariant scalable keypoints (BRISK), and five other methods for feature descriptors which are BRIEF, BRISK, SIFT, SURF, and ORB. These methods are used in 23 different combinations and it was possible to obtain meaningful and consistent comparison results using some performance criteria defined in this study. All of these methods are used independently and separately from each other as being feature detector or descriptor. The performance analysis shows the discriminative power of various combinations of detector and descriptor methods. The analysis is completed using five parameters such as (i) accuracy, (ii) time, (iii) angle difference between keypoints, (iv) number of correct matches, and (v) distance between correctly matched keypoints. In a range of 60°, covering five rotational pose points for our system, “FAST-SURF” combination gave the best results with the lowest distance and angle difference values and highest number of matched keypoints. The combination “SIFT-SURF” is obtained as the most accurate combination with 98.41% of correct classification rate. The fastest algorithm is achieved with “ORB-BRIEF” combination with a total running time 21303.30 seconds in order to match 560 images captured during the motion with 127 dataset images.