Computer Graphics and Visualization Lab

Department of Computer Science at Purdue University

ModelCamera

Real-Time Modeling from Dense Color and Sparse Depth
Summary

The ModelCamera is a fast, easy to use, and inexpensive 3D scene modeling system. The ModelCamera acquires dense color (720x480 video frames) augmented with sparse depth (7x7 to 11x11 depth samples). The frames are registered and merged into an evolving model at the rate of five frames per second. The model is displayed continually to provide immediate operator feedback.

 

SDDV Modeling

We propose an automated modeling approach based on sampling the scene sparsely from a dense set of acquisition viewpoints. The sparse data quickly accumulates to generate models with good scene coverage. The sparse depth is acquired efficiently and robustly which enables an interactive, operator-in-the-loop, pipeline.

We implemented this sparse depth/dense viewpoints (SDDV) approach in a modeling system. The system acquires scenes with complex geometry and complex reflective properties from thousands of viewpoints in minutes. The resulting model has a compact memory footprint and it supports photorealistic rendering at interactive rates. The system is robust, yet it does not require displacing scene objects or altering scene lighting conditions.

 

Typical acquisition path with 3684 viewpoints.

Acquisition system

Rendering of a complex plant model acquired in 10 minutes with our system

 

Paper

 

M. Mudure, V. Popescu, “1001 Acquisition Viewpoints – Efficient and Versatile View-Dependent Modeling of Real-World Scenes”, submitted to Eurographics 2007   PPT(9.5 MB)

To see a gallery of models acquired with our system, click here

 

Large Scale Modeling

 We are currently working to extend our modeling system to scanning room-size environments using the SDDV approach. Our acquisition device already has the capability to scan large environments. For an effective implementation of the SDDV approach in the inside-looking-out-case, our view-dependent modeling system must be disocclusion error resistant, must  have bounded cost and must support incremental modeling.

For our desired view-dependent modeling we propose to use occlusion cameras, a class of non-pinhole cameras that produce disocclusion-resistant reference images. We will develop a panoramic occlusion camera that will provide a disocclusion error-resistant model of an entire room. We will investigate rasterization for our camera model, such that we can take advantage of the GPU and render in a feed-forward fashion

We will demonstrate the efficiency and effectiveness of the proposed system by acquiring, in less than a week, a photorealistic model of the interior of a 100-room building on our campus.

 

 

 

 

Publications

 

People

 

 

 

[Return to Main Page]