skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: High-resolution Urban Image Classification Using Extended Features

Abstract

High-resolution image classification poses several challenges because the typical object size is much larger than the pixel resolution. Any given pixel (spectral features at that location) by itself is not a good indicator of the object it belongs to without looking at the broader spatial footprint. Therefore most modern machine learning approaches that are based on per-pixel spectral features are not very effective in high- resolution urban image classification. One way to overcome this problem is to extract features that exploit spatial contextual information. In this study, we evaluated several features in- cluding edge density, texture, and morphology. Several machine learning schemes were tested on the features extracted from a very high-resolution remote sensing image and results were presented.

Authors:
 [1]
  1. ORNL
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
Sponsoring Org.:
Work for Others (WFO)
OSTI Identifier:
1050899
DOE Contract Number:
DE-AC05-00OR22725
Resource Type:
Conference
Resource Relation:
Conference: ICDM Workshop on Spatial and Spatiotemporal Data Mining (SSTDM), Vancouver, Canada, 20111211, 20111211
Country of Publication:
United States
Language:
English
Subject:
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE; CLASSIFICATION; LEARNING; MINING; MORPHOLOGY; REMOTE SENSING; RESOLUTION; TEXTURE; classification; extended image features

Citation Formats

Vatsavai, Raju. High-resolution Urban Image Classification Using Extended Features. United States: N. p., 2011. Web.
Vatsavai, Raju. High-resolution Urban Image Classification Using Extended Features. United States.
Vatsavai, Raju. 2011. "High-resolution Urban Image Classification Using Extended Features". United States. doi:.
@article{osti_1050899,
title = {High-resolution Urban Image Classification Using Extended Features},
author = {Vatsavai, Raju},
abstractNote = {High-resolution image classification poses several challenges because the typical object size is much larger than the pixel resolution. Any given pixel (spectral features at that location) by itself is not a good indicator of the object it belongs to without looking at the broader spatial footprint. Therefore most modern machine learning approaches that are based on per-pixel spectral features are not very effective in high- resolution urban image classification. One way to overcome this problem is to extract features that exploit spatial contextual information. In this study, we evaluated several features in- cluding edge density, texture, and morphology. Several machine learning schemes were tested on the features extracted from a very high-resolution remote sensing image and results were presented.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2011,
month = 1
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • The proliferation of several machine learning approaches makes it difficult to identify a suitable classification technique for analyzing high-resolution remote sensing images. In this study, ten classification techniques were compared from five broad machine learning categories. Surprisingly, the performance of simple statistical classification schemes like maximum likelihood and Logistic regression over complex and recent techniques is very close. Given that these two classifiers require little input from the user, they should still be considered for most classification tasks. Multiple classifier systems is a good choice if the resources permit.
  • An operational monitoring application is described for providing information on vegetation composition and its change over time in a wetland restoration site. High spatial resolution multispectral image data are acquired on near anniversary dates over the Sweetwater Marsh in San Diego County Kinematic GPS data and an orthorectified base image provide the basis for rectifying, registering and mosaicing the images to produce seamless mosaics for the restoration site. An iterative unsupervised classification routine was then applied to each mosaic to identify marsh cover types. The same marsh cover types were defiried for each year`s image mosaic with individual class accuraciesmore » ranging from 30% to 100%. The marsh cover classification maps then provided a basis for implementing a post-classification comparison change detection strategy. A change image and summary statistics were produced where each pixel was assigned to a unique class based on its nosh cover type at date I and then at date 2. The post-classification comparison enabled areas of change and no-change to be clearly identified, along with the nature of change. The cross-tabulation of time sequential maps of marsh cover types indicate areas of stable habitat and areas changing towards or away from restoration objectives.« less
  • The advent of high spatial resolution satellite imagery like Quick Bird (0.6 meter) and IKONOS (1 meter) has provided a new data source for high resolution urban land cover mapping. Extracting accurate urban regions from high resolution images has many applications and is essential to the population mapping efforts of Oak Ridge National Laboratory's (ORNL) LandScan population distribution program. This paper discusses an automated parallel algorithm that has been implemented on a high performance computing environment to extract urban regions from high resolution images using texture and spectral features
  • Numerical modeling of air flow and pollutant dispersion around buildings in the urban environment is a challenging task due to the geometrical variations of buildings and the extremely complex flow created by such surface-mounted obstacles. Building-scale air flows inevitably involve flow impingement, stagnation, separation, a multiple vortex system, and jetting effects in street canyons. Lawrence Livermore National Laboratory (LLNL) and Los Alamos National Laboratory (LANL) have developed two complementary, robust computational fluid dynamics (CFD) models, FEM3MP by LLNL and HIGRAD by LANL, for such purposes. Our primary goal is to support emergency response planning, vulnerability analysis, and development of mitigationmore » strategies for chem-bio agents released in the urban environment. Model validation is vitally important in establishing the credibility of CFD models. We have, in the past, performed model validation studies involving simpler geometries, such as flow and dispersion past a cubical building [1] and flow around a 2-D building array [2]. In this study, wind tunnel data for a 7 x 11 array of cubical buildings [3] are used to further validate our models.« less
  • Problems of resolving high Reynolds number flow features occurring in aircraft flow fields are discussed. The nature of different features is characterized to provide guidance on how best to mesh them, both from local and global perspectives. An overlying mesh approach to their resolution is proposed and is illustrated by consideration of the shear layer surrounding a jet emerging from a powered nacelle.