Abstract

This thesis addresses the topic of semi-automated extraction of urban road networks from high-resolution satellite imagery. Research on this topic is mainly motivated by the use geographic information systems in transportation (GIS-T), and the need for reliable data acquisition methods and to update GIS-T databases. To this end, 1-m spatial resolution IKONOS imagery provides a new data source to collect the spatial models of citywide road networks. In this thesis, a novel methodology of a semi-automated road extraction using high-resolution satellite imagery over urban areas is developed. The main objective of this research is to extract urban road networks from a single IKONOS image. To detect the road features from a highly complex scene, a multiscale analysis of the optimal image was performed. To extract roads and their networks, the knowledge of road geometry is exploited in an interactive environment. The key advantage of the developed method is the full employment of a human and a computer's abilities for fast and precise road extraction from high-resolution satellite imagery. The results show that the presented method enables reliable road extraction over urban areas. The potential applications exemplified in case studies indicate that the high-resolution satellite imagery offers an efficient and precise source for geographic and transportation databases. Based on this research, the limitations and future work for the prototype system are discussed.

Highlights

  • 1.1 MotivationFrom a practical point of view, research on automatic road extraction in urban areas is mainly motivated by the use of geographic information systems in transportation (GIS-T) and the need for data acquisition and update for GIS-T databases

  • By employing change detection technique provided by ERDAS Imagine 8.5, the difference between the Principal components analysis (PCA) image and Decorrelation stretch (DS)-PCA image can be visually examined

  • The green areas are where the digital numbers (DN) in DSPCA image is larger than the DN in PCA image; the red areas are where the DN in DS-PCA image is smaller than the DN in PCA image; black areas are of no change

Read more

Summary

Motivation

From a practical point of view, research on automatic road extraction in urban areas is mainly motivated by the use of geographic information systems in transportation (GIS-T) and the need for data acquisition and update for GIS-T databases. The new generation o f high-resolution imaging satellites (e.g., IKONOS, QuickBird) have presented more details and precision on urban man-made features (e.g., roads, buildings). Many existing approaches, working on highresolution satellite imagery, are based on the algorithms developed for lower resolution satellite imagery following a downgrading preprocessing to reduce the image spatial resolution. This compromises the accuracy of the results and the full exploitation o f high-resolution data. Extracting roads and their networks from high-resolution satellite imagery and taking advantage of most data spectra should be a new research topic in remote sensing

Problem Statement
Research Objectives
Thesis Organization
LITERATURE REVIEW
Multiscale Edge Detection
Parametric Feature Detection
Automatic Methods
Use of Single Image
Use of Multisource Data
Use of Multiscale Images
Use of Combination of Multisource Datasets and Multiscale Images
Active Tracing
Snakes
Template Matching
Classification
Summary of Existing Methods
METHODOLOGY
Input Data Description
General Strategy for Semi-automated Road Extraction
Interpretation of Image Data
Standard Deviation Stretch
D Stretched data file values 255 standard Deviadon Stretch
D N Stretched
Time-Scale Analysis
Geometric Feature Detection
Extraction of Simple Horizontal Circular Curves
Extraction o f Reverse and Compound Curves
Extraction of Spiral Curves
Extracting Second Roadside
Connection to Desktop GIS Environment
Study Areas
Decorrelation Stretch
Principal Components Analysis
Results and Discussion
Wavelet Edge Detection
Extraction of Simple Circular Curves
Extraction of Road Networks
Performance Evaluation
CONCLUSIONS AND RECOMMENDATIONS
Summary
Limitations
Recommendations for Further Research
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call