Worldwide Pose Estimation Using 3D Point Clouds

Abstract

Abstract. We address the problem of determining where a photo was taken by estimating a full 6-DOF-plus-intrincs camera pose with respect to a large geo-registered 3D point cloud, bringing together research on image localization, landmark recognition, and 3D pose estimation. Our method scales to datasets with hundreds of thousands of images and tens of millions of 3D points through the use of two new techniques: a co-occurrence prior for RANSAC and bidirectional matching of image features with 3D points. We evaluate our method on several large data sets, and show state-of-the-art results on landmark recognition as well as the ability to locate cameras to within meters, requiring only seconds per query.

Fast and automatic city-scale environment modelling using hard and/or weak constrained bundle adjustments

“…A building front coordinate system associated to each facade of the 3D building models. 2 This coordinate system is defined as shown in Fig. 1b.…”

“…For example, Q r represents the 1 The DEM used is a geometric model where each road is presented by a plane. 2 The 3D building models used are a geometric model where each facade is presented by a plane.…”

“…Such methods guarantee the continuity of the localization even if the geolocation information is locally unavailable. For example, several methods such as [2,6,7] exploit a collection of georeferenced internet photos to create 3D landmarks databases through a SfM algorithm. However, the resulting databases are mostly focused on some tourist attractions observed from pedestrian viewpoint while the landmarks must be evenly distributed along the whole road network from a driver viewpoint to ensure an on-line and accurate localization of a car.…”

“…Nevertheless, these solutions use the geolocation provided by a standard GPS whose accuracy remains insufficient for a high quality service in urban context. To overcome this problem, vision-based solutions exploiting geo-referenced 3D landmarks [1][2][3] seem to be a promising alternative. These landmarks can be obtained through an structure from motion (SfM) process or a simultaneous localization and mapping (SLAM) algorithm.…”

Incremental image set querying based localization

“…Based on the pose of the image, high level analysis could be applied [5][6][7][8][9][10][11][12][13][14][15][16]. Generally, there are three key steps in a single image-based localization system [17,18]: (1) 2D feature extraction (e.g., SIFT [19]) from the querying image, (2) matching between these 2D features and the pre-built 3D feature point cloud, and (3) camera pose estimation by solving a perspectiven-point (PNP) problem [20][21][22]. The 3D feature point cloud is usually reconstructed from many captured images at offline stage by using a conventional 3D reconstruction algorithm [4,23].…”