Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Alignment of continuous video onto 3D point clouds.

Wenyi Zhao1, David Nister, Steve Hsu

  • 1Vision Technologies Lab, Sarnoff Corporation, 201 Washington Rd., Princeton, NJ 08540, USA. wzhao@sarnoff.com

IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 27, 2005
PubMed
Summary

We developed a new method to align video point clouds with 3D sensor data, improving 3D modeling for complex scenes like forests. This approach enhances 3D model texturing and offers practical opportunities for creating detailed 3D models.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Underwater Image Enhancement via Intelligent Optimized Multi-Exposure Image Fusion.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same author

SEMA6A inhibits tumor progression and boosts anti-tumor immunity via blocking the ISG15/TGFβ axis in colorectal cancer.

Molecular biomedicine·2026
Same author

Genome-wide identification of WRKY members in elephant grass (<i>Cenchrus purpureus</i>) and its expression profiling under aluminum stress.

PeerJ·2026
Same author

Designing T-cell Engagers: Trade-offs Between Activity and Safety.

Protein & cell·2026
Same author

TD loss: Taylor expansion of Dice loss for robust medical image segmentation.

Medical image analysis·2026
Same author

Identification of PLXNC1 as a novel biomarker for consensus molecular subtype 4 in colorectal cancer.

Genes & diseases·2026

Area of Science:

  • Computer Vision
  • 3D Reconstruction
  • Geospatial Information Science

Background:

  • Current 3D modeling often aligns 2D images to existing 3D models.
  • This method struggles with complex, unstructured environments like forests.
  • Aligning directly with 3D sensor data is challenging.

Purpose of the Study:

  • To present a general framework for aligning continuous video onto 3D sensor data.
  • To enable accurate 3D modeling of complex scenes by aligning point clouds.
  • To introduce a novel modeling-through-registration approach.

Main Methods:

  • Computed 3D point clouds from video using motion stereo and camera pose estimation.
  • Aligned video-derived point clouds directly onto 3D sensor point clouds.

Related Experiment Videos

  • Developed a novel modeling-through-registration approach to fuse data.

    • Main Results:

    • Demonstrated advantages of the proposed registration algorithm for texturing models in large-scale semi-urban environments.
    • Showcased the ability to align video before 3D model construction.
    • Initial experiments with real data illustrated the approach's potential.

    Conclusions:

    • The proposed point cloud alignment framework effectively handles complex scenes difficult to model with traditional methods.
    • This approach offers new practical opportunities for 3D modeling by enabling pre-model alignment.
    • The novel fusion of 3D sensor and video data shows significant potential for enhanced 3D reconstruction.