Fast Image-Based Localization using Direct 2D-to-3D Matching

Image-based localization is an important problem in
Computer vision. Its applications include localization and
navigation for both pedestrians [22, 31, 13] and robots
[6, 5], Augmented Reality [1, 3], and the visualization of
Photo collections [26]. Image-based localization is also an
important part in the pipeline of higher-level computer vision
tasks such as semantic object annotation [9] and can
be used as an initial pose estimate to speed up large-scale
Reconstructions from Internet photo collections [27].
Traditionally, large-scale image-based localization has
Been treated as an image retrieval problem. After finding
those images in a database that are most similar to the query
image, the location of the query can be determined relative
To them [22, 31]. The huge progress achieved in the field
of image retrieval enables the use of an increasing number
of images for the representation of real world scenes
[25, 19, 20]. However, the localization accuracy obtained
this way cannot be better than the precision of the GPS
Positions available for the database images.

• www.fuvision.com
• www.fucraft.com
&bul
l; www.fuanimedia.com
•
• Contact for more information: 9003206648; 9360029290
•
• Coimbatore Address:
• No: 96,Hope College bus stop,
• Opp Krishna sweets, Peelamedu,
• Avinashi road, Coimbatore.
•
• Madurai Address:
• near BELL Hotel,
• Rajan Complex – 3rd floor
• Tamil Sangam Road,