HOME PUBLICATIONS SMART SURVEILLANCE 3D FACE RECOTNIGION 3D OBJECT RECOGNITION 3D MODELING INTERNET KEY EXCHANGE PICTURES

Automatic Correspondence and Registration of Range Images for 3D Modeling


Automatic registration of range images is a fundamental problem in 3D modeling of free-from objects. Registration is usually performed in two steps. First, the range images (views) are coarsely registered and then this registration is iteratively refined with an algorithm like ICP for example. Automatic coarse registration can be performed either through system calibration or through feature matching (also known as automatic correspondence establishment). Our contribution is in the area of automatic correspondence establishment. The following paper describes our automatic pairwise correspondence algorithm.

Ajmal S. Mian, M. Bennamoun and R. Owens, "A Novel Representation and Feature Matching Algorithm for Automatic Pairwise Registration of Range Images", International Journal of Computer Vision (IJCV), vol. 66(1), pp. 19--40, 2006. [pdf] Copyright © Springer Verlag.

We also extended our above work to multiview correspondence which we define as a one-to-many correspondence (or matching) approach whereby one view is simultaneously matched with multiple views. Our multiview correspondence algorithm can automatically establish correspondences between N unordered views of an object with O(N) complexity. Unordered means that the order in which the views were acquired is unknown and hence there is no prior knowledge about which view pairs overlap. The following paper describes our multiview correspondence algorithm.

Ajmal Mian, M. Bennamoun and R. Owens, "3D Model-based Object Recognition and Segmentation in Cluttered Scenes", IEEE Transactions in Pattern Analysis and Machine Intelligence (IEEE TPAMI), vol. 28(10), pp. 1584--1601, 2006. [pdfCopyright © IEEE.

The data we used for our 3D modeling experiments is available for download below. Please cite the above two papers if you use the data.

3D Models


The 2.5D views of the above models can be downloaded as zip files in PLY format. Click on the object name to download its complete 3D model and on v1 (for example) in the same row to download its view 1.
 
Chef v1 v2 v3 v4 v5 v6 v7 v8 v9 v10 v11 v12 v13 v14 v15 v16 v17 v18 v19 v20 v21 v22
Parasaurolophus v1 v2 v3 v4 v5 v6 v8 v9 v10 v11 v12 v13 v14 v15 v16
T-rex v1 v2 v3 v4 v5 v6 v7 v8 v9 v10 v11 v12 v13 v14 v15 v16 v17 v18 v19 v20 v21
Chicken v1 v2 v3 v4 v5 v6 v7 v8 v9 v10 v11 v12 v13 v14 v15 v16

Iterative Closest Point (ICP) Algorithm (Matlab Code)
icp.m
icp2.m