Passive Recovery of Scene Geometry for an Unmanned Ground Vehicle.

Abstract

The goal of this project was to develop techniques for constructing three-dimensional descriptions of outdoor scenes to support the navigational needs of an Unmanned Ground Vehicle (UGV), operating both during the day and at night. In this report we describe our progress in four areas: Stereo Evaluation - We performed an in-depth evaluation of three representative stereo techniques by analyzing the results on 49 stereo pairs. Scene Sketch - We developed a high-level representation of an outdoor scene, which we call a 'scene sketch'., that can describe the semantic, as well as geometric properties of a three-dimensional scene. Spatiotemporal Filtering - We developed a technique for increasing the resolution and robustness of passive range sensors by integrating stereo and motion analysis. FLIR Stereo - We demonstrated the effectiveness of stereo analysis applied to infrared data, which makes passive night driving possible. (MM)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 1995
Accession Number
ADA303483

Entities

People

  • Martin A. Fischler
  • Robert C. Bolles

Organizations

  • SRI International

Tags

Communities of Interest

  • Autonomy
  • Sensors

DTIC Thesaurus Topics

  • Filtration
  • Geometry
  • Ground Vehicles
  • Recovery
  • Test And Evaluation
  • Three Dimensional
  • Unmanned
  • Unmanned Ground Systems
  • Unmanned Ground Vehicles
  • Vehicles

Readers

  • Geodesy
  • Image Processing and Computer Vision.
  • Neural Network Machine Learning.

Technology Areas

  • Autonomy