In this Page...
For a solution to the "Vision Problem" ... For earlier results, published references are given." For recent results, examples are shown."
Low-Level Feature Extraction:
Useful to extract cues from non-panchromatic images such as SAR, IFSAR, HYDICE, DEMs, LIDAR and HalfTones.
- Laplacian-of-Gaussian filtering
- Sub-pixel precision zero-crossings
- Feature fusion for macro edge detection
- Sub-Pixel precision line matching
References:- Fast Convolutions with Laplacian-of-Gaussian Masks. Chen, Huertas, Medioni, PAMI.
- Edge Detection with Sub-Pixel Precision. Huertas, Medioni. PAMI
- Automatic Registration of Color Separation Film. Huertas, Medioni, Wilson. Machine Vision and Applications
Airport Runway Detection:
- General techniques for Linear Features Extraction.
- Detection and description of runways and taxyways from monocular aerial images.
- Automated 3-D Site model construction
Examples:- Boston LOGAN: image and runways... an example of fully automated detection.
- New York JFK: Markings detected to verify runway hypotheses. Note that center "runway" is not a runway now, thus not reported as such.
References:- Runways: Huertas, Cole, Nevatia. CGIP, IUW
- Taxiways: Huertas, Nevatia. IUW
Harbor Pier Detection:
- General techniques for Perceptual Grouping.
- Detection and description of Piers in harbors from monocular aerial images.
- Automated 3-D Site model construction
Example:- Navy facility, San Diego: Image, water boundary, seeds, pier seeds, pier groups, piers
References:- Grouping Fields for Perceptual Organization: A. Huertas, DARPA 1987 IUW
Evidence of Construction:
- General techniques for Perceptual Grouping.
- Detection and description of Foudation and Frame Stages of construction of buildings from monocular aerial images.
- Automated 3-D Site model construction
Example:- Los Angeles area: Image & detected site, proximity groups... an example of automated detection.
References:- Grouping Startegies: Huertas IUW
Building Detection:
- General techniques for automated map-making and photo-interpretation.
- Detection and description of buildings from monocular aerial images.
- Detection and description of buildings from multiple aerial images.
- Automated 3-D Site model construction
- Pre-Assisted Systems
- Post-Assisted Systems
- Learning Systems
- User Interaction
- MVS (Automatic Multi-View System)
- MVS (Interactive Multi-View System)
- ABERS (Automatic Multi-View System)
Example:- 20-sec video of automated and post-assisted detection (10 mouse clicks required).
References:- Monocular: Lin, Huertas, Nevatia. CVPR86,ASCONA95,ASCONA97,CGVIU
- Stereo: Mohan, Chung, Nevatia. PAMI,CVPR
- Multiview: Noronha, Kim, Huertas, Nevatia. CVPR, IUW
- Multisensor: Huertas, Kim, Nevatia. IUW
Change Detection:
- Automated site model validation
- Change Detection
- Model Updating (See Building Detection above)
- Site Monitoring (See Aicraft detection below)
Example:- 15-sec video of model validation and change detection and reporting.
References:- Huertas, Nevatia. ASCONA95, ICCV98
Aircraft Detection:
- Automated Presence Verification
- Aircraft Detection
- Site Monitoring
Examples:- C-130 Hercules: Edges, model and match
- C-130 Hercules: Image, edges, model, model match and edge match
- C-130 Hercules: Image, edges, model, model match and edge match
- DeHavilland: Image, edges, model, model match and edge match
- Missing aircraft: Edges, poor match
References:- Model-Based Aircraft Recognition in Perspective Aerial Imagery. Huertas, Mourani, Medioni. SCV95(371-376).
Sensor Fusion and Information Integration:
- Extraction of cues for focus of attention
- Use DEM, IFSAR, HYDICE, SAR, LIDAR, IR, for Man Made Structures
- Main sensor is Electro Optical (EO). Suitable to Geometric Analysis
- Use cues at all levels of processing (low level filtering, high level reasoning)
- Experiments with DEM, DTE, DSM (Digital Elevation)
- Experiments with IFSAR (Interferometric SAR)
- Experiments with HYDICE (Hyperspectral)
- Experiments with LIDAR (Laser Ranging)
- Ft Hood Example: EO only, EO+DEM, EO+HYDICE
Appearance Learning:
- Learn appearance of automatically detected objects
- Look for weaker objects having similar appearance
- Appearance in terms of first order pixel statistics
- Data can be anything: panchromatic, color, radar, depth, HYDICE, IR, IFSAR
- Experiments using panchromatic images
Recent Papers:
"Multisensor Integration for Building Modeling"With Z. Kim and R. Nevatia. CVPR - Hilton Head Island, SC, June 2000.
"Detecting Changes in Aerial Views of Man-Made Structures"With R. Nevatia. Image Vision and Computing Journal, May 2000.
"The MURI Project for Rapid Feature Extraction in Urban Areas"With Z. Kim and R. Nevatia. ISPRS - Munich, Germany, Sept. 1999.
"Use of Hyperspectral Data with Intensity Images for Automatic Building Modeling"with D. Landgrebe and R. Nevatia. 2nd International Conf. on Information Fusion.
Mountain View, California, July, 1999.
"Use of IFSAR with Intensity Images for Automatic Building Modeling"with Z. Kim and R. Nevatia. Submitted to ICCV 99, Greece.
"Model-Based Building Detection and Description: 1997-98"R. Nevatia and A. Huertas. Principal Investigator Overview, APGD Proyect.
DARPA Image Understanding Workshop 1998, Monterey, California.
"Use of Cues from Range Data in Building Modeling"A. Huertas and R. Nevatia, APGD Proyect, DEM, IFSAR, IFSARE, SAR cueing
DARPA Image Understanding Workshop 1998, Monterey, California.
"Use of HYDICE Cues in Building Modeling"A. Huertas and R. Nevatia, MURI Project, DEM, HYDICE cueing
MURI Annual Review Research Report 1998, Los Angeles, California.
"Detecting Changes in Aerial Views of Man-Made Structures,"A. Huertas and R. Nevatia, Matching, Change Detection.
IEEE Proceeedings ICCV98, India 1998.
"A System for Building Detection from Aerial Images,"A. Lin, A. Huertas & R. Nevatia. Monocular Building Detection
Ascona Workshop II, Ascona Switzewrland, April 1997.
"Detection of Buildings from Monocular Images,"A. Lin, A. Huertas & R. Nevatia, Monocular Building Detection
Ascona Workshop, Ascona, Switzerland, March 1995.
Maintained by Andres Huertas, (huertas@iris.usc.edu).