Towards Fast Image-Based Localization on a City-Scale

Abstract

Abstract. Recent developments in Structure-from-Motion approaches allow the reconstructions of large parts of urban scenes. The available models can in turn be used for accurate image-based localization via pose estimation from 2D-to-3D correspondences. In this paper, we analyze a recently proposed localization method that achieves state-of-the-art localization performance using a visual vocabulary quantization for efficient 2D-to-3D correspondence search. We show that using only a subset of the original models allows the method to achieve a similar localization performance. While this gain can come at additional computational cost depending on the dataset, the reduced model requires significantly less memory, allowing the method to handle even larger datasets. We study how the size of the subset, as well as the quantization, affect both the search for matches and the time needed by RANSAC for pose estimation.

Geometric-visual descriptor for improved image based localization

“…In our case the query is given by a stereo camera so it's impossible to test on the classical datasets. Therefore, we select from the state of the art two methods who use visual vocabulary to estimate the pose ( [16] , [15]) and we test them on the databases presented on the figure 3. For the query images we keep only the features which are triangulated after stereo matching.…”

“…On the other hand, direct methods proceed by matching descriptors between the input query and the 3D model built by a SfM or SLAM algorithm. [8] Sattler et al [16] propose a direct matching method based on visual words to establish the correspondence between the query and the 3D scene. Because an image contains fewer primitives than the whole 3D model, Sattler et al [15] improve their framework by combining the 3Dto 2D and 2D to 3D matching strategies to increase the number of correspondences.…”

“…As proposed by [6] to evaluate the results, we consider a query as well registered in the model only if we obtain more than 12 inliers after RANSAC. Our proposed geometric descriptor can be used with most of the state of the art methods and in the experimental part we show results obtained with three different methods: [16], [17] and our own indirect method.…”

Camera Pose Voting for Large-Scale Image-Based Localization

“…While 2D-to-3D search is inherently more reliable than 3D-to-2D matching [21], stateof-the-art approaches use the latter to recover correspondences missed or rejected during the former [7,16,22]. This enables them to better counter the problem that the ratio test rejects more and more correct matches for larger datasets due to the increased descriptor space density [16,23]. Recently, alternatives [16,27] to aggressive outlier filtering during the matching stage have been proposed.…”

CNN Models for Accurate Pose Localization

“…Beginning with direct methods, these techniques rely on correspondences established by directly matching query features with points from the 3D point cloud. Sattler et al [25], [27] [26] have enhanced matching strategies, including visual vocabulary-based approaches and bag-of-visual-words methods. Indirect methods, on the other hand, focus on finding nearest neighbors using Content-Based Image Retrieval (CBIR) algorithms and then matching query features with 3D point cloud data from retrieved key images.…”