Automatic Alignment of Vector Data and Orthoimagery for The National - PowerPoint PPT Presentation

Automatic Alignment of Vector Data and Orthoimagery for The National Map Craig A. Knoblock, University of Southern Cal Cyrus Shahabi, University of Southern Cal Ching-Chien Chen, Geosemble Technologies E. Lynn Usery, US Geological Survey

Outline • Introduction & Motivation – The National Map • Our Approach to align vector and imagery – Approach overview – Improvements over our previous approach • Related Work • Conclusion and Future Work

Introduction • Geospatial data sources have become widely available • Automatically and accurately integrating and aligning two spatial datasets is a challenging problem Street maps Road network Orthoimagery ( in raster format ) ( in vector format ) ( in raster format )

Motivation: Vector and Imagery Integration � Challenges � Different projections, accuracy levels, resolutions result in spatial inconsistencies Lat / Long Lat / Long Lat / Long Lat / Long

Motivation: The National Map • The National Map is a government effort to make geospatial data available for 133 urban areas of the US for Homeland Security • Purpose is to make these integrated datasets available to government organizations to support crisis response and emergency planning, etc. • There are no automated techniques for aligning vector data with orthoimagery and this is a very labor intensive task.

Motivation: The state of the art • Traditionally, the problems of vector-imagery and map-imagery alignment have been in the domain of GIS and Computer Vision • In GIS literature – The alignments were previously performed manually • Commercial products: ESEA MapMerger ESRI ArcView; Able R2V ; Intergraph I/RASC • In Computer Vision literature – Alignment was performed automatically based on image processing techniques • Often required significant CPU time • Accuracy quite poor

Road network The example • The Data Sets (for the National Map) – USGS high resolution color imagery – Road vector data from DOT, MO USGS 0.3m/p color imagery They are misaligned, and there is no global transformation

The Vector-Imagery conflation approach Lat / Long Triangulation and Filtering Rubber-Sheeting Technique Final control points Intermediate control points Control Point Detection Lat / Long Lat / Long Lat / Long

Finding Control Points Using Localized Template Matching Learned color distribution for road/off road pixels Road classification road width and road directions Matching by Correlation road width and road directions

Filtering Control Points Using Vector Median Filter (VMF) – View the control point pair displacement as vector – Using a fixed ratio ( k %) to keep control point pairs that have similar displacement as the median one Y (meter) k% control-point vector Vector median 10 X (meter) -10 10 -10

Results: NAVSTREETS + High-res Image Original Conflated NAVSTREETS NAVSTREETS Completeness 44.9 % 74.4 % Correctness 47.9 % 85 % Positional Accuracy

Road Classification Used in Localized Template Matching • Before: Bayes classifier based on Hue component of learned road/off- road pixels • Improved: Support Vector Machine (SVM) classifier based on all color channels (R,G,B) of learned road/off-road pixels – Much fewer “false positives” and more “true positives” Road-classified pixels Road-classified pixels Original imagery based on SVM classifier based on Bayes classifier

Filtering Control Points Using Vector Median Filter (VMF) • Improved: Dynamic determine the ratio – Investigate the cluster around the Y (meter) median vector Kept control-point vector (clustered vectors – Accommodate more control point pairs around the median vector) Most of the vectors are close to the median vector. Vector median This forms a cluster around the median vector 10 X (meter) -10 10 -10

The Vector-Imagery conflation approach: Triangulation and RubberSheeting

Improved results: comparing with results based on previous technique Red lines: Original roads Yellow lines: Conflated results based on previous technique Blue lines: Conflated results based on improved technique

More improved results: comparing with results based on previous technique

Related Work • Vector to imagery conflation – Utilizing matched polygons [Hild et al. 98] – Utilizing matched lines [Filin et al. 00] – Utilizing matched junction-points [Flavie et al. 00] – All above solutions • Require lots of CPU time • Utilize vector data only for verifying detected features not for extracting features – Commercial products: ESRI AreView • Pick control points manually

Conclusion • Accomplishments Refinement of pattern recognition procedures for identifying the road intersections in the images Refinement of the filtering procedures for the ground control points Development of methods for matching across image panels Overall improvement of the accuracy of the transformed transportation data to match the images

Future Work • Address the alignment of road vector data with highways • Apply the same techniques to automatically align: – Vector Parcel Data – Hydrographic Data – Elevation Data

Automatic Alignment of Vector Data and Orthoimagery for The National - PowerPoint PPT Presentation

Automatic Alignment of Vector Data and Orthoimagery for The National Map Craig A. Knoblock, University of Southern Cal Cyrus Shahabi, University of Southern Cal Ching-Chien Chen, Geosemble Technologies E. Lynn Usery, US Geological Survey

Automatically and Accurately Conflating Road Vector Data, Street Maps and Orthoimagery

Data Driven Collaboration Data Drives Collaboration in the GIS Community Orthoimagery

National Digital Orthoimagery Program Subcommittee Charter July 20, 2010 Purpose The NDOP

Lingua-Align: An Experimental Toolbox for Automatic Tree-to-Tree Alignment

Vector Forward Mode Automatic Differentiation on SIMD/SIMT architectures Jan H uckelheim,

Using support vector machines and state-of-the-art algorithms for phonetic alignment to identify

Automatic Classification of Automatic Classification of Audio Data Audio Data Carlos H. C.

Data driven Ontology Alignment Data driven Ontology Alignment Nigam Shah nigam@stanford.edu

CS145: INTRODUCTION TO DATA MINING 7: Vector Data: K Nearest Neighbor Instructor: Yizhou Sun

CS145: INTRODUCTION TO DATA MINING 6: Vector Data: Neural Network Instructor: Yizhou Sun

RNA structure alignment by a unit-vector approach Emidio Capriotti Marc A. Marti-Renom

CS145: INTRODUCTION TO DATA MINING 5: Vector Data: Support Vector Machine Instructor: Yizhou Sun

Automatic Alignment and Annotation Projection for Literary Texts Uli Steinbach Ines Rehbein

CS145: INTRODUCTION TO DATA MINING 09: Vector Data: Clustering Basics Instructor: Yizhou Sun

Sequence Alignment (chapter 6) The biological problem l Global alignment l Local alignment l

Alignment in C Seminar Effiziente Programmierung in C Sven-Hendrik Haase Universit at

Robosat: an Open Source and efficient Semantic Segmentation Toolbox for Aerial Imagery @o_courtin

e e J ( L I T E R A T U R E He1lo. Or should I just say Hi! Great Expectations

Imagery Fictive Motion types Emanation Pattern Path Frame-relative motion

Imagery Imagery Perception-like experiences accompanying language comprehension or

Zebra Crossing Detection from Aerial Imagery Across Countries Daniel Koester, Bjrn Lunt, Rainer

Detecting and Analyzing Solar Panels in Switzerland using Aerial Imagery (SolAI) Adrian Meyer

DeepStereo: Learning to Predict New Views from the Worlds Imagery Example video Deep networks

UNDERSTANDING IMAGE QUALITY AND TRUST IN PEER-TO-PEER MARKETPLACES Xiao Ma [1] Lina Mezghani [2*]