Topic
Scale-invariant feature transform
About: Scale-invariant feature transform is a research topic. Over the lifetime, 6779 publications have been published within this topic receiving 197891 citations. The topic is also known as: SIFT.
Papers published on a yearly basis
Papers
More filters
TL;DR: In this paper , the authors rely on Python and Computer vision to study the detection efficiency of various corner detection methods in different environments and find that the ORB detector can be competent in most situations and is currently the fastest and stable feature point detection algorithm.
Abstract: In-depth research on feature detection technology affects people's modern life. Modern artificial intelligence can act as the eyes of human beings and efficiently filter out effective information from complex pictures. Corner detection has now evolved into a tool for efficient image scanning. People's increasingly stringent requirements for image processing continue to promote the birth of new technologies. Corner detection methods have been improved and perfected, and have experienced detectors such as Harris, FAST, Scalriant Feature Transform (SIFT), Speeded Up Robust Feature (SURF), Binary Robust Invariant Scalable Keypoints (BRISK), KAZE, AKAZE and Oriented FAST and Rotated BRIEF (ORB). In the face of many mainstream detectors, the main research purpose of this paper is to rely on Python and Computer vision to study the detection efficiency of various detectors in different environments. In this experiment, thirteen groups of pictures were selected, and after being flipped, complicated, and blurred respectively, they were detected by different detectors to obtain the results. Finally, by comparing the feature detection points, detection time and other factors, this study found that the ORB detector can be competent in most situations and is currently the fastest and stable feature point detection and extraction algorithm. On the other hand, the BRISK detector can handle highly blurred images more efficiently.
21 Aug 2020-The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
TL;DR: The implication of the findings of this research is that the choice of feature descriptors must be informed by the imaging conditions of the image registration analysts, as different descriptors perform differently under different conditions.
Abstract: . Automatic detection and extraction of corresponding features is very crucial in the development of an automatic image registration algorithm. Different feature descriptors have been developed and implemented in image registration and other disciplines. These descriptors affect the speed of feature extraction and the measure of extracted conjugate features, which affects the processing speed and overall accuracy of the registration scheme. This article is aimed at reviewing the performance of most-widely implemented feature descriptors in an automatic image registration scheme. Ten (10) descriptors were selected and analysed under seven (7) conditions viz: Invariance to rotation, scale and zoom, their robustness, repeatability, localization and efficiency using UAV acquired images. The analysis shows that though four (4) descriptors performed better than the other Six (6), no single feature descriptor can be affirmed to be the best, as different descriptors perform differently under different conditions. The Modified Harris and Stephen Corner Detector (MHCD) proved to be invariant to scale and zoom while it is excellent in robustness, repeatability, localization and efficiency, but it is variant to rotation. Also, the Scale Invariant feature Transform (SIFT), Speeded Up Robust Features (SURF) and the Maximally Stable Extremal Region (MSER) algorithms proved to be invariant to scale, zoom and rotation, and very good in terms of repeatability, localization and efficiency, though MSER proved to be not as robust as SIFT and SURF. The implication of the findings of this research is that the choice of feature descriptors must be informed by the imaging conditions of the image registration analysts.
TL;DR: Wang et al. as mentioned in this paper proposed an accurate and robust image matching approach using fewer training data in an end-to-end manner, which could be used to estimate the pose error.
Abstract: As the fundamental problem in the computer vision area, image matching has wide applications in pose estimation, 3D reconstruction, image retrieval, etc. Suffering from the influence of external factors, the process of image matching using classical local detectors, e.g., scale-invariant feature transform (SIFT), and the outlier filtering approaches, e.g., Random sample consensus (RANSAC), show high computation speed and pool robustness under changing illumination and viewpoints conditions, while image matching approaches with deep learning strategy (such as HardNet, OANet) display reliable achievements in large-scale datasets with challenging scenes. However, the past learning-based approaches are limited to the distinction and quality of the dataset and the training strategy in the image-matching approaches. As an extension of the previous conference paper, this paper proposes an accurate and robust image matching approach using fewer training data in an end-to-end manner, which could be used to estimate the pose error This research first proposes a novel dataset cleaning and construction strategy to eliminate the noise and improve the training efficiency; Secondly, a novel loss named quadratic hinge triplet loss (QHT) is proposed to gather more effective and stable feature matching; Thirdly, in the outlier filtering process, the stricter OANet and bundle adjustment are applied for judging samples by adding the epipolar distance constraint and triangulation constraint to generate more outstanding matches; Finally, to recall the matching pairs, dynamic guided matching is used and then submit the inliers after the PyRANSAC process. Multiple evaluation metrics are used and reported in the 1st place in the Track1 of CVPR Image-Matching Challenge Workshop. The results show that the proposed method has advanced performance in large-scale and challenging Phototourism benchmark.
TL;DR: Chen et al. as discussed by the authors proposed an unsupervised V-SLAM light enhancement network (UVLE-Net) to enhance the quality of low-light images, which significantly increased the brightness and contrast of the images, allowing for easy detection of feature points.
Abstract: Low-light Image Enhancement for Construction Robot Simultaneous Localization and Mapping Xinyu Chen, Yantao Yu Pages 116-123 (2023 Proceedings of the 40th ISARC, Chennai, India, ISBN 978-0-6458322-0-4, ISSN 2413-5844) Abstract: Visual Simultaneous Localization and Mapping (V-SLAM) is commonly employed as a method in construction robots due to its efficient and cost-effective nature for information acquisition. However, detection and positioning using V-SLAM face significant challenges in low-light construction scenes, such as underground garages or dim indoor locations, where detecting enough valid feature points can be difficult leading to navigation failure. In this study, we propose an Unsupervised V-SLAM Light Enhancement Network (UVLE-Net) to address this issue by enhancing the quality of low-light images. Subsequently, we incorporated the robust Shi-Tomasi method in ORB-SLAM2 to detect feature points and utilized the sparse optical flow algorithm to track them. The use of UVLE-Net significantly increased the brightness and contrast of the images, allowing for easy detection of feature points. Furthermore, the Shi-Tomasi method and sparse optical flow algorithm were found to be effective in improving the ability of feature point extraction and tracking in low-light conditions. To validate the robustness and superiority of our approach, we carried out comparison experiments with other enhancement techniques, using publicly available and real-world construction datasets. Keywords: Low-light enhancement, Simultaneous Localization and Mapping, Retinex, construction robots DOI: https://doi.org/10.22260/ISARC2023/0018 Download fulltext Download BibTex Download Endnote (RIS) TeX Import to Mendeley
TL;DR: The potential for the computer vision models in combination with 3D imaging to be free of bias and to detect potential geoengineered formations in the future was showed and positive results showed for these algorithms about the similarities of both features.
Abstract: Ahuna Mons is a 4 km particular geologic feature on the surface of Ceres, of possibly cryovolcanic origin. The special characteristics of Ahuna Mons are also interesting in regard of its surrounding area, especially for the big crater beside it. This crater possesses similarities with Ahuna Mons including diameter, age, morphology, etc. Under the cognitive psychology perspective and using current computer vision models, we analyzed these two features on Ceres for comparison and pattern-recognition similarities. Speeded up robust features (SURF), oriented features from accelerated segment test (FAST), rotated binary robust independent elementary features (BRIEF), Canny edge detector, and scale invariant feature transform (SIFT) algorithms were employed as feature-detection algorithms, avoiding human cognitive bias. The 3D analysis of images of both features’ (Ahuna Mons and Crater B) characteristics is discussed. Results showed positive results for these algorithms about the similarities of both features. Canny edge resulted as the most efficient algorithm. The 3D objects of Ahuna Mons and Crater B showed good-fitting results. Discussion is provided about the results of this computer-vision-techniques experiment for Ahuna Mons. Results showed the potential for the computer vision models in combination with 3D imaging to be free of bias and to detect potential geoengineered formations in the future. This study also brings forward the potential problem of both human and cognitive bias in artificial-intelligence-based models and the risks for the task of searching for technosignatures.