3d street

Abstract

Methods and apparatus are provided for acquiring an image to be positioned with a textured 3D model and aligning the acquired image with the model while using texturing information from the model to obtain a new estimated pose. This enables the acquired images to be positioned with the 3D model using the new estimated pose. In one aspect, the aligning includes rendering the image in the 3D model at the estimated pose and then aligning the acquired image with the rendered image. In another aspect, the aligning includes projecting the acquired image onto the 3D model at the estimated pose and then aligning the projected acquired image with the textured 3D model. In yet another aspect, the aligning includes projecting the acquired images onto the textured 3D model, rendering the first image in the original 3D model and rendering the second image in the 3D model on which the acquired images are projected , and then align the first and second images.

Description

Technical field [0001] The present invention relates to a method and device for using a textured 3D model to locate captured images, especially street-level images. Background technique [0002] In the prior art, geographic information systems that provide digital maps are well known. Generally, a three-dimensional map is formed based on an image captured (or acquired) by an aircraft scanning a part of the terrain to be 3D modeled. When capturing an image, the position of the camera can be obtained by GPS, and the image is further time stamped. The aircraft is further equipped with an inertial measurement unit (IMU) so that the angle of rotation of the aircraft (which is also called roll, pitch and yaw) can be recorded. Therefore, the position and attitude of the camera are recorded for each captured image. [0003] The position and posture of the camera (called navigation data) can be considered to perform stereo processing of the captured image. The image pair including the o...

AI Technical Summary

Problems solved by technology

[0006] Although 3D modeling using aerial imagery usually produces high-quality localization, street-level 3D modeling often suffers from low-quality localization

Factors such as shadowing of GPS signals e.g. by obstructions, signal distortion and drift of IMU data in relatively variable motion of vehicles at street level deteriorate measurements at ground level

This makes the location of recorded street-level imagery imprecise

Further, the mechanical and optical properties of a given actual camera differ from those of a camera assumed to be the same, which yields erroneous measurements

A further problem is that aligning images captured at very different angles is cumbersome due to the difficulty in finding overlapping image data

Therefore, when street-level imagery is projected onto a 3D model derived from aerial imagery, there is a mismatch because the pose of the ground-level camera used to capture the street-level imagery does not match the georeferenced details of the aerial 3D model major risk of

Images